Nondestructive Evaluation of Composite Railroad Ties and Bridge Components Using Infrared Thermography

Transcription

1 Nondestructive Evaluation of Composite Railroad Ties and Bridge Components Using Infrared Thermography Srinivas Majiga Problem Report submitted to the College of Engineering and Mineral Resources at West Virginia University in partial fulfillment of the requirements for the degree of Master of Science in Civil Engineering Udaya B. Halabe, Ph.D., P.E., Chair Hota V. S. GangaRao, Ph.D., P.E. Hema J. Siriwardane, Ph.D., P.E. Department of Civil and Environmental Engineering Morgantown, West Virginia 2010 Keywords: Infrared Thermography, Pulsed Thermography, FRP, Composites

2 ABSTRACT Nondestructive Evaluation of Composite Railroad Ties and Bridge Components Using Infrared Thermography Srinivas Majiga This study focuses on the review and application of infrared thermography technique for detecting subsurface defects such as debonds in timber railroad ties and bridge components wrapped with Fiber Reinforced Polymer (FRP) composites. The study also includes a comprehensive literature review on the application of Infrared thermography technique for condition assessment of various materials and structural components. Infrared thermography is a very effective Nondestructive Evaluation (NDE) technique that utilizes the heat energy emitted by an object to characterize its subsurface conditions. It is one of the NDE techniques that is becoming increasing popular for testing field bridges and other components. This problem report investigates the use of digital infrared thermography on FRP composite wrapped timber railroad components under both the laboratory and field conditions. The field trips were conducted to test pile and pile caps in several timber railroad bridges in Moorefield, West Virginia and to test composite railroad ties in the laboratory and field. The lab testing of the composite railroad ties using infrared thermography was conducted during the manufacturing process to evaluate the condition of the ties prior to field installation. Subsequent field testing of the railroad ties was conducted to assess if any debonds were formed in the ties during service. The infrared tests were conducted with the help of both ThermaCAM S60 and InfraCAM SD infrared cameras manufactured by FLIR Systems. The latter camera is a lower cost and lighter weight model, and this study demonstrated that such a camera can also perform very well during field testing and produce very good data quality for debond detection. The results reveal that infrared thermography is a very effective and easy to use technique for detection of subsurface debonds in timber bridge and railroad components wrapped with FRP composites. Availability of low cost infrared cameras can make the technique popular among contractors, and widespread use of the technique can help to ensure continued structural integrity of FRP wrapped components.

3 ACKNOWLEDGEMENTS I would have never accomplished this research study without the continuous encouragement and support from my advisor and Advisory and Examining Committee (AEC) Chair, Dr. Udaya B. Halabe. His invaluable comments and suggestions during the preparation of this problem report are sincerely appreciated. Sincere thanks are also extended to Dr. Hota V. S. GangaRao and Dr. Hema J. Siriwardane for serving as members of my AEC. I am deeply grateful to my parents for their encouragement during the entire period of my study. I would also like to thank my sister who stood by me all this time. I am very grateful to the Department of Civil and Environmental Engineering, West Virginia University for the educational experience I have received during my MSCE Program. The financial support that I received in the form of research and teaching assistantships from West Virginia University during my graduate study is gratefully acknowledged. iii

4 TABLE OF CONTENTS ABSTRACT... ii ACKNOWLEDGEMENTS... iii TABLE OF CONTENTS... iv LIST OF FIGURES... viii LIST OF TABLES... xiv 1 INTRODUCTION Background Research Objectives Scope Report Organization LITERATURE REVIEW Optimum NDT using IRT for Defected Concrete (Zi et al. 2008) Laboratory Experiments Heating Sources Analysis of the Thermography data Results and Discussions Conclusions Pulsed Phase Infrared Thermography (Maldague et al. 1996) Introduction Pulse Infrared Thermography Modulated Infrared Thermography Pulsed Phase Infrared Thermography (PPT) Results and Discussions iv

5 2.2.6 Conclusions Pulsed Thermography Simulation: 1D, 2D and 3D Electro-Thermal Model Based Evaluation (Jena et al. 2006) Introduction Electro-thermal Approach to Pulsed Thermography Electro-thermal Circuit for Heat Conduction Sample Information Simulations Results and Discussions Conclusions New Absolute Contrast for Pulsed Thermography (Pilla et al. 2002) Introduction Reconstructed Sound Area Temperature Images New Absolute Contrast (NAC) Time Considerations Experimental Data Results and Discussions Conclusions Pulse Thermography Applied on a Complex Structure Sample: Comparison and Analysis of Numerical and Experimental Results (Susa et al. 2007) Introduction Experiment Numerical Modeling Results and Discussions Conclusions Advances in Pulsed Thermography (Shepard 2001) v

6 2.6.1 Introduction Noise Reduction, Analysis and Compression of Raw Data Pulsed Thermographic Image Synthesis Simultaneous Processing of Multiple Image Sequences Conclusions Detection of Air Blisters and Crack Propagation in FRP Strengthened Concrete Elements using Infrared Thermography (Hu et al. 2002) Introduction Experimental Program Thermographic Results and Observations Conclusions Detecting of Defects in Polymeric Materials using Pulsed Infrared Thermography (Szczepanik et al. 2008) Introduction Experimental Results and Discussions Conclusions Active Infrared Thermography applied to Detection and Characterization of Non Emergent Defects on Asphalt Pavement (Dumoulin et al. 2002) Introduction Laboratory Test and Numerical Simulations Results and Discussions Conclusions NDE of Composites Delamination by Infrared Thermography (Songling et al. 2003) Introduction vi

7 Principle of Infrared Thermography Experiments Results and Discussions Conclusions INFRARED FIELD TESTING Field Testing and Evaluation of Timber Railroad Bridge Components Using Infrared Thermography Introduction Infrared Testing Equipment Description of the Bridges and Infrared Field Testing Results Conclusions Laboratory and Field Testing of FRP Composite Railroad Ties Introduction Laboratory Testing, Analysis and Results Field Testing of FRP Composite Railroad Ties Conclusions CONCLUSIONS AND RECOMMENDATIONS CONCLUSIONS RECOMMENDATIONS FOR FUTURE RESEARCH REFERENCES vii

8 LIST OF FIGURES Figure 2.1 Equipment set-up (Zi et al. 2008)....5 Figure 2.2 (a) Infrared camera (b) Halogen lamp and (c) Infrared lamp (Zi et al. 2008)...5 Figure 2.3 (a) Void defects and (b) Debond defects (Zi et al. 2008)....6 Figure 2.4 The evolution of the dimensionless temperature of CID1 series heated by the sun light for 5 A.M. to 6 P.M. (Zi et al. 2008)....8 Figure 2.5 The evolution of the dimensionless temperature of CID2 series heated by the sun light for 5 A.M. to 6 P.M. (Zi et al. 2008)....9 Figure 2.6 The evolution of the dimensionless temperature of CID3 series heated by the sun light for 5 A.M. to 6 P.M. (Zi et al. 2008)....9 Figure 2.7 Experimental set-up for infrared thermography experiments (Maldague et al. 1996) Figure 2.8 Three-dimensional plots over deep inclined slot and strip of white paint Figure 2.9 Plastic specimen: non uniform heating Figure 2.10 Aluminum specimen: simulated corrosion Figure 2.11 One-dimensional equivalent electrical model of heat conduction (Jena et al. 2006) Figure 2.12 Two-dimensional equivalent electrical model for heat conduction (Jena et al. 2006) Figure 2.13 Three-dimensional equivalent electrical model for heat conduction (Jena et al. 2006) Figure 2.14 Dimensional layouts of the (a) Mild steel, and (b) CFRP samples (Jena et al. 2006) Figure D, 2D and 3D simulation results of mild steel (top two plots) and CFRP (Jena et al. 2006) Figure 2.16 (a) Small positive error on time t e (b) Small positive error on time t e and (c) Error compensated (Pilla et al. 2002) Figure 2.17 Plexiglas specimen : (a) ΔT[i,j](t) and ΔTs[i,j](t) plotted for 3 locations over,.. 33 Figure 2.18 (a) New absolute contrast image at t = 6.40 s, (b) PPT phase image at f = 0.5 Hz (plastic specimen) (Pilla et al. 2002) viii

9 Figure 2.19 Raw thermogram (graphite-epoxy specimen) at t = 0.57, 1.17, 1.53 s (Pilla et al. 2002) Figure 2.20 New absolute contrast image (graphite-epoxy specimen) at t = 0.57, 1.17, 1.53 s (Pilla et al. 2002) Figure 2.21 PPT phase image (graphite-epoxy specimen) at f = 1 Hz (Pilla et al. 2002) Figure 2.22 Sample plate drawing with specified type, size and defect position (Susa et al. 2007) Figure 2.23 Experimental set-up (Susa et al. 2007) Figure 2.24 Sample plate drawing with specified type, size and defect position (Susa et al. 2007) Figure 2.25 Thermograms obtained from pulse experiment on a sample, left - at t=2s and right at t = 20s after the heat pulse has been applied (Susa et al. 2007) Figure 2.26 Surface temperature distribution as obtained by numerical simulation corresponding to t=2s (left) and t = 20s (right) after the heat pulse has been applied (Susa et al. 2007) Figure 2.27 Surface temperature decay curves above the defective and non-defective area, experimental and numerical results; crushed core defect; left side of the plate (Susa et al. 2007) Figure 2.28 Surface temperature decay curves above the defective and non-defective area, experimental and numerical results, core unbound defect, right side of the plate (Susa et al. 2007) Figure 2.29 Thermal contrast evolutions in time, experimental and numerical results; crushed core defect; left side of the plate (Susa et al. 2007) Figure 2.30 Thermal contrast evolution in time, experimental and numerical results; extra adhesive defect; left side of the plate (Susa et al. 2007) Figure 2.31 Thermal contrast evolution in time, experimental and numerical results, core unbound defect, right side of the plate (Susa et al. 2007) Figure 2.32 Thermal contrast evolution in time, experimental and numerical results, delaminations defect, right side of the plate (Susa et al. 2007) ix

10 Figure 2.33 Thermal contrast evolutions in time influence of non-uniform heating, experimental and numerical results; crushed core defect; right side of the plate (Susa et al. 2007) Figure 2.34 Conventional thermographic images processing of graphite epoxy thermal resolution target Figure 2.35 Logarithmic temperature-time plot of points on steel slabs ranging in thickness from 0.25 to 1.0. Each plot trace displays nonlinear behavior immediately after flash heating, followed by linear behavior, and a deviation from linearity at a time that is correlated to the slab thickness (Shepard 2001) Figure 2.36 Comparison of Reconstructed 2 nd derivative (top) and raw 100 frame average (bottom) images of an adhesive disbond in a graphite epoxy panel. Flash reflection artifacts in the raw image are eliminated in the synthetic image (Shepard 2001) Figure 2.37 a) Synthetic 2 nd derivative of graphite epoxy thermal resolution target (extracted from same data set as the images in Figure 2.34). (b) Synthetic depth image of the resolution target (Shepard 2001) Figure 2.38 (a) Simultaneously processed 1 st and 2 nd derivative images of hollow channels deep in a 4 x 3 Inconel panel. Each image comprises a 5 x 4 image array. The derivative images of the entire panel are created in less than 2 seconds (Shepard 2001) Figure 2.39 Illustration of embedded blister locations (Hu et al. 2002) Figure 2.40 (Plate 1) CFRP plate concrete specimens, where the artificial blisters are embedded between the interface of composite and substrate (Hu et al. 2002) Figure 2.41 (Plate 2) Thermal images taken at various distances for the concrete samples plated by CFRP (see Plate 1 for the corresponding actual photograph) (Hu et al. 2002) Figure 2.42 Configurations of test beams (Hu et al. 2002) Figure 2.43 (Plate 3) Actual photograph of concrete samples, after laminated CFRP Figure 2.44 (Plate 4) Thermal image taken during 45 to 55 kn cyclic load Figure 2.45 Thermal image taken after failure occurred (at the cyclic load range of kn) (Hu et al. 2002) x

11 Figure 2.46 (Plate 6) Actual photograph for 1.2m concrete beam laminated by GFRP taken after failure occurred (Hu et al. 2002) Figure 2.47 Specimens for Testing Szczepanik et al. (2008) Figure 2.48 Schematic draw of thermovision research stand: (a) Specimen heating process and (b) Observing specimen s surface by thermovision camera (Szczepanik et al. 2008) Figure 2.49 The thermal of polyethylene specimen registered twelve seconds after the end of heating process the first signs of subsurface defects (Szczepanik et al. 2008) Figure 2.50 The thermal of polyethylene specimen registered sixty seconds after the end of heating process clearly visible defects manifested by temperature incensement (Szczepanik et al. 2008) Figure 2.51 The thermal of polymethacrylate (methylate) specimen registered one hundred fifty seconds after the end of heating process (Szczepanik et al. 2008) Figure 2.52 The thermal image of laminate registered one hundred fifty seconds after the end of heating process (Szczepanik et al. 2008) Figure 2.53 The thermal image of steel specimen registered two seconds after the end of heating process (Szczepanik et al. 2008) Figure 2.54 The dependences of temperature on cooling time for areas with and without defect. (a) Polyethylene (b) polymethacrylate (methylate) and (c) laminate (Szczepanik et al. 2008) Figure 2.55 (a) View of pine wood defects (b) Sample front face viewed (Dumoulin et al. 2002) Figure 2.56 Experimental thermograms for a 300s step heating duration (Dumoulin et al. 2002) Figure 2.57 Infrared images simulated for a 2620 W.m-2 heat flux density pulse during 300s (Dumoulin et al. 2002) Figure 2.58 Square defect depth maps in meter: (a) simulation and (b) measurement (Dumoulin et al. 2002) Figure 2.59 A sketch map of heat conduction (Songling et al. 2003) Figure 2.60 The block map of infrared testing system (Songling et al. 2003) Figure 2.61 Defect temperature gradient curves (Songling et al. 2003) xi

12 Figure 2.62 Defect relative temperature gradient curves (Songling et al. 2003) Figure 2.63 Infrared images of honeycomb aluminum composites (Songling et al. 2003) Figure 3.1(a) Picture of the ThermaCAM S60 infrared camera Figure 3.2 Shop heater Figure 3.3 Location of Moorefield, WV Figure 3.4 Photograph of Fort Runs bridge (new #36.7, old #568) Figure 3.5 Infrared setup in the field Figure 3.6 (a) Photograph of the first pile with a metal bolt, and (b) Infrared image showing the metal bolts as cold spot Figure 3.7 (a) Photograph of the second pile and (b) infrared image Figure 3.8 (a) Photograph of the pile cap and (b) infrared image Figure 3.9 Photograph of Dumpling Runs Bridge (new #37.4, old #570) Figure 3.10 (a) Photograph of the first pile and (b) infrared image Figure 3.11 (a) Photograph of the second pile and (b) infrared image Figure 3.12 (a) Photograph of the third pile and (b) infrared image Figure 3.13 (a) Photograph of the fourth pile and (b) infrared image Figure 3.14 (a) Photograph of the fifth pile and (b) infrared image Figure 3.15 (a) Photograph of the second pile and (b) infrared image Figure 3.16 (a) Photograph of the pile cap and (b) infrared image Figure 3.17 (a) Photograph of the pile cap and (b) infrared image Figure 3.18 Photograph of Lilly Pond bridge (new #42.4, old #574) Figure 3.19 (a) Photograph of the first pile and (b) infrared image Figure 3.20 (a) Photograph of the first, second and third piles and (b) infrared image Figure 3.21 (a) Photograph of the second and third piles and (b) infrared image Figure 3.22 (a) Photograph of the fourth pile and (b) infrared image Figure 3.23 Photograph of Durgon Bridge (new #46.2, old #583) Figure 3.24 (a) Photograph of the first pile and (b) infrared image Figure 3.25 (a) Photograph of the second pile and (b) infrared image Figure 3.26 (a) Photograph of the first and second pile and (b) infrared image Figure 3.27 Schematic of the composite tie Figure 3.28 Wooden ties after encasing with GFRP composite xii

13 Figure 3.29 (a) Photograph of the first tie and (b) infrared image Figure 3.30 (a) Photograph of the composite tie and (b) infrared image Figure 3.31 (a) Photograph of the first tie and (b) infrared image Figure 3.32 (a) Photograph of the second tie and (b) infrared image Figure 3.33 Location of the seven molded composite ties, (a) two at the southern end, (b) three in between, and (c) two at the northern end Figure 3.34 (a) Photograph of the first tie and (b) infrared image Figure 3.35 (a) Photograph of the second tie and (b) infrared image Figure 3.36 (a) Photograph of the third tie and (b) infrared image Figure 3.37 (a) Photograph of the fourth tie and (b) infrared image Figure 3.38 (a) Photograph of the fifth tie and (b) infrared image Figure 3.39 (a) Photograph of the sixth tie and (b) infrared image Figure 3.40 (a) Photograph of the seventh tie and (b) infrared image Figure 3.41 (a) Photograph of the first and second FRP wrapped wooden ties, (b) Close up photograph Figure 3.42 (a) Photograph of the first FRP wrapped wooden tie and (b) infrared image Figure 3.43 (a) Photograph of the second FRP wrapped wooden tie and (b) infrared image xiii

14 LIST OF TABLES Table 2.1 Displacements of 1.2m RC beam subjected to static and cyclic loading tests Table 2.2 Thermophysical properties of materials used for numerical simulations (Dumoulin et al. 2002) Table 2.3 Testing results (Songling et al. 2003) Table 3.1 Timber bridge names and numbers xiv

15 1 INTRODUCTION 1.1 Background For years, civil engineers have been in search for alternatives to steels and alloys in the construction industry to combat against the high costs of repair and maintenance of structures damaged by corrosion, harsh environment, and dynamic/fatigue loading. For example, cost estimates for maintenance of highway bridge components composed of steel-reinforced concrete are very high. Therefore, composite materials, formed by the combination of two or more distinct materials in a microscopic and/or macroscopic scale, are gaining increasing popularity in the field of civil engineering. Fiber Reinforced Polymer (FRP) is a relatively new class of composite material manufactured from fibers and resins and has proven efficient and economical for the rehabilitation of deteriorating structures in civil engineering, as well as for new construction. The increasing use of composite materials in civil engineering infrastructure can be attributed to many advantages offered by composites such as light weight, excellent corrosion and fatigue resistance, high strength, and high impact resistance. The superior mechanical properties of FRP composites make them ideal for widespread applications in civil infrastructure. One of the major applications of FRP composite materials involves the repair and rehabilitation of damaged or deteriorating structures. Several companies across the world are beginning to wrap damaged bridge piers and steel-reinforced columns with FRP fabrics to improve the structural integrity, prevent collapse, and to prevent buckling of the reinforcement. In addition to increasing the load carrying capacity, wrapping the structural components with FRP fabrics enhances their ductility behavior which is very important, especially in seismic regions. A proper bond (near one hundred percent) between the FRP wrap and the underlying component's surface is of primary importance for appropriate functioning of the FRP wrap, i.e., for effective transfer of load from the underlying component to the wrap through confining action. Discontinuity between the wrap and the underlying component's surface could be developed during FRP wrapping of the structural component and during the service life of the component due to various reasons, such as presence of moisture, vehicular loads 1

16 etc. Such a discontinuity is commonly referred as a debond. Early detection of such debonds makes the repair work less expensive and ensures continued structural integrity. In recent years, Nondestructive Testing (NDT) techniques have become more and more popular for maintenance and quality assurance in manufacturing processes. This study focuses mainly on the applicability of Infrared Thermography (IRT) for detecting debonds in piles and pile caps of timber railroad bridges wrapped with Glass Fiber Reinforced Polymer (GFRP) fabrics. The study also includes application of IRT technique for debond detection in composite railroad ties, both during the manufacturing stage and during service. Previous research studies conducted at West Virginia University had demonstrated the usefulness of IRT in detecting subsurface debonds in FRP composite bridge decks (Vasudevan 2004) and FRP wrapped concrete cylinders (Dutta 2006). These studies were conducted using a highend infrared camera. The field studies in the current study were conducted using a low-cost infrared camera to demonstrate the usefulness of low-end infrared cameras that have now become widely available. These low end models can make the IRT technique popular among contractors and field inspectors. 1.2 Research Objectives The objectives of this research study are as follows: To conduct a comprehensive literature review of the recent advances on the application of infrared thermography for subsurface defect detection To demonstrate the feasibility of using a low-cost infrared camera for subsurface defect detection in FRP wrapped timber components in the field. 1.3 Scope Literature review on infrared thermography technique and its applications have been carried out. This review has focused on the more recent publications, primarily during This research involved the extensive use of the state-of-the art digital infrared camera which produced a sequence of infrared images as well as a low-cost infrared camera which produced "single shot" infrared images in radiometric JPEG format. Laboratory infrared testing was conducted on GFRP composite railroad ties. Field infrared testing was conducted on GFRP wrapped components (e.g., timber piles and pile caps) in timber railroad bridges as well as on GFRP composite railroad ties. These infrared tests revealed the presence of 2

17 embedded debonds at various locations and also served to demonstrate the usefulness of the low-end infrared camera for field testing. 1.4 Report Organization This problem report is organized into four chapters. Chapter 1 presents the background, objectives and scope of this study. Chapter 2 presents critical reviews of the literature dealing with the infrared thermography technique, and recent developments and applications of IRT on various concrete and composite bridges. Chapter 3 gives a detailed description of both the high-end and low-end digital infrared cameras that were used during the laboratory and field testing. Chapter 4 presents the conclusions of this research and recommendations for future studies. Finally, a listing of all the references cited in this study is provided at the end. 3

18 2 LITERATURE REVIEW This chapter reviews the literature on application of infrared thermography for defect detection in concrete and composite members. The focus of this work is on reviewing the recent advances, so most of the literature cited here has been published during Optimum NDT using IRT for Defected Concrete (Zi et al. 2008) Zi et al. (2008) experimentally illustrated the infrared thermography technique to detect the defects of concretes. He considered two different kinds of defects such as voids located 10 to 30 mm below the surface and debond between concrete surface and Fiber Reinforced Polymer (FRP) which is used to strengthen the concrete and increasing its ductility. Three different sources of heating where considered which includes sun light, halogen lamp and infrared lamp Laboratory Experiments Equipment set-up The equipment set-up is shown in Figure 2.1 Equipment set-up (Zi et al. 2008).. The infrared camera is used for the experiment which is produced by FLIR. The resolution of the infrared camera is 320 x 240 pixels. Thermal sensitivity of the device is in between 0.8 o C and 30 o C. The imaging performance i.e., instantaneous field-of-view for the camera has a spatial resolution of 1.3 µrad (IFOV). It can detect infrared radiation in the spectrum range of 7.5 to 13 microns and has an image frequency (capture rate) of 50/60 Hz. The camera was operated by a Windows XP personal computer. 4

19 Figure 2.1 Equipment set-up (Zi et al. 2008). Three different heat sources were used: the natural sun light, a halogen lamp and an infrared lamp as shown in Figure 2.2. (a) (b) (c) Figure 2.2 (a) Infrared camera (b) Halogen lamp and (c) Infrared lamp (Zi et al. 2008). Specimen set-up The concrete block specimens are used for the experiments which are made up of ordinary plain concrete. The maximum aggregate size of the concrete was 20 mm. The water/cement ratio was 0.47 and the compressive strength of the concrete was 27 MPa. These values were measured by the standard procedures. Four different sizes of the defects were considered with diameters 20, 40, 60 and 80mm for both void and delamination. The void defects were placed in three different depths of 10, 20 and 30mm (Figure 2.3). The voids in the concrete specimen are made as follows: Defect-shaped Styrofoam attached to the tips of 5

20 1cm diameter nozzle at the locations of the defect was placed at the desired locations before the fresh concrete was cast. After the concrete was hardened, a solvent was carefully dropped through the nozzle to solve out the Styrofoam imbedded in the concrete. The delamination (defect) was made by removing the adhesive at the position of the defect on the surface of the hardened concrete block before the sheets were attached (Kamoi et al. 2004). Ordinary concrete has a high emissivity as So the surface temperature of the void specimen was captured by the infrared camera directly without any surface treatment. For specimens with delamination, the surface treatment was required to capture the surface temperatures using the infrared camera. The surface emissivity of FRP sheet was only about 0.7, which would influence the thermography. For this reason, the specimen s surface was treated with a black lacquer spray which has high emissivity of 0.8 to 0.95 (Jeff and Hamilton 2003). This led to increased image quality and more accurate temperature readings. (a) (b) Figure 2.3 (a) Void defects and (b) Debond defects (Zi et al. 2008) Heating Sources The three heating sources that were used were the natural sun light, a halogen lamp and an infrared lamp. The wattage rating of the halogen and infrared lamps were not available in the paper by Zi et al. (2008). Because the heat supply by the sun light is a function of time, the surface thermography of the specimens was captured from 5 A.M. to 6 P.M. continuously in every 1 hour when the specimen was heated by the sun light. The halogen lamp and the infrared lamp were placed 5 cm above the top of the specimens to heat up the specimens for 3 min. After heated by the heating sources, the specimens were put 6

21 below the infrared camera for the test. The thermographic pictures were taken at one second intervals (Weritz et al. 2005) Analysis of the Thermography data From the test data, time-dependent thermography results could be produced. The position of defect was identified as the temperature change in the thermograph as the temperature change in the thermograph as shown in Figure 2.4, Figure 2.5 and Figure 2.6. To investigate the results closely, it is better to monitor the change of the temperature of the defected region. To reduce the influence of the ambient temperature, a dimensionless temperature change ( T) given by the following equation (Zi et al. 2008) was used. T T T d r T (2.1) r where T r is the reference temperature and Td is the is the temperature of the defective region. The reference temperature is the surface temperature of the sound (defect-free) region. Then the evolution of the dimensionless temperatures with respect to time was obtained. To reduce the noise of the result, the temperatures were averaged over a small circular area of 2 cm diameter (Zi et al. 2008) Results and Discussions All of the four sizes of the voids located at a depth of 10mm below the surface could be easily identified from the temperature changes by any of the three heating sources. Infrared heating gave clearer thermographic image than heating by the halogen lamp. In the case of the sun light, the dimensionless temperature was most noticeable between 9:00A.M. to 1:00P.M. However, no information on the month (time of the year) was available for solar heating related experiments in the paper by Zi et al. (2008). By adjusting the contrast scale in the infrared images, the boundary of the defects can be seen more prominently. This ability to adjust the contrast in the recorded images is one of the major advantages of using the digital infrared system when compared to the non-digital infrared system. An opposite trend was observed between 5:00 P.M. to 6:00P.M. i.e., the temperatures above the defects appear to be colder than the surrounding defect-free areas. At this time the specimen is undergoing a cooling cycle and hence the defects show up as lower temperature areas. 7

22 It was found that the voids in 20 and 30mm depths could be identified with the sun light only when the specimens were investigated between 9:00A.M. to 1:00P.M. But the temperature changes of the voids smaller than the depth were not that noticeable compared to larger voids. It seems that the voids more than 20mm below the surface could not be identified by the halogen-lamp heating. The infrared lamp gave better performance than the halogen lamp. However, the infrared lamp did not work either, for the voids that were 30mm below the surface. The results for the delamination defects between the concrete and FRP sheet were calculated. It shows that the sun light did not result in enough temperature changes for the thermographic method. Both the halogen lamp and the infrared lamp were effective to enhance the thermographic differences due to the existence of the delamination defect. Just after heating, the temperature differences were monitored clearly, and the results are shown in Figure 2.4, Figure 2.5 and Figure 2.6. Time (hour) Figure 2.4 The evolution of the dimensionless temperature of CID1 series heated by the sun light for 5 A.M. to 6 P.M. (Zi et al. 2008). 8

23 Figure 2.5 The evolution of the dimensionless temperature of CID2 series heated by the sun light for 5 A.M. to 6 P.M. (Zi et al. 2008). Figure 2.6 The evolution of the dimensionless temperature of CID3 series heated by the sun light for 5 A.M. to 6 P.M. (Zi et al. 2008). 9

24 2.1.5 Conclusions The voids near the concrete surface can be identified easily by using the thermography method. For solar heating, the optimum time for acquiring infrared data is between 9:00AM to 1:00PM. It was also determined that the voids were detected clearly when the diameter of the voids was larger than the depth. In absence of sun light, an artificial heating will be needed. The infrared lamp was found to be better than the halogen lamp. The defect between the concrete surface and FRP sheet can be identified by using the thermography method. The heating using sun light was not effective for such a problem. Both the halogen lamp and the infrared lamp were found to be more effective for sharpening the thermographic information. 2.2 Pulsed Phase Infrared Thermography (Maldague et al. 1996) Maldague et al. (1996) has given a new approach which combines the advantages of both Pulse infrared Thermography (PT) and Modulated infrared Thermography (MT). In a view of nondestructive evaluation (NDE), the specimen or the material is pulse-heated as in pulse infrared thermography and the mixture of different frequencies of all the thermal waves applied into the specimen are altered by performing the Fourier transform of the temperature evolution over the field of view. The maximum phase image with more important features is: deeper probing, less influence of surface infrared and optical characteristics, rapid image recording (pulse heating, surface- wide inspection), with possibility to inspect high thermal conductivity specimens. With the help of above features, several results were calculated and the theory is discussed as well (Maldague et al. 1996) Introduction Infrared thermography is the most efficient nondestructive evaluation (NDE) method with an increasing span of applications in the recent years. An external thermal stimulation is brought and launched to the part of the specimen for the inspection and analysis of the thermal response to this stimulus and the readings are recorded by an infrared detector to provide information about the internal structure of the part (such as thermal properties and presence of defects in the specimen). Most common types of thermal stimulation are the 10

25 Pulse infrared Thermography (PT) and the Modulated infrared Thermography (MT). In this paper, both PT and MT principles are briefly explained. And also, explained the principles of pulse phase thermography (PPT) and experimental results are presented respectively. This clearly explains that PPT combines the advantages of PT and MT without sharing their limitations. Specifically, PPT offer deeper thickness of probing under the surface, less sensitivity to optical and infrared specimen surface features, better defect shape resolution, non-necessity to know a prior position of a non-defect area in the field of view to compute contrast image, ability to inspect high thermal conductivity specimens than the other methods (Maldague et al. 1996) Pulse Infrared Thermography In experiments related to the PT, a small pulse of energy is brought to a point and this launches a thermal front which propagates under the surface by diffusion according to the Fourier diffusion equation. This pulse of energy is generated by means of laser beam, lamp, flash, air or water jets (a warm or a cold stimulation which makes the temperature differential). The actual pulse energy duration ranges from microseconds, milliseconds to seconds depending upon the thickness of specimen or material and also its thermal properties (especially the thermal conductivities K). The analysis is conducted on the specimens or materials either during the rising surface temperature or during the decay. In PT, a useful relationship relates the thermal propagation time t to the depth z of a subsurface defect. 2 z t (2.2) where is the material thermal diffusivity [m 2 /s], α = K/(ρC), is mass density and C is specific heat (Maldague et al. 1987). One of the main advantages of the PT approach is the ease of deployment in the field; and it is also fast since operating in a pulse-transient regime, and whole area of the specimen can be analyzed with single pulse which is shown in Figure 2.7. Moreover, as stated by equation 2.2, typical NDE thickness for most common materials requires a minimum observation time of less than one minute. A difficulty in PT is the computation of the thermal contrast C which requires computing temperature differences in images with respect to a non-defect area requiring some a priori knowledge about the images. 11

26 et al. (1996): To calculate the thermal contrast C(t) the following equation was used by Maldague Ti ( t) Ti ( t0 ) C( t) (2.3) T ( t) T ( t ) s s 0 where indexes i and s refers respectively to temperatures over a suspected defective area and over a sound area. The thermal contrast C (t) is computed with respect to temperature distribution at time t 0 just before heating (to suppress the adverse contributions from the surroundings) and normalized by the behavior of a sound area: a unit value will be obtained over a non defect area. Obviously, C(t) is function of time and for quantitative analysis, its maximum value C max (computed at t max ) is of interest (Maldague et at. 1996). Figure 2.7 Experimental set-up for infrared thermography experiments (Maldague et al. 1996). 12

27 2.2.3 Modulated Infrared Thermography In Modulated infrared Thermography (MT) or Lock-in Thermography the specimen is subjected to sinusoidal temperature stimulation (Busse et al and Busse 1994). In the stationary regime, the specimen thermal response is also described by a sinusoidal regime whose magnitude and phase mostly depends on the input frequency. In these conditions a highly attenuated and dispersive wave which is sometimes referred to as a thermal wave stands inside the material. These thermal waves were already investigated by Fourier and Angstrom in the 19th century but the use of such waves for NDE has caught the interest of researchers recently because of the non-contact way of generating and detecting these waves (Maldague et al. 1996) In a point investigation, the modulated laser beam is applied on the test specimen and heated while the corresponding thermal infrared emission witnesses the resulting oscillating temperature field. The magnitude and the phase of the thermal wave can be observed by means of an infrared detector with proper optics and with a lock-in amplifier used to synchronize the input and output signals. This principle was recently extended to surface thermal wave imaging in which a given area of the specimen surface is exposed to sinusoidal illumination, for instance by means of a modulated heating beam (Figure 2.7). All the data points are monitored with an infrared image on the analyzed specimen; the lock-in amplifier is not required since the detection delay is obtained by computations done on the digitized data. If the reference function and output signal are of sinusoidal nature are present only few data points per cycle are required, these points allows to compute both the magnitude (A) and the phase shift (Φ) with respect to the reference modulation equations 2.3 and 2.4. The magnitude image (A) is proportional to both local optical and infrared surface features. Then is the phase shift (Φ) image which is related to the propagation time delay which is independent of optical or surface features. As for depth range probing is limited close to surface features due to the high damping of the thermal wave. However, it is easily shown that phase shift (Φ) image can probe nearly twice the thickness probed by magnitude (A) image. MT also shares the same advantages as the PT and with following differences. Better depth resolution depending up on the modulation frequency which can be tuned respectively (high modulation frequencies implies the response to the near surface region), 13

28 Insensitivity to surface artifacts with some problems for practical implementations (more complicated set-up), Very slow in the acquisition process, for example, 2 minutes are required to probe a 2mm thick carbon fiber reinforced plastic (CFRP) at a modulation frequency of 0.03 Hz; where as a PT approach requires about 30seconds to obtain the maximum thermal contrast image with obviously a less clearer resolution than of the MT phase image (Busse 1994) Pulsed Phase Infrared Thermography (PPT) It s very easy to show how PT is related to MT. It is well known from transform analysis that a ideal Dirac pulse δ(t) in the time domain has an infinite flat spectrum in the frequency domain: δ(t) 1. It is said that in MT, a single frequency is tested in the stationary regime. Where as in PT, a rectangular pulse is used instead of an ideal Dirac pulse, thus the frequencies in the specimen are not of constant amplitudes since for a rectangular pulse of width τ and amplitude A, centered at time t = 0, the frequency is expressed as A Sin( u ) / u equation 2.4 where u is the frequency variable. The possibility of thermal wave frequencies to unscramble the spectrum in PT experiment is attractive since such processing could combine together advantages of both PT and MT. The procedure for finding the pulsed phase infrared thermography (PPT) is mainly based on the Fourier transform. The sequence of infrared images to process witnessing the temperature decay following the initial thermal pulse is obtained as in conventional PT experiments and for each pixel (i,j), the temporal evolution f(x) is extracted from the image sequence (where x is the index in the image sequence). Then the discrete Fourier transform is computed with respect to the well known formula. 1 F( u) f ( x) exp[ j2 ux / N] R( u) ji ( u) (2.5) N N 1 n 0 where R(u) and I(u) are respectively the real and imaginary components of F(u). Finally, the phase is computed for each of the transformed terms using 1 I( u) ( u) tan (2.6) R( u) Typically frequencies range from 0 and to 1 / x (where x is the time interval between images, i.e., the sampling rate). In experimental setup the minimum value of x is 14

29 19 ms corresponding to the maximum acquisition rate of 52 Hz while the frequency increment u is given by the equation u 1 / N x. For instance, a sequence of N = 32 images yields a frequency increment u of 1.6 Hz (with x = 19 ms). An important point here is the transient regime (MT acquisition is done in the permanent regime). Instead of finding phase (Φ) images at a particular frequency, the best way is to calculate the maximum value of the phase since the final values of phase Φ and amplitude A have not yet settled. Such Φ max image is obtained by considering, for all the pixels (i,j), the maximum value of the phase computed with equation Results and Discussions They have conducted different experiments on PPT to investigate its potentials. The heating pulse can be obtained by two ways i.e., using 7.2 KW lamp housing apparatus (τ = 10 s, x = 1.5s) or quicker high power flashes 12.8 KJ (τ=10 ms, x = 80 ms). The camera used in the experiment was a focal plane array IRC-160 operating in the 3 to 5 μm. Digital images were obtained on 12-bit format and all the experiment synchronizing is computerized (Figure 2.7). (a) Less sensitivity to optical surface disturbances and deeper subsurface probing A specimen was particularly designed to test PPT Φ max images for these properties. The specimen is designed using the Figure 2.9(a) which is made of Nylon with an inclined slot milled dimensions (150 x 5 mm) from the rear side (subsurface slot depth varies from 5 to 15 mm across the specimen of size: 120 x 150 x 15 mm). The front surface coated with high emissivity paint (~ 0.9). Moreover, an optical disturbance is added on the tested surface as an oblique strip of white paint (~ 5 mm width) over the deepest section of the slot. Figure 2.8(a) and (b) show respectively 3D representation of C max represented in equation 2.3 and Φ max images. The field of view is about 3 x 3 cm. While the slot is not visible in the C max image, it is well observed, across the field of view in the image (in Figure 2.8, the slot depth spans from about 14.7 to 12.7 mm respectively from top to bottom). Moreover the optical disturbance is well seen in the C max image (Figure 2.8(a)) while its effect is attenuated in the case of the Φ max image (Figure 2.8(b)). Additionally, the strong effect presents in the contrast image is absent in the Φ max image as discussed in section 3 for MT (Marinetti et al. 1995). 15

30 (b) Less sensitivity to infrared surface disturbances In Figure 2.9 a plastic specimen of 6 mm thickness covered with high emissivity paint (~0.9) was flash-heated with a single flash (shown in Figure 2.7) thus making a strong left to right heating non-uniformity. A cross-shaped defect was milled on the back of the specimen, 1 mm under the front surface. Figure 2.9(a) is the image in equation 2.3 and Figure 2.9(b) is the Φ max image. The strong heating non-uniformity is well seen in the C max image (recorded 10 seconds after the heating is done) while it is less apparent in the Φ max image. In addition, the shape definition of the defect is much better resolved for the Φ max image as explained below. (c) Quick acquisition time to inspect high thermal conductivity specimens In case of high thermal conductivity specimens such as the inspection of aluminum aircraft skin, quick heating schemes are necessary. This is evident from equation 2.2, which indicates that the thermal propagation time is inversely proportional to diffusivity of the material. Figure 2.10 shows the results of experiments conducted on a 1 mm thick aluminum sheet in which back holes of 20 mm diameter were drilled simulating corrosion of 13 % (left) and 80 % (right). In the C max image (recorded at 0.52 seconds after the end of heating), the defects exhibit a different apparent size due to the spreading of the thermal front. This can also be explained by multiple internal reflections the thermal waves experience inside the specimen. It is believed that this phenomenon is not present in the images because only the maximum phase is considered at each pixel location (instead of the cumulative effects of all the various frequencies): a better definition of defect shape is thus noted. These tests show that images share advantages of both MT images (deeper probing, less sensitivity to specimen optical and infrared surface feature) and PT experiments (fast acquisition). Moreover, images do not require any knowledge of the image such as position of a nondefect area which is needed in the case of contrast images. Finally, PPT is easier to deploy in field applications since being suited either to low and high thermal conductivity specimens (stable high frequency surface-wide modulated heating as required in MT experiments is not easy to achieve to inspect high thermal conductivity specimens (Maldague at el. 1996)). 16

31 Deep slot: 3D view of the contrast image Deep slot: 3D view of the phase image (a) (b) Figure 2.8 Three-dimensional plots over deep inclined slot and strip of white paint (a) C max contrast and (b) phase Φ image (Maldague et al. 1996). Contrast Image Phase Image (a) (b) Figure 2.9 Plastic specimen: non uniform heating (a) C max contrast and (b) phase Φ image (Maldague et al. 1996). 17

32 Contrast Image Phase Image (a) (b) Figure 2.10 Aluminum specimen: simulated corrosion (a) C max contrast and (b) phase Φ image (Maldague et al. 1996) Conclusions The Fourier transform based approach presented by Maldague et al. (1996) evades the drawbacks of the Pulsed infrared Thermography (PT) in the NDE, in terms of thickness of probing under the surface and sensitivity to optical and infrared specimen surface features. Modulated infrared Thermography (MT) and Pulse infrared Thermography (PT) advantages are combined through unscrambling the various frequencies launched in pulse infrared thermography to conserve only the maximum phase components in the corresponding Fourier spectrum. Some of the additional advantages were explained such as better defect shape resolution, non-necessity to know a priori position of a non-defect area in the field of view, and ability to inspect high thermal conductivity specimens. The experimental results demonstrated the advantages of this approach while some theory is introduced as well to explain the observed behavior. 2.3 Pulsed Thermography Simulation: 1D, 2D and 3D Electro-Thermal Model Based Evaluation (Jena et al. 2006) Jena et al. (2006) presented the uses of electrical and thermal quantities in the form of electro-thermal models for one-dimension (1D), two-dimension (2D) and three-dimension (3D) simulations of active thermography. Pulsed thermography is one of the efficient NDT method in which temperature decay at the surface of a material is captured after heating the specimen with a small duration pulse for small amount of time. In this paper, the results of 18

33 the simulations are compared and analyzed for mild-steel and carbon-fiber reinforced plastic (CFRP) samples. Decay in amplitude with depth and the effect of lateral heat diffusion for 1D, 2D and 3D are calculated Introduction Active thermography process uses an external stimulus to reveal the sub surface defects in a specimen which are testing. This paper talks about the electro thermal modeling and SPICE simulations of pulsed thermography for 1D, 2D and 3D simultaneously. The material is divided into small parts along the length for calculation of Heat conduction in 1D. Each element has its own resistance (R) and capacitance (C) values which are obtained from electro-thermal model. Similarly for 2D and 3D models are made by dividing it into parts along the cross-sectional dimensions. To model transient heating of the sample, current sources are employed followed by SPICE simulation. By using the above approach the simulation results are reported and analyzed. It also examines the decay in amplitude and thermal propagation delay of thermal waves. Variation in the stimulated results is shown for different models (Jena et al. (2006)) Electro-thermal Approach to Pulsed Thermography By using the equivalence elementary laws of heat and electricity, the RC equivalent of a specimen is calculated considering the electro-thermal quantities. Thus, heat conduction problem is converted into an equivalent electrical problem; where, voltage plays the same role as temperature. The electrical equivalent values of the resistance (R) and capacitance (C) for the mild steel and CFRP sample are calculated using the relationship given below. l R (2.7) KA C cal (2.8) Here l is the length in meters measured from top surface, A is the area of the defect in m 2, K is the thermal conductivity (W/m 0 C), is the density (kg/m 3 ) and c is the specific heat of the material (J/ 0 Ckg). The heat conduction is modeled in all three dimensions 1D, 2D and 3D by dividing the cross section areas into appropriate smaller parts, each with its own equivalent R and C values which are calculated and shown in above equations 2.7 and

34 2.3.3 Electro-thermal Circuit for Heat Conduction Heat flow through a thermally conductive material is described as gradient transport which depends on three quantities, they are: material conductivity, material crosssectional area, and the spatial gradient of temperature. The larger the conductivity, gradient, and the cross-section area, the faster the heat flow will be. In forming the equivalent electrical circuit i.e., Heat and Electricity the following analogies are used Q = ΔT/R th (Heat transfer side) (2.9) I = ΔV/R (Electrical side) (2.10) Here Q, ΔT and R th are the rate of heat transfer, temperature difference and thermal resistance all in thermodynamic units, and I, ΔV and R are the current, voltage difference and electrical resistance all in electrical units correspondingly. One-dimensional Heat Transmission In this case of one-dimensional heat transmission, the heat flow is simulated as a 1D, which is time dependent process and mainly along the x-axis. Heat is considered as negligible when is flows in y-direction and z-direction. Based up on electro-thermal analogy, Length is divided into smaller parts to prepare heat conduction, with equivalent resistance (R) and capacitance (C) values for each part. Equivalent electrical model of 1D transient heat conduction is showed in the Figure The sample model is provided with the heat source simulator at one end and a convection resistance R conv at the end of each RC node. Taking the effect of other modes of heat transfer into consideration i.e., convection. The equivalent electrical resistance, which corresponds to the thermal resistance offered by convection mode of heat transfer, is called as convection resistance. Convection is modeled to an equivalent resistance once the respective coefficients are known. R conv models heat loss due to convection from the surface of the sample and can be easily calculated by the equation 2.11 i.e., from the concept of convection heat transfer. 1 R conv (2.11) hs 20

35 Here S is the surface area and h is the convective heat transfer coefficient. The heat transfer coefficient S mainly depends up on the temperature, physical dimensions, and position of the surface. Figure 2.11 One-dimensional equivalent electrical model of heat conduction (Jena et al. 2006). Two-dimensional Heat Transmission A similar approach of 1D modeling is extended to 2D, by considering the sample as a 2D model entity shown in Figure In this modeling, the surface of the sample is divided into smaller parts along parallel to the direction of incident convection and their resistance R and capacitance C values are calculated using the equations 2.7 and 2.8 as explained earlier. As in the 1D case, the surface elements or parts in front of incident radiations are connected to current sources (which are proportional to incident heat flux and surface area of the element). The 2D model is a link between 1D and 3D models, and has its own limited significance which is shown in the Figure

36 Figure 2.12 Two-dimensional equivalent electrical model for heat conduction (Jena et al. 2006). Three-dimensional Heat Transmission Heat conduction in this sample can be further modeled by dividing it into small cuboids as it is 3D which cannot be divided into the parts and which is shown in Figure Electrically equivalent resistances (R) and capacitances (C) in the different directions of the cuboids are shown in the Figure Connecting all cuboids forms a 3D RC network of the sample together. Elements with the similar valued current sources are connected up in the front surface. Based on this approach even multilayered samples can be modeled by adopting appropriate resistance (R) and capacitances (C) values in different directions. Figure 2.13 Three-dimensional equivalent electrical model for heat conduction (Jena et al. 2006). 22

37 2.3.4 Sample Information For comparison Mild steel and CFRP samples have been considered. These two materials are considered due to their varying thermal properties. Mild-Steel and CFRP The simulated mild-steel sample has the following dimensions: 10.4 cm in length, 9.9 cm in breadth and 1 cm thick. The CFRP sample is 25.2 cm in length, 15.5 cm in breadth and 0.4 cm thick. The CFRP laminate reinforcing consists of bonding the CFRP strips with a high-strength epoxy resin as the adhesive. Both of the samples are shown in Figure Schematic of the mild-steel sample Schematic of the CFRP sample Figure 2.14 Dimensional layouts of the (a) Mild steel, and (b) CFRP samples (Jena et al. 2006) Simulations For present simulations Transient (also known pulsed) regime has been adapted, which consists of flash lamp or lasers for generating the heat pulse and a high speed IR camera system for data collection and analysis. The samples are divided into sections, and the RC network generated from the electro-thermal modeling has been simulated by a circuit simulator SPICE (Simulation Programming with Integrated Circuit Emphasis). The simulation was performed on an evaluation version of PSPICE, which has limitation of 64-nodes. For each element RC values are calculated individually. The convection resistances were used at boundaries of the RC network to model thermal insulation at boundaries and to make the circuit complete for simulation. Incident heating on the sample has been modeled through current sources: duration 15s (including rise and fall times of 1s each) with pulse amplitude 10 ampere. Heating is 23

38 along length (across 20 pieces) i.e. heating is on the surface (breadth thickness). During simulations heat/current stimulation parameters are kept same to facilitate proper comparison. Having modeled the sample in terms of R, C and I elements, SPICE based simulation of the circuit is undertaken. SPICE calculates voltages (temperature) at every node of the circuit as well as current flow (heat flow) through all elements. In order to study the node adjacent to heat source node has been taken into consideration. A transient (pulse) analysis mainly deals with the behavior of these parameters as a function of time. Consideration of anisotropy Many materials are anisotropic. Consider CFRP, parameters such as density, specific heat and thermal conductivity values are different in parallel and perpendicular directions to the fibers. The effect of anisotropy in CFRP material can be modeled in this approach. The resistance (R) and capacitance (C) value mainly depends up on the direction of the fibers and varying accordingly Results and Discussions It is important to evaluate the thermal evolution predictions among the three models for both the mild-steel and CFRP samples. The comparisons between 1D, 2D and 3D models based on simulated data highlights the prediction regarding the thermal propagation (time) decay and amplitude decay with depth shown in the Figure 2.15 in the graphs. Here the anisotropy has not been invoked for CFRP. It may be pointed out at the outset of analysis and discussions that these comparisons do not bring out the relative merits of these models for predicting phase and group delay. Figure 2.15 shows four graphs, which has node voltages (temperature) on y-axis and time on x-axis, for comparison of amplitude and time delay in mild steel and CFRP samples using 1D, 2D and 3D RC models. Here second node is considered in all cases. The top two graphs correspond to mild-steel: the topmost comparing 1D and 2D evolution at the chosen nodes, and the next graph plotting 1D and 3D evolutions. The 1D plot is repeated in both graphs for comparison. The bottom two graphs show the simulated plots for the CFRP sample: the bottom most graph showing the 1D and 3D model based response, while the graph above pertains to 24

39 1D and 2D model. Again the 1D plot in both is repeated for reference and ease of comparison. While comparing the graphs, two points need to be focused on: the peak amplitudes and the time at which they occur. In general, the RC ladder network spreads out as we go from 1D to 2D to 3D, i.e., RC ladder combination increases. This accounts for the drop in peak amplitude as we go from 1D to 2D to 3D simulation for a given material sample, and is a consequence of the expected lateral heat diffusion which is also practically observed. On the time front, this results in comparatively less time being taken to propagate in 1D, little more in 2D and the most time in 3D Conclusions Electro-thermal modeling based simulations has been carried out and 1D, 2D and 3D RC models are compared respectively. Decay in the amplitude with depth as well as increase in peak propagation delay obtained from 1D, 2D and 3D results and physical explanations are clearly shown. The graphs clearly show the comparisons of amplitude decay and time delay for both specimens. The RC model has been extended with the addition of convection resistance and can also include effect of anisotropy. 25

40 Figure D, 2D and 3D simulation results of mild steel (top two plots) and CFRP (Jena et al. 2006). 26

41 2.4 New Absolute Contrast for Pulsed Thermography (Pilla et al. 2002) Pilla et al. (2002) proposed a new absolute thermal contrast method for pulsed infrared Thermography (PT) in this corresponding paper. It does not require the priori knowledge on the specimen or sample because it is mainly based on the computations of reconstructed defect-free images. To account for possible delays in the acquisition time it needs a small correction. Experiments are conducted on both the Plexiglas and graphite-specimens and results are presented. Comparison between Pulse infrared thermography (PT) and Pulsed Phase Thermography (PPT) phase images are also presented along with a discussion on the advantages of the proposed method Introduction Pulsed infrared Thermography (PT) is a common thermal stimulation approach in Active Infrared Thermography (Maldague, X., (2001) Theory and Practice of Infrared Thermography for Nondestructive). Among all the well known established thermogram processing techniques, thermal contrast computations are more convenient in order to enhance subsurface defect visibility, and also to enable some quantitative extractions such as defect depth, size and thermal properties which helps in the data acquisition and analysis of the specimen. Thermal contrast definitions imply the knowledge of a sound area within the infrared camera field of view (Maldague, X., 2001). This is one of the main drawbacks since it requires good information which is not always available. Moreover, in these thermal contrast computations, the temperature on the specimen is assumed to be uniform in all directions and it has been reduced to a scalar value. Sometimes, it is fictional if the energy deposited over the specimen is not uniform or if the thermal properties of the material (such as thermal effusivity) will not change over the specimen. In all these cases, traditional contrast computation leads to inaccuracies (Grinzato et al. (2002) NDE of Porosity in CFRP by multiple thermographic techniques ) To overcome all the difficulties and obtain the reliable thermal contrast, a new technique is proposed in this paper. And also to compute the local temperature of a sound area Ts [i,j] (t) by means of an heat transfer mode (semi-infinite body assumption) and without a priori knowledge of sound area locations (here[i,j] represents location within the field of 27

42 view, T the temperature, t the time and subscript s stands for sound area). Second, the inaccuracy of the acquisition time is also taken into account and corrected Reconstructed Sound Area Temperature Images In Pulsed infrared thermography (PT), the specimen is subjected to a thermal pulse at t 0, which lasts up to t r and a thermogram sequence is recorded with the infrared camera. After recording the surface of the specimen with the infrared camera, it takes few seconds before the defects manifest themselves due to the traveling time of the thermal waves within the specimen (Maldague, X., 2001). Assume that t is a time between t f and the time at which the first temperature spot related to subsurface defects appears. In this case, we have Ts [i,j] (t ) = T [i,j] (t ). Next we compute Ts [i,j] for all the rest of the temporal sequence. Assuming a Dirac pulse applied to a semi-infinite body, the 1D Fourier equation is solved as (z is the depth variable, z = 0 corresponds to the surface, Q is the injected energy at the surface, e is the thermal effusivity of the sample and ΔT is the temperature increase from t = 0). T Q inf inite body( z 0, t (2.12) e t semi ) As it is clearly known that the solution provided by the above equation 2.12 diverges as time elapses and also as plate thickness enlarges with respect to the non-semi-infinitebody case. For instance, consider in case of Plexiglas, it was found the error obtained with the equation 2.12 is less than 2% for observation time of less than 2 seconds and a plate thickness less than the 1 mm. in case of an AI specimen, error is less than 20% for time of 0.02 second and 2 mm plate thickness. Nevertheless, equation 2.12 is a good approximation. At time t, the temperature of the sound area Ts [i,j] (t ) is given by the equation: Q[ i, j] Ts[ i, j] ( t') T[ i, j] ( t') (2.13) e t' [ i, j] Assuming the injected energy over the specimen is changing relatively smoothly, equation 2.12 stands and allows in extracting Q/e locally: Q e [ i, j] [ i, j] t. T ( t') (2.14) ' [ i, j] From the above the temperature of the sound area can be defined locally as a function of t. 28

43 Ts Q[ i, j] t' t' ( t). T[ i, j] ( t'). T[ i, j] ( ') e t t t [ i, j] t [ i, j] (2.15) Performing above equation 2.15 computations over all the surface (all locations [i,j] and for a whole temporal sequence) reconstructs the ideal defect-free thermogram sequence New Absolute Contrast (NAC) Introducing the well known absolute thermal contrast definition C ac : C ac [ i, j] ( [ i, j] t t) T ( t) Ts( ) (2.16) The new proposed thermal contrast definition is first generalized over all thermogram locations: C ac [ i, j] ( [ i, j] [ i, j] t t) T ( t) Ts ( ) (2.17) We can then rewrite the above equation 2.17 in terms of T as follows: ac C t) { T ( t) T (0 )} { Ts ( t) T (0 )} (2.18) [ i, j] ( [ i, j] [ i, j] [ i, j] [ i, j] Where T (0 [ i, j] ) is the absolute temperature of the specimen at location [i,j] before heating (i.e., at t = 0-), this is generally referred to as cold image. Obviously we also have T 0 ) Ts (0 ) since no defects are present at t = 0- and both cancel out in equation [ i, j] ( [ i, j] 2.18 so that equation 2.16 written as. C ac [ i, j] ( [ i, j] t t) T ( t) Ts( ) (2.19) Finally, the New Absolute Contrast is obtained as: C t' t) T[ i, j] ( t). T[ i, j ( ') (2.20) t ac [ i, j] ( ] t Time Considerations. a Since equation 2.13 as a form of T c t, in logarithmic space it can be rewritten as. log( T) a log( t) b where a and b constants of a given location [i,j] on the specimen Q surface and with a = -1/2 and b = log( ). The good thing with such representation is that e t the logarithmic time behavior of defect-free region is a line with -1/2 slope. 29

44 In equation 2.20 time t and t are referenced to t f. Any error t e and t f will affect NAC computations, especially when a logarithmic space is chosen. For instance Figure 2.16 (a) and Figure 2.16 (b) show such effect for positive and negative values of t e. In the figure the curve ΔT vs. (t+) is plotted (case of a plastic plate, no defect). The straight corresponds to the ideal 1/2 slope Ts(t) case. Obviously, the actual data does not match the ideal case and such error t e will affect further NAC computations. The solution to this is to find the error t e and then to compensate it by computing: C ( t t, t' t ) Instead of C nac t, ') nac [ i. j] e e ( [ i. j] t Normally, such procedure should lead to a curve with a linear behavior, at least at the earlier times (due to unaccounted thermal lateral flow, thermal losses), see for instance Figure 2.16 (c). Such adjustment is made dynamically thanks to a dedicated computer program by modifying t e so that both lines coincide together. Interestingly, in another method is shown to deal with a finite pulsed width in case the pulse is not fast enough. (a) 30

45 (b) (c) Figure 2.16 (a) Small positive error on time t e (b) Small positive error on time t e and (c) Error compensated (Pilla et al. 2002). 31

46 2.4.5 Experimental Data Plexiglas plate The 4 mm-thick plate contains 6 flat-bottom holes of depth 1, 1.5, 2, 2.5, 3, 3.5 mm, all with 10 mm diameter. This academic specimen was flash heated for 15 ms (with two flashes of 6.4 KJ of electrical energy per flash) and 200 thermograms were recorded (from t = 0.1 to t = 20 s). The experimental apparatus is described elsewhere. Figure 2.17 (a) shows ΔT vs. time for three locations over the specimen (first 2 curves over defect and last one over sound area). Error t e was corrected as explained in previous section. The computed equation values Q/e is respectively 7.97, 6.72 and 6.05 confirming the non-uniform heating visible on the raw thermogram of Figure 2.17 (b). The dots at t = 0.9 s correspond to t. The straights represent: t' Ts[ i, j] ( t). T[ i, j] ( t') (2.21) t And the dotted curves ΔT[i,j](t). The NAC corresponds, respectively for the three cases, to the difference between the dotted curve and the straight as in equation Figure 2.18 (a) shows the NAC at t = 6.4 s. For reference, the phase image at frequency f = 0.5 Hz in Pulsed Phase Thermography (PPT) is provided in Figure 2.18 (b) as well. As it is seen both images are similar with defects showing over a relatively flat background (compare with raw thermogram of Figure 2.17 (b)). The absolute contrast image is not provided as it is similar to Figure 2.17(a) (since a constant value is simply subtracted from each image equation C ac [ i, j] ( [ i, j] t t) T ( t) Ts( ) this does not modify the relationship between defects and background). 32

47 (a) Figure 2.17 Plexiglas specimen : (a) ΔT[i,j](t) and ΔTs[i,j](t) plotted for 3 locations over, (b) black dots represent the acquisition time for each thermogram, crosses indicate the acquisition time of (b). (b) Raw thermogram at t = 6.40 s (Pilla et al. 2002). (b) (a) (b) Figure 2.18 (a) New absolute contrast image at t = 6.40 s, (b) PPT phase image at f = 0.5 Hz (plastic specimen) (Pilla et al. 2002). Graphite-epoxy plate The next specimen is a graphite-epoxy plate of 5 plies. Simulated defects as a Teflon insert and a thin void were embedded at the manufacturing stage. Fifty images were acquired from 0.55 s to 1.53 s. The same experimental apparatus as in previous section was used. Figure 2.19 shows three raw thermograms recorded at different times. Figure 2.20 shows the NAC for the same three moments. Both defects are clearly seen as also a possible delamination or epoxy-richer zone on the right side of the specimen. For reference, Figure 2.21 shows the phase image at frequency f = 1 Hz in PPT. In this case, the NAC provides better results. 33

48 Figure 2.19 Raw thermogram (graphite-epoxy specimen) at t = 0.57, 1.17, 1.53 s (Pilla et al. 2002). Figure 2.20 New absolute contrast image (graphite-epoxy specimen) at t = 0.57, 1.17, 1.53 s (Pilla et al. 2002). Figure 2.21 PPT phase image (graphite-epoxy specimen) at f = 1 Hz (Pilla et al. 2002). 34

49 2.4.6 Results and Discussions As seen, the NAC is a novel processing technique that improves the signal to noise ratio using a simple and fast - no complex computations in equation algorithm by computing locally the sound area values Ts. In fact, it brings the following advantages to PT: For a given time t, only two thermograms (at t and at t of interest) are needed to compute the NAC while for instance the PPT approach requires many more ( ). In quantitative studies (ex: to extract defect depth), one approach consists to compute the maximum contrast. Since this is computed precisely with the NAC (ex: the local injected energy is taken into account), higher reliability is achieved. Since the background is made flat, defects are segmented easily, especially in case of non uniform heating or local emissivity variations. Ts computed in the NAC can be applied to other contrast definitions as well. Thus, the known Absolute thermal Contrast is strongly enhanced thanks to a better definition of the sound area Ts. The Absolute Contrast is only taken as an example to show the interests about computing this new Ts[i,j](t ). The local sound area Ts[i,j](t ) can be applied to any equation involving a sound area Ts, especially it can enhance most of type of contrasts and not only the absolute contrast. In 1987, found out a powerful way to show defects they called it the Normalized Apparent Effusivity (NAE). For instance, an advantage of the NAE was its ability to maintain a constant background from image to image. Interestingly, the local Ts[i,j](t ) helps explaining the NAE. In fact, the NAE is the inverse of the Running Contrast using Ts[i,j](t ). From this point, the NAE can be seen as a particular case of a general method presented in this article Conclusions In this paper a New Absolute Contrast (NAC) procedure is derived and proposed based on the reconstruction of the temperature of the sound area. It was also shown how important is to adjust uncertainties in the acquisition time. Results were presented on two specimens with comparisons with PPT. Due to its distinctive advantages, it is believed that the interests of the local Ts[i,j](t ) will be well received by the AIRT, PT community. 35

50 2.5 Pulse Thermography Applied on a Complex Structure Sample: Comparison and Analysis of Numerical and Experimental Results (Susa et al. 2007) Susa et al. (2007) experimentally illustrated the Pulse infrared Thermography (PT) applications on complex structure samples. Modeling complex structure samples were inspected by the infrared thermography and using these results the identification of the defect characteristics is done. This paper also analyzes the results obtained by PT experiments on a complex structure samples with defects of different types and sizes located at different depths. The two different types of honeycomb panel are considered with inserted defects of known size, position and type for the sample testing. The sequences of thermograms are noted which are obtained from the experiment and are used to extract the surface temperature evolution curves above the defective and non-defective sample areas. The temperature evolution curves obtained from the experiments were used for comparison with the results obtained by numerical modeling. And also the results obtained from the thermal contrast evolution curves were used to analyze the differences between the experimental modeling and through modeling. For the purpose of finite element analysis, a tested sample model is made so that the finite element method (FEM) could be used to solve the problem of transient heat transfer occurring in experimental conditions. Unknown parameters such as power density, convective heat transfer coefficients and sample surface emissivity used in the numerical model and which were adjusted close to the results obtained from the experiments and the surface temperature decay curves were extracted from the numerical model results. The results obtained from both were analyzed and discussed. Possibilities for improving the results and further research activities are proposed Introduction In the recent years, infrared thermography has been benefited from rapid technical development of imaging systems. The increased performance of infrared cameras when spatial and temperature resolution are considered has led to defect detection improvements. Despite the fact that the defect detection efficiency increased, defect characterization procedures are still a wide area of research. In pulse experimental procedures the defect 36

51 characteristics using different mathematical models as a means of predicting the defect behavior within the sample subjected to the experimental conditions. Since the physical nature of the heat transfer occurring during the experiment was well known to be governed by the differential equation of the transient heat transfer, the main problem of this approach is to find the solution to this equation that would permit the comparison of the both experimentally and theoretically obtained results. Simplifications were used in many cases that considered the heat transfer only in one direction (1D heat transfer models) that were then solved analytically (Delpech et al. 1994). Authors also used iterative methods to find solutions to the 1D problem defined analytically (Delpech et al. 1992). Moreover, the most of them neglected heat losses from the surface in order to further simplify the solution (Maldague et al. 1987). Some authors expanded the analytical solutions of 1D model onto 2D models using thermal quadrupoles and Laplace transformations (Maillet et al and 1994). The analytical disturbance method for the solution in 3D models has also been proposed (Bendada et al. 2005). A numerical method is also used based on finite differences in cases where the symmetry was taken as a constraint: in such a way a 2D model of heat transfer was obtained that was equal to a 3D heat transfer model in cylindrical coordinates and it was solved using finite differences. Control volumes were used for the 2D heat transfer problems and another numerical solution was proposed for corrosion evaluation using 3D heat transfer conditions. Finally the use of FEM was reported as an interesting tool for modeling pulse experiment heat transfer. This work concentrates on modeling of complex structure samples in an attempt to verify the possibilities of using FEM for purposes of solution retrieval for corresponding established of mathematical model in case of more complex samples with multiple different defect types present. In this paper it is experimentally illustrated the infrared thermography technique to detect the defects of concretes Experiment In order to obtain the data needed to develop a numerical model evaluation, a pulse experiment was performed on a complex composite structure sample with inserted defects. Sample description as well as the experimental set-up description is given next. 37

52 Sample tested The sample under inspection is a calibration plate made of two honeycomb panels having different densities. The honeycomb core is placed between two layers of carbon fiber reinforced plastics (CFRP) with a foam adhesive layer between them of known type and thickness. Defects of different sizes and types were inserted at different locations within the sample so that two sample halves with two different densities of honeycomb core were symmetrical with respect to the location of the defects. The sample with defects is shown in Figure It is seen that the three different sized TEFLON defects are inserted within the upper graphite epoxy layer representing the first type of defect. The second defect type was represented by two different sized TEFLON defects which are inserted between the adhesive and the honeycomb core. Extra foam adhesive is applied in regions of specified dimensions to account for the third defect type. Finally, the crushed core defect represents the fourth type of defect that could be found in the sample with the honeycomb crushed in the specified region just below the adhesive foam layer. Experimental set-up The experimental set-up is shown in Figure The experiment was conducted in a reflection mode since it was judged that the sample was too thick for the transmission mode to be successfully employed. Two high power (6.4 kj), low duration Balcar FX 60 flash lamps were used as excitation sources. The pulse duration considered here is around 10 ms. Data acquisitions were made at frequency of fps so that maximum number of images are acquired, with sufficient time duration of the experiment. An infrared 14 bits ThermaCAM TM Phoenix camera from FLIR Systems, InSb 640x512 FPA, with Stirling closed cycle cooler operating in the 3-5 nm range was used here for the experimental analysis. 38

53 Figure 2.22 Sample plate drawing with specified type, size and defect position (Susa et al. 2007). Figure 2.23 Experimental set-up (Susa et al. 2007) Numerical Modeling Solving the transient or pulse heat transfer equations provides the theoretical results for temperature evolution of the sample which is subjected to a pulse experiment. The finite 39

54 element method is one of the powerful numerical tools that enable the solution of complex nonlinear and nonsymmetrical mathematical problems which are governed by the partial differential equations such as heat transfer by conduction, convection and radiation with temperature dependant thermal properties of materials involved. In order to solve the given differential equation, a model geometry corresponding to the tested sample was defined and its calculation domain is divided into finite elements that represent base elements on which the equation solutions are found. The mathematical model used as well as the model geometry and mesh are presented next. Mathematical model used For the problem under consideration 3D heat transfer was taken into account. The differential equation to be solved on the model domain yields: T c p ( k T ) 0 (2.22) t The corresponding initial (2) and boundary conditions included heat transfer by convection and radiation from the object surfaces (3) as well as the heat source applied on the front surface during the first 10ms of the experiment (4). These conditions yielded the following equations (Susa et al. 2007): T( x, y, z, t 0) T n ( k T ) h n ( k T ) q conv o ( T h amb amb conv T) C( T ( T amb 4 amb T T) C( T 4 4 amb ) T 4 ) (2.23) Where T is temperature, T amb is ambient temperature, x, y, z are the space coordinates, is density, C is sample surface emissivity, k is the material heat conductivity, h conv is the o r convective heat transfer coefficient, c p is the material heat capacity and t is time. Geometry and meshing The model geometry is defined to correspond to the tested sample for this experimental analysis. All the dimensions considered here in the model were taken from the plate specifications which are described earlier in this paper. The defect size and disposition respects the plate specifications. In this case of the defect type named crushed core where no specification was available with respect to the defect thickness, 1 mm was taken as the 40

55 assumed realistic value that could be expected. The unstructured mesh used consisted of tetrahedral elements. Where larger temperature gradients were expected an adjustment of the mesh parameters is made and permitted a different degree of mesh refinement in those regions. In addition to this, a model geometry scaling is used in this region which enables the large differences in plate dimension proportions to be taken into account so that sufficient mesh refinement was also achieved in the model direction corresponding to plate thickness, much smaller than the two other plate dimensions. Both the model geometry as well as the meshed model used here is shown in the Figure Figure 2.24 Sample plate drawing with specified type, size and defect position (Susa et al. 2007). Model parameters Material properties that were used in this model are either taken from past literature or values specified by producers. In the case of CFRP, the differences in thermal properties are considered into the model with respect to the fiber layout. In addition, for materials with significantly temperature dependant properties, this dependence was also taken into account. Honeycomb properties are determined in a specific way so as to represent the mean value of the materials of which the honeycomb is composed. For this purpose the honeycomb density provided by the manufacturer is used to determine air and aluminum proportions in each of the two honeycomb types. These proportions are helpful to obtain the average properties of two different honeycombs that the sample is composed. The value of the sample surface emissivity coefficients which are used in the model is confirmed by comparing the value of the temperature measured by contact thermometer and the one measured by infrared camera. Moreover, the ambient temperature measured in the room was used in the numerical model 41

56 both as a boundary and an initial condition since it is assumed that the plate was in equilibrium with the environment and therefore at room temperature before the experiment started. Convective heat transfer coefficients used in the model corresponding to values recommended in literature for natural convection in still air environment. The density of the heat flux delivered by the heat source, as well as its shape was adjusted so that the numerical results fit as much as possible with the experimental ones (Susa et al. 2007) Results and Discussions Both experimental and numerical results for different defect types are presented below. Thermograms and surface temperature distribution obtained numerically Two pairs of corresponding sample surface temperature distributions obtained experimentally and from numerical models are shown in the respective Figure 2.25 and Figure Temperature scales are adjusted so that the maximal contrast is obtained for experimental images in order to enhance the visibility of defects. The range of temperatures for the corresponding surface temperature distribution on numerically obtained results was adjusted to correspond to the experimental data temperature scale so that the images can be directly compared. Several observations can be made from the thermograms. First, when earlier thermograms are taken into account, the influence of the honeycomb structure on the surface temperature distribution is noticeable, and thus the honeycomb cells can clearly be distinguished on the images. On the other hand, since in the numerical model the structure is only taken into account via equivalent thermal properties as described earlier, the structure existing inside the panel cannot be reflected in the surface temperature distribution obtained using the model (Susa et al. 2007). The exact location of the points representing defective and non-defective areas in the experimental data should be chosen carefully at the same time, because the thermal print of the structure will significantly influence the extracted temperature decay curves during the time when the honeycomb structure is visible on the surface. Moreover, it is clearly seen that highly non-uniform heating was present in the experiment data. In fact, looking at the early thermograms from the readings, it is concluded that heating is much stronger towards the center of the plate in a horizontal and a bit stronger on the right-hand side of the plate which 42

57 can seen by varying the color intensity this is because of the thermal properties of the denser right hand honeycomb structure. This is clearly seen in results obtained by numerically, though not so noticeable mostly due to the fact that in the model uniform heating over the whole surface was applied. Figure 2.25 Thermograms obtained from pulse experiment on a sample, left - at t=2s and right at t = 20s after the heat pulse has been applied (Susa et al. 2007). Figure 2.26 Surface temperature distribution as obtained by numerical simulation corresponding to t=2s (left) and t = 20s (right) after the heat pulse has been applied (Susa et al. 2007). 43

58 Time evolution of the temperature decay curves The temperature decay curves for defective and non-defective areas both obtained experimentally and numerically are shown in the Figure 2.27 and Figure In Figure 2.27 results for the defect type crushed core are depicted. The larger difference in behavior of the defective area curve is partly due to non-uniform heating and partly to the fact that no exact specification was available on the thickness of the defect. On the other hand, Figure 2.28 shows the results obtained for the defect type core unbound. The overall difference in temperature levels of the experimental and numerical results is again due to non-uniform heating, but despite certain differences, it can generally be concluded that the decay curves exhibit relatively similar behavior (Susa et al 2007). Figure 2.27 Surface temperature decay curves above the defective and non-defective area, experimental and numerical results; crushed core defect; left side of the plate (Susa et al. 2007). 44

59 Figure 2.28 Surface temperature decay curves above the defective and non-defective area, experimental and numerical results, core unbound defect, right side of the plate (Susa et al. 2007). Thermal contrast evolution in time Evolutions of the thermal contrast obtained experimentally and through modeling for each of the four different defect types and two different defect sizes are presented in the following Figure 2.29 to Figure The first two Figure 2.29 and Figure 2.30 are obtained for defects imbedded in the left-hand side of the panel, while the last two Figure 2.31 and Figure 2.32 shows the behavior of the defects on the right-hand-side of the plate. It can be concluded that results of the modeling for defects of type crushed core and core unbound correspond relatively well to the experimental data. More or less the same conclusion can be applied in the case of the delaminations defects, although after just a few seconds following the heat pulse, the thermal contrast obtained was so small that the defects were barely visible, both in case of experiment and simulation. On the contrary, in the case of the extra adhesive defect, the behavior of the defect as obtained from experimental data shows no 45

60 correspondence to the behavior obtained by modeling. Comparing the results obtained by pulse infra thermography (PT) with those obtained by other methods, it was concluded that the defect does not behave as expected when its simulation thermal properties are considered. This fact raised doubt in accuracy of the data with respect to the thermal properties of the adhesive that was used to simulate defects. All these properties were taken into considerations under the specifications given with the plate but they can vary widely depending on the final condition of the adhesive mass once applied to the plate. Figure 2.29 Thermal contrast evolutions in time, experimental and numerical results; crushed core defect; left side of the plate (Susa et al. 2007). 46

61 Figure 2.30 Thermal contrast evolution in time, experimental and numerical results; extra adhesive defect; left side of the plate (Susa et al. 2007). 47

62 Figure 2.31 Thermal contrast evolution in time, experimental and numerical results, core unbound defect, right side of the plate (Susa et al. 2007). 48

63 Figure 2.32 Thermal contrast evolution in time, experimental and numerical results, delaminations defect, right side of the plate (Susa et al. 2007). Finally, to illustrate the impact of the non-uniform heating on the experimental results, and also to point out the difficulties that causes for obtaining experimental data which can truly be comparable to the numerical results, corresponding Figure 2.33 gives the thermal contrast curves for defect type crushed core located in the left-hand part of the plate. Data depicted in cyan and green represents the thermal contrast obtained experimentally for the smaller of the two defects. The only difference is in the choice of the reference sane area points. In general, for all of the defects, the non-defective area has been chosen at the same distance from the center of each defect and from the defect center horizontally in the direction opposite from to plate centre. The distance corresponded to half the distance between the centers of the two same defect types. The same was done to obtain the thermal contrast depicted in cyan. The result (thermal contrasts of comparable maximal value for two 49

64 defects of different sizes) is rather surprising if non-uniform heating is not taken into account. Therefore, in order to obtain the thermal contrast depicted in green, the same reference points were taken for both defects, those in the middle of the defects. Clearly, the smaller defect now shows a lower maximal thermal contrast and thus, the conclusion can be drawn that non-uniform heating has an important impact on the extracted data quality. Figure 2.33 Thermal contrast evolutions in time influence of non-uniform heating, experimental and numerical results; crushed core defect; right side of the plate (Susa et al. 2007) Conclusions The results obtained from modeling complex composite structures with different defects presented. It has been proven that different types of defects have significantly different thermal responses when imbedded in the same structure. Therefore research concerning the possible identification of the defect type by analyzing these specific differences in temperature decay and thermal contrast curves is under way. Non-uniform 50

65 heating clearly represents one of the main obstacles for easier comparison of the experimental and numerical results needed for numerical model validation. Since the quantity of the heat delivered to the sample surface directly influences the temperature decay curves and in addition the resulting thermal contrast, a way to overcome this drawback will be examined next. Since it is not possible to eliminate the non-uniformity in heating, by taking this experimental condition into account, it will be considered as the continuation of this work. 2.6 Advances in Pulsed Thermography (Shepard 2001) Shepard (2001) in this paper clearly explains the use of pulsed infrared thermography advances in the recent years. The use and implementation of pulsed infrared thermography as an NDE solution for manufacturing and in-service applications has increased dramatically in the past couple of years and enables the advances in IR camera and computer technology enormously. However, the pulsed infrared thermographic data analysis and processing has been remained unchanged. These methods include image, averaging, subtraction, decision, slope calculation and contrast methods like peak contrast and peak slope time mapping which are very important for analysis purpose. An Alternative approach is also used for the pulsed infrared thermographic data analysis, based on developing a parametric equation for the time history of each pixel. The resulting synthetic image provides increase in spatial and temporal resolutions, and significantly extends the range of defect depths and sample configurations to which pulsed thermography can be applied. Moreover, it also reduces the amount of data that must be manipulated and stored, so that an entire array of image sequences from a large structure can be processed simultaneously Introduction The growth of pulsed infrared thermography technique in NDE can be attributed to advances in infrared IR camera technology in the recent years, which has provided increased in sensitivity, spatial resolution, and frame rates in commercial cameras that have decreased in size and cost, compared to earlier, less capable predecessors. The rapid evolution of the infrared IR camera has served to increase the domain of applications for pulsed infrared thermography in terms of sample materials, defect depths and sizes. With the increased 51

66 processing speeds in the recent generations the data transfer rates of PC s made faster data analysis and calculation of the results with respect to time. The ability to transfer digital image data directly from the IR camera to the PC in real time, and the ability to perform computationally intensive operations such as the FFT on entire sequences of image data within a few seconds, offers unprecedented opportunity for processing and analysis of the raw data from the IR camera. Figure 2.34 Conventional thermographic images processing of graphite epoxy thermal resolution target Left: The average preflash image is subtracted from a 100 frame average beginning 11 seconds after pulse heating. Center: Slope image of a 100 frame sequence beginning 11 seconds after pulse heating. Right: Peak contrast image (Shepard 2001). Despite of all these important advances of PT, the methods for processing and analyzing pulsed infrared thermographic image data have not evolved at a corresponding pace. After sending a pulse of ray on to the specimen, the sequences of images are collected. The basic operations have to be performed (Shepard 2001): 1) Image averaging: A linear average of a group of consecutive images from the postpulse sequence is displayed. The averaging serves to increase the signal to noise content of the image. 2) Image Subtraction: The average of consecutive images acquired at one time after the pulse is subtracted from an average acquired at a different time. Typical operations include subtraction of the pre-pulse average, or subtraction of a late post-pulse 52

67 average from an earlier one. Subtraction reduces the effort of non-uniform heating and fixed pattern artifacts from the camera focal plane. 3) Image division: The average of consecutive images acquired at one time after the pulse is divided by an average acquired at a different time. 4) Slope calculation: The time rate of change of each pixel over a specified interval of time is calculated and displayed. The slope image reduces non-uniform heating effects and flash lamp reflection from the sample, but it yields a noisy image that improves as a larger time interval is analyzed. 5) Contrast calculations (2-3): A contrast curve is generated for each pixel (by subtracting the time history of a reference pixel from the time history of each pixel) and the peak contrast time, or the peak slope time for each pixel is displayed as an image. Although these methods have been the subject of numerous investigations, commercial application has been limited. In practice, contrast based methods are extremely susceptible to temporal noise in the raw IR data, and the placement of the reference pixel or region Noise Reduction, Analysis and Compression of Raw Data Figure 2.35 Logarithmic temperature-time plot of points on steel slabs ranging in thickness from 0.25 to 1.0. Each plot trace displays nonlinear behavior immediately after flash 53

68 heating, followed by linear behavior, and a deviation from linearity at a time that is correlated to the slab thickness (Shepard 2001). Several factors are involved in the analysis and processing of pulsed infrared thermographic data such as: a) Noises in the IR camera signal; b) Large dynamic ranges of the signal due to the flash pulse; and c) Saturation or nonlinear effects that often occur immediately after the flash. Transformation to logarithmic domains helps to solve these types of problems. Natural logarithm is ideally suited for pulsed infrared thermography. Surface temperature of a semi-infinite sample subjected to an instantaneous heat pulse is shown in the equation below. Q T (2.24) 0.5 e( t) Applying the natural logarithm to the above equation 2.24 for both sides we get. ln( T) ln( Q/ e) 0.5ln( t) (2.25) We can clearly observe in the logarithmic behavior, the first term depends on the input energy (Q) and the effusivity of the sample (i.e., the product if thermal conductivity, density, and heat capacity), and by a second term depends only on time with a slope of -0.5 in straight line. In particular, the logarithmic transform: a) scales the time evolution of each pixel so that the early flash effects are less dominant than they are in the linear domain; b) establishes the behavior of a normal defect-free pixel as a straight line with slope 0.5; c) acts as a low pass filter that suppresses the high frequency temporal noise from the IR camera, leaving the lower frequency thermal events unperturbed. The logarithmic behavior varies with the ideal linear behavior due to poor camera calibration, reflection artifacts and convection. However, the logarithmic behavior exhibits remarkable consistency, the defect free areas are nearly linear in the pixel representation, and pixels corresponding to subsurface defects depart from the near-linear signature at a particular time that is correlated to the depth of the defect (this time is essentially the peak slope time that is detected in time domain contrast curves). The good and bad pixels are observed using the logarithmic behavior and relatively their time evolution. Approximation is done for the logarithmic time dependence of a pixel by a set of orthogonal functions. For instance, a polynomial has been provided for an experimental data, where 54

69 ln( T( t)) o a a1 ln( t) a2[ln( t)] a3[ln( t)] a4[ln( t)] (2.26) In this example, we have specifically used a low order polynomial, as the inclusive of higher orders serves only to replicate noise that appears in the later, low amplitude data, and nonlinearity in the early data. Once the time evolution of each pixel has been approximated by equation 2.26, or a similar function, we can synthesize the original data, since T( t) exp[ ao a ln( t) a [ln( t)] a [ln( t)] a [ln( )] ] (2.27) t There are significant advantages to this process, compares to conventional processing of the raw data: 1) The synthetic T(t) reproduces the true thermal behavior of the signal, suppressing high frequency noise. 2) Calculation of derivatives and detection of inflection points in the time evolution of each pixel can be accomplished algebraically. 3) The time evolution of a pixel can be retreated by using only the coefficients of the polynomial, reducing storage and data handling requirements. 4) It is possible to interpolate the behavior of a pixel at times that are in between actual acquisitions Pulsed Thermographic Image Synthesis Figure 2.36 Comparison of Reconstructed 2 nd derivative (top) and raw 100 frame average (bottom) images of an adhesive disbond in a graphite epoxy panel. Flash reflection artifacts in the raw image are eliminated in the synthetic image (Shepard 2001). A PC software (MOSAIQ TM ) helps in a synthetic data set from raw data can convert a 400 frame, 320 x 240 pixels 12-bit data sequence in approximately 15 seconds, which is considered as extremely fast. The images are directly seen in the synthetic format, or a synthetic time sequence movie can be seen using this software. Of course, it is also possible to perform the conventional processing operations, as described earlier on the synthetic data 55

70 set. However, in most cases, the substantial improvement over these operations is obtained by viewing the 1 st or 2 nd derivatives of the synthetic data. The derivative images significantly reduce artifacts that occur as the result of reflection of flash hardware off her sample surface (the flash hardware cools at a slower rate than the sample, and more or less static compare to changes due to conduction) (Shepard 2001). (a) (b) Figure 2.37 a) Synthetic 2 nd derivative of graphite epoxy thermal resolution target (extracted from same data set as the images in Figure 2.34). (b) Synthetic depth image of the resolution target (Shepard 2001) Simultaneous Processing of Multiple Image Sequences For each pixel the synthetic data sequence is constructed from the coefficient of polynomials and which helps in saving the space on the disk by not saving the large amount of processing and sequential data, for instance the set of coefficients for a 400 frame, 12-bit data sequence from a 320 x 240 pixel camera is reduced to a 4.5 MB data file, compared to 61.4 MB for the raw data sequence. The size of the synthetic file does not depend on the length of the raw sequence. The significant reduction in the amount of data that must be stored and manipulated allows more than one file to be processed at the same time. For instance, results from the inspection of a large structure could be processed and analyzed in a single operation, as opposed to the current time-consuming practice of analyzing each constituent image of the part separately and pasting the results into a graphics program. 56

71 (a) (b) Figure 2.38 (a) Simultaneously processed 1 st and 2 nd derivative images of hollow channels deep in a 4 x 3 Inconel panel. Each image comprises a 5 x 4 image array. The derivative images of the entire panel are created in less than 2 seconds (Shepard 2001). For inspection of large structures that require multiple shots, the actual acquisition process is often accomplished quickly, while the processing and analysis steps are repetitive and far more time consuming. In these cases, the ability to process the entire data set for the part simultaneously represents a considerable reduction in the total inspection time Conclusions The approach that is described allows significant signal to noise improvement, and also data reduction, artifact suppression and the opportunity to perform mathematical operations quickly and without generating additional noise. This approach can be used on digital data acquired directly from the IR camera, or alternately, on analog data that has been digitized with a video frame grabber. The application of the synthesis process we have described provides considerable improvement to the analog 8-bit signals. The use of the synthetic approach results in the removal of much of the noise in the temporal signal. As a result, the fixed pattern (spatial) noise in each image may appear to be more prominent. In fact, we have not amplified the fixed pattern noise; we have simply removed the temporal noise that obscured it. However, the fixed pattern noise can be systematically removed using subtraction, division, or Fourier methods. 57

72 2.7 Detection of Air Blisters and Crack Propagation in FRP Strengthened Concrete Elements using Infrared Thermography (Hu et al. 2002) Hu et al. (2002) in this paper clearly explains the possibility of (1) Detection of air voids between the advanced composite materials like Fiber Reinforced Polymer (FRP) and concrete substrate and (2) Predicting crack initiation and propagation in a reinforced concrete (RC) beam in the initial stages of the failure. Both this studies are carried out successfully with the infrared thermography (IRT). The artificial blisters such as air-voids with known sizes are embedded between the interface of FRP and concrete which can be detected by the infrared thermography remotely up to distance of 20meters from the specimen. The RC beam, which was clean in initial condition, is subjected to continuously static or cyclic loading tests. The preliminary results shows the damaged region of the RC beam, and it is partially strengthened by glass fiber reinforced polymer (GFRP), which was mainly used to cover the cracks in the RC beam, was clearly identified using an IR thermal imaging system. The anticipated failure plane was proven to be identical to the actual failure of the test beam Introduction It is known fact that the care and maintenance of the infrastructure is a main issue of the mere future and a major dispute in the present centuries. The construction materials used in construction and infrastructure can often be regarded as at different stages of integrity or distress, which is due to: (a) prolonged period of use, (b) over-loading at or beyond serviceability limit, (c) flawed initial design, (d) poor workmanship and/or site supervision and (e) aggressive environmental or chemical attacks. The demolition and reconstruction of the constructed structures is impractical, resource intensive, socially disruptive and environmentally unacceptable. So, the alternative way remains the best which is re repairing and making more strengthen and maintain the existing structures and the built environment. Every retrofitting and upgrading activity in the construction materials should be thoroughly checked and should be preceded by an appropriate materials integrity/distress evaluation for a meaningful and cost-effective outcome. Therefore, with such an increasing need in the construction industry, the non- 58

73 destructive testing (NDT) techniques, such as infrared thermography (IR) thermography, is the best and most effective equipment which is being developed for use as predictive and preventive maintenance tools. This research is mainly designed to develop a rich and reliable inspection procedure to monitor the damage in building, bridges and other infrastructures and avoid them from the failure. This will help the practicing engineers to (a) assess the structural performance (b) verify quality assurance in final products and services, and (c) make decision in whether to take preventative strengthening and pre-emptive maintenance action, or allow the structure to become dysfunctional Experimental Program Case Study I Detection of blisters a) Insertion of blisters into CFRP and substrate interface The 1mm thick blisters are made with the help of rings which are cut from plastic piping, and embedded between the interface of Carbon Fiber Reinforced Polymer (CFRP) and concrete specimen. The commercially available pipes have diameters of 16, 18, 20 and 30 mm respectively with negligible variation in respective diameters, see Figure 2.39 Note: Sizes of bubbles, d: diameter blister 1: d = 16mm, blister 2: d = 18mm, blister 3: d = 20mm and blister 4: d = 30mm After preparing the concrete surface, the cut blisters were lightly pressed to the concrete substrate before bonding the laminate and squeezing out unwanted natural airbubbles with a wall-paper type roller. All bonded surfaces are allowed to cure for more than one week before testing. Figure 2.40 shows the Plate 1 which is actual visual image of composite specimens with the embedded artificial voids. 59

74 Front Elevation Figure 2.39 Illustration of embedded blister locations (Hu et al. 2002). b) Detection of blisters between CFRP and substrate interface Under known site conditions, due to the presence of the air voids in the concrete element it may not create a perfect bond between the FRP and results in the poor workmanship and it also lessens the strength of the concrete element. The relevant studies have been reported in the references by Andreou et al. (2000), Delpak et al. (2001), Shih et al. (2002). The air has lower thermal conductivity than concrete (or the cement paste) after the usage of detection method. Therefore, a heated section of laminated FRP, backing to a blister will remain warmer due to reduced thermal conduction. The thermographic images were captured through an IR camera using Active Thermographic Approach (ATA) when the element is still in the warmer condition, which requires an external thermal perturbation to stimulate thermal distribution in the objects, Maldague (1993) and Hu (2000). The Active Thermographic Approach (ATA) is achieved by deploying either radiant heat, from powerful electric light bulb/lumps, or electric resistance heating elements attached to the bonded FRP surface. 60

75 Figure 2.40 (Plate 1) CFRP plate concrete specimens, where the artificial blisters are embedded between the interface of composite and substrate (Hu et al. 2002). (a) Image taken at the distance of 2.5m (b) Image taken at the distance of 10m 61

76 (c) Image taken at the distance of 2.5m (d) Image taken at the distance of 10m Figure 2.41 (Plate 2) Thermal images taken at various distances for the concrete samples plated by CFRP (see Plate 1 for the corresponding actual photograph) (Hu et al. 2002). Plates 2 (a)-(d) show the thermal image of the concrete beams, strengthened by CFRP materials, taken at different distances. Case Study II Prediction of crack initiation a) Preparations of experimental set-up A 100 x 200 x 1200 mm concrete beam was reinforced in tension with three T10mm bars and laminated partially by Glass Fiber Reinforced Polymer composite (GFRP) material. The GFRP sheets were applied through wet lamination using a two-part epoxy based adhesive. In addition, the reinforced concrete (RC) beam was placed on the jack and subjected to a 3 point loading condition, as shown in Figure The load is gradually increased up to a certain limit of ultimate load and then continuously decreased through cyclic loading with a specified time period. In this study the load peak-to-peak amplitude was set to be 20% of static load. The frequency of vibration was set at 3Hz. The displacement readings at test beam center, were recorded after completion of each phase of static and cyclic loading sequence, see Table

77 (a) Elevation (b) Cross-section Figure 2.42 Configurations of test beams (Hu et al. 2002). Table 2.1 Displacements of 1.2m RC beam subjected to static and cyclic loading tests. b) Laboratory thermographic measurements The thermal imaging system used for this part of work, CEDIP, for which the accuracy of the measurement performance is about 1 0 C, 1%, has a thermal sensitivity of 0.02 o C at 25 o C. A series of thermal infrared images were taken during the each phase of loading in order to identify the potential failure area through various stages of thermographic monitoring process, it was expected to record the generated heat through friction and fretting between cracks. The Passive Thermographic Approach (PTA) was therefore adopted for this 63

78 study, (Hu et al. 2001). The natural heat was expected to be release at the locations where the defects exists; therefore, there is no need of additional thermal stimulation Thermographic Results and Observations Comparison of blisters sizes from measured and IR images (Case Study I) From the thermal image in Plate 2 (a) to (d) shown in Figure 2.41, it is clearly possible to locate the positions of the artificial blisters where the areas with higher temperature existed. Even though the distance between the surface of samples and the transmission line of IR lens is increased up to 20m; the locations of air-voids covered by CFRP were still identified using the infrared IR camera. If the bond line between CFRP and concrete (excluding blister locations) has no air voids, four circular shapes or patterns with higher temperature are expected to be observed, see Figure However, the distorted hot areas have appeared and registered by images displayed in Plate 2 (a) to (d). This is due to the bad workmanship which results in the undesired thin gap around the blisters between CFRP and the concrete. This was confirmed after the removal of CFRP plates from the beams as seen in Figure 2.43 Plate 3. Figure 2.43 (Plate 3) Actual photograph of concrete samples, after laminated CFRP (see Plate 1 for its corresponding photograph) (Hu et al. 2002). 64

79 The specimens with the areas devoid of adhesive (highlighted by a marker pen), where the external heat flow supplied from the surface is supposed to be trapped, are clearly identified. By comparing the actual photograph shown in Figure 2.43 with the thermal image represented in Plates 2 (a) to (d) shown in Figure 2.41, the potential of thermographic technique in detecting the defect locations in FRP strengthen specimens/ structures is justified. Prediction of failure plane in RC concrete beam due to cyclic loading tests (Case Study II) The continuous cyclic load is expected to accelerate the formation of cracks, and hence initiate the generations of the heat through the fiction at the areas, where the discontinuities are likely to be formed. Thermal images taken during the cyclic loading level of kn is shown in Plate 4 i.e., Figure 2.44, where the area with triangular shape, highlighted by a white circle, indicates the dissipated energy due to the friction of the crack tips. From this image, it is predictable that the failure would take place within the region with the higher temperature and the possible failure plane would be about 45 0 along the triangular plane. Plate 5 shown in the Figure 2.45 has a thermal image of the test sample taken immediately, after the failure has occurred. It is observed that the actual failure plane shown in Figure 2.46 which is Plate 6 is almost identical to the predicted one anticipated in Plate 4. 65

80 Figure 2.44 (Plate 4) Thermal image taken during 45 to 55 kn cyclic load. Figure 2.45 Thermal image taken after failure occurred (at the cyclic load range of kn) (Hu et al. 2002). 66

81 Figure 2.46 (Plate 6) Actual photograph for 1.2m concrete beam laminated by GFRP taken after failure occurred (Hu et al. 2002) Conclusions After conducting the many experiments in the current studies, the following conclusions can b summarized: Depending on the specifications of infrared IR thermography, the blister locations between GFRP and concrete substrate could be identified remotely from a distance of up to 20m. Accurate detection of poor workmanship in wet-lamination and bonding processes of composite materials to concrete soffit was possible by using infrared thermography. The size of the blisters could also be estimated if the distance between the IR thermography and the surface of the object is available. The dissipate energy (due to dysteretic action e.g. cyclic loading effect), between the potential failure planes, was identified using IR thermography. The region of concrete beam, which had been strengthened by GFRP, was identified by IR thermography immediately prior to fracture due to cyclic loading. Subject to future trails, the current thermographic techniques can provide an unparalleled opportunity in locating concealed cracks as yet invisible to unaided eyes. 67

82 It is clear the IR thermographic techniques can provide the maintenance engineers with both quantitative and qualitative information relating to the state of structural damage. 2.8 Detecting of Defects in Polymeric Materials using Pulsed Infrared Thermography (Szczepanik et al. 2008) Purpose: The aim of this paper is to determine the possibility of the use of Non-Destructive infrared Thermography (NDT) technique to detect the defects in polymeric materials and steel and compare their results. To one of the non-invasive technique which is Infrared Thermography (IRT) is used for monitoring the temperature changes in the material during cooling down or heating processes, which are determined by measuring the infrared emission from the surfaces of the materials. In this paper the subsurface defects of specimens are made from polymeric materials such as PE, PMMA, laminate experimentally detected and directly displayed by thermographic images are presented. Design/methodology/approach: Szczepanik et al. (2008) developed a real-time noninvasive technique using pulsed infrared (IR) thermography for measurement of the temperature of polymer materials and described. In this study 16 specimens were heated during specific time using infrared lamp as the heating source. After the heating is specimen for couple of minutes, the specimen s surface temperature was measured during cooling down process by thermovision camera; next defects were detected by means of thermographic images analysis with infrared thermography. Findings: The experimental results have demonstrated that radiation heating and thermographic images analysis is effective method for revealing defects in the polymeric materials. Research limitations/implications: It is not possible to detect defects at a long time of heating of researched material because it results in uniform temperature on whole surface of specimen. 68

83 Practical implications: It is possible to detect subsurface defects in polymeric materials by infrared thermography method. It is possible to see the defects on thermographic image, but the determination of their geometry and position is restricted and not very precise, it requires specific skills and as well as long labor consuming attempts. The specimen s area with defect show higher temperature than area without defect also cooling down process proceeds longer in the area with defect Introduction To detect the defects in the polymer material specimens there are many different nondestructive testing techniques (NDT), but so far ultrasounds occupy one of the leading places (Wrobel et al and 2007). The usage of thermography technique is increased enormously in the present days. The Infrared thermography has applications in ecology, medicine (for cancer testing), rescue, and buildings, it also helps in observing thermal process and in material testing and also in monitoring of manufacturing and transforming process in casting. The main purpose of non destructive testing (NDT) technique is to determine defects of various type and size and their properties at unknown depths. It is not possible with one technique to detect the defects in the specimen thus various techniques are used. Taking into account the way of thermal process activation thermography is classified into two categories: Passive thermography, where objects outside temperature distribution and changes are observed without observer interference Active thermography which consists in observing of researched object s thermal answer to external thermal impulse stimulation being a function of time and in registering this answer by means of thermograph. There are several types of thermography depending on method of thermal activation Pulsed thermography, Lock-in thermography with modulate heating, Pulsed phase thermography. 69

84 The term thermography and thermovision include testing methods based on registering infrared part of radiation spectrum emitted by body which then is converted by special camera into a color map of temperature. The thermovision system allows to measure temperature remotely and in many places at once. Therefore, the suitability of infrared thermography (IRT) to nondestructive testing depends on the ability in detecting the temperature variation or thermal contrast induced by a defect in the structure. So, the defect visibility depends on several factors which involve the defect geometry (mainly defect diameter and position in testing material), the relative thermal characteristics (e.g. thermal conductivity, thermal diffusivity) between the defect and the host material, the way thermal stimulation is performed and the sensitivity of the infrared imaging system used. The presence of a defect at a certain depth interferes with the heat flow causing local surface temperature variations Experimental The main aim of this research is to determine the defects in polymeric materials and steel specimens using non-destructive thermographic testing (NDT). Materials Thermographic research was carried on twelve polymeric materials specimens and four steel specimens. In compare purpose holes with the same geometry were made in each specimen at known depths and distances. The photograph of four specimens is shown in the Figure Figure 2.47 Specimens for Testing Szczepanik et al. (2008). The specimens in the form of rectangular prism with dimensions of 23.5 x 35 x 150 mm and with through holes with different diameter placed in distance of 3 mm from searched 70

85 surface were tested. For good heat emission, all specimens are uniformly painted with black matt coating on the surface. Methodology The specimens are tested using Pulsed infrared thermography (PT) technique. This technique is used in the research belonging to activate thermography (AT) with static thermal activation. To carry out this testing a special stand is prepared which is shown in Figure This stand includes infrared radiator (IR) (Victory Lightning) (1) insulating shield in the form of frame (2) with specimen mounted (3) and thermovision ca) Specimens were activated by short time heating. After applying short thermal impulses on the surface of the materials, the infrared thermography technique helps in detecting the thermal changes during the mera (4 cooling down process. Specimen s surface temperature during cooling down process was observed in areas above defects and in areas without defects. The specimens are located at a constant distance of 80 mm from radiation source. Together with the end of the heating process the surface temperature recording procedure begun Results and Discussions Thermal imaging technique helps in obtaining the surface temperature changes. It is not possible to present all the thermographic images and temperature measurement results at a time so only chosen examples which is shown in Figure 2.49 presents the thermographic image taken 12 seconds after the end of heating process id done. The range of temperature between the defect and defect free surfaces is nearly about five degrees. The subsequent photographs which are listed in Figure 2.50 to Figure 2.53 shows thermographic images registered after 60, 150 and 2 seconds after the beginning of temperature measurement for the different materials and geometry specimen by range two degrees. 71

86 Insulation shield with specimen (a) Infrared Radiation (IR) (b) Thermovision camera Insulation shield with specimen Figure 2.48 Schematic draw of thermovision research stand: (a) Specimen heating process and (b) Observing specimen s surface by thermovision camera (Szczepanik et al. 2008). 72

87 Figure 2.49 The thermal of polyethylene specimen registered twelve seconds after the end of heating process the first signs of subsurface defects (Szczepanik et al. 2008). Figure 2.50 The thermal of polyethylene specimen registered sixty seconds after the end of heating process clearly visible defects manifested by temperature incensement (Szczepanik et al. 2008). 73

88 Figure 2.51 The thermal of polymethacrylate (methylate) specimen registered one hundred fifty seconds after the end of heating process (Szczepanik et al. 2008). Figure 2.52 The thermal image of laminate registered one hundred fifty seconds after the end of heating process (Szczepanik et al. 2008). 74

89 Figure 2.53 The thermal image of steel specimen registered two seconds after the end of heating process (Szczepanik et al. 2008). Figure 2.54 shows the temperature changes of the heated side of the specimen at, area with defects and area without defects. For every specimen these curves are prepared. Observed dependences allowed to distinguishing initial stage of temperature decrease almost linearly with time. The speed of cooling is bigger in area without defect than area with defect for all specimens. It is the result of bigger thermal resistance of the defect then tested material. (a) 75

90 (b) (c) Figure 2.54 The dependences of temperature on cooling time for areas with and without defect. (a) Polyethylene (b) polymethacrylate (methylate) and (c) laminate (Szczepanik et al. 2008) Conclusions The experimental results have been demonstrated that radiation heating and thermographic images analysis is effective method for revealing defects in the polymeric materials. It is possible to see the defects on thermographic image, but the determination of their geometry and position is restricted and not very precise. It requires specific skills and as 76

91 well as long labor-consuming attempts (tests; probations). The specimen s area with defect show higher temperature than area without defect also cooling down process proceeds longer in the area with defect. It is not possible to detect defects after long time of heating of researched material because it results in uniform temperature on whole surface of the specimen. 2.9 Active Infrared Thermography applied to Detection and Characterization of Non Emergent Defects on Asphalt Pavement (Dumoulin et al. 2002) Dumoulin et al. (2002) in this paper explains about the Pulse Thermography analyses which are helpful in detecting the non emergent defects in asphalt concrete which is used for road pavements. Experimental and numerical experimentations were run on this specific heterogeneous material. Only experimental data acquired with an un-cooled microbolometer camera were used. Data were processed using a semi-infinite heat transfer model in order to determine the depth of the defects. Finally, a discussion on the influence of performances of the IR camera employed versus potential detection of subsurface defects is finally proposed in correlation with the investigated domain of pavement materials Introduction For the diagnosis and the maintenance of the French roads and bridge networks different non destructive testing techniques are used and for pavement structure it s currently under investigation. These methods help in detecting the hidden defects from the surface. Indeed, defects such as un-sticking zones between the top layer and the structural ones could induce a quick deterioration of the pavement surface. When the defect is punctual, or is located on a small road stretch, some efficient repair might solve the problem. Sometimes, it involves a larger number of kilometers, and then renewal of the road is the solution. So, their detection as soon as possible has therefore a great importance and is an important challenge. It might imply significant cost savings and avoid major traffic disruptions. There is a large variety of materials used for road construction, from asphalt to cement concrete. They have different granular constitution and therefore different thermal and radiative properties. These variations could also appear inside the same layer due to construction constraints. Such 77

92 heterogeneities would induce additional difficulties in the use of active thermography. Active infrared thermography for the detection of defects has been now used for many years for non-destructive control of materials such as metals, composites and so on, as described in the literature (Maldague, 2001). Its application to civil engineering materials like cement concrete, slightly porous and almost homogeneous was shown in (Maierhofer, Ch., 2006). First extension of such approach for bitumen concrete material was presented in (Marchetti, M., 2008), where the challenge was to sort relevant signal of defects among the pristine porosity and heterogeneity of such a material. So, in this paper we present Pulse Thermography analysis conducted both on experimental and numerical experimentations applied to porous and heterogeneous asphalt concrete pavement materials. We also present a model for depth retrieval of subsurface defect. Finally, combining numerical simulations with experiments allow us to discuss on the influence of the sensitivity of the IR detector used onto the potential detection of sub surface defects Laboratory Test and Numerical Simulations A description of the experimental test bench is given in this paragraph. It is followed by one of the numerical simulations undertaken in the case of the two defects sample. Experimental apparatus a) Bitumen Concrete samples During this study, active infrared thermography is used to detect the defects in the laboratory frame. Semi-granular asphalt pavement materials were considered. Their nature was of the most commonly used material either on national roads and highways. Samples were made of granular materials with a bitumen matrix. They consisted in parallelepipeds with following dimensions 10 cm x 18 cm x 50 cm. Defects made of wood were included while manufacturing the road samples. During the thermal analysis process, these defects were either left or removed. The defect shapes were a parallelepiped and a pyramid. They were located at different depths under the samples surfaces (1.3 cm for the parallelepiped, 1.3 cm and 6 cm for the pyramid). These inclusions were inserted into the road pavement samples in such a way they should not thermally affect each other which is shown in Figure

93 (a) Figure 2.55 (a) View of pine wood defects (b) Sample front face viewed (Dumoulin et al. 2002). (b) b) Bitumen Concrete samples The experimental setup consists of two halogen lamps of 500 W each. The heating phase time ranges between 60 s to 1 hour. A reflector is used to get a constant flux density over entire sample surface which is constant during the trials and It ranges between 3000 W.m -2 (short pulse duration 60s to 300 s) down to 200 W.m -2 (long pulse duration 30 minutes to 1 hour). The infrared camera was located at a distance of 0.8 m from the sample surface consistent with its optical characteristics having a focal length 36 mm. In the present paper, the results obtained with a FLIR S65 camera equipped with an un-cooled microbolometer FPA detector of 320 x 240 sensitive elements in the spectral bandwidth μm are discussed and analyzed. c) Experimental temperature fields evolution Infrared images were acquired at a frequency of 1 Hz. Higher power density was used for lower pulse duration (1, 2, 3 or 5 min). Lower heat density was used for longer time experiment coupled to longer pulse duration (30 min up to 1 hour). Associated infrared images were acquired during half an hour up to two hours. Tests have shown that defects close to the surface could be directly detected during the heating phase with the help of infrared images. It was in the case of parallelepiped one with a flat surface parallel to the heat flow. But, for pyramid one no direct detection was observed. The heterogeneity of the pavement samples did not significantly affect the global heat transfer within the structure. Nevertheless, thermal behavior of aggregates had its own thermal signature. Some heating heterogeneity was observed on sample edges and has to be connected to natural convection development during thermal relaxation phase. Figure

94 presents infrared thermograms (temperature are shown in Kelvin), acquired with the FLIR S65 camera, at the end of a heat pulse of 300 s and 510s farther. The heat flux density applied to pavement sample was round 3000 W.m -2. at t = 300s at t = 510s Figure 2.56 Experimental thermograms for a 300s step heating duration (Dumoulin et al. 2002). It can be observed on the infrared thermogram that the natural convective heat transfer development at the surface of the sample is non homogeneous. Aggregates thermal signature can also be observed for infrared thermogram at t = 300s. Numerical Simulation a) Numerical Simulation description The modeling part consists of applying a step function of 2620 W.m -2 heat flux density s for short duration pulses (60s) and of 220 W.m -2 for long pulse duration (3600 s) to the front face of the sample. For these numerical test case temperature field on the surface and inside the sample as a function of time were computed. The geometry of the sample considered is corresponding to the sample containing two defects. The thermal characteristics of materials, used for numerical simulations are presented in Table 2.2. Furthermore, ambient temperature was considered as constant (equal to 20 C), and the sample supposed to be submitted to a global heat exchange coefficient h = 10 W.m -2.K -1 on its front and rear faces. Lateral faces are insulated. Table 2.2 Thermophysical properties of materials used for numerical simulations (Dumoulin et al. 2002). 80

95 b) Numerical Temperature fields evolution Figure 2.57 shows temperature field (in Kelvin) evolution for a density heat pulse of 2620 W.m -2 during 300 s and thermal relaxation. Due to the good homogeneity of material used for computation, the aggregates and natural convection did not affect the temperature field as previously observed for experimental data. at t = 300s at t = 510s Figure 2.57 Infrared images simulated for a 2620 W.m-2 heat flux density pulse during 300s (Dumoulin et al. 2002) Results and Discussions Results obtained on simulations and experiments are presented and discussed in the case of a step heating of 300s. Pulse Thermography analysis To compare defective and non defective areas, contrast maps are used and are calculated by different approaches (Maldague, X. 2001). Then in a second step, knowledge of their positions is generally used for the depth calculation approach. Here we assume that the location of the defect was not known and we compute overheating maps evolution using the temperature map acquired before the step heating at t 0 for each pixel shown in the equation 2.28 of the infrared thermogram which is a conventional approach in wall heat transfer determination by IR thermography (Dumoulin et al. 1995) also known as cold image subtraction (Maldague, X. 2001). Tx, y ( t) Tx, y ( t) Tx, y ( to ) (2.28) So at this stage a set of overheating maps are determined and will be used for depth retrieval using the heat transfer model presented hereafter. 81

96 Defect depth retrieval In a first approach heat transfer in pavement materials can be assumed to behave like heat transfer in semi-infinite bodies. A care must be taken at the thermal diffusion time when making calculation and experiments in order to keep within the validity domain of such model. Furthermore, by neglecting natural heat transfer on the surface of the inspected material, the heat equation system to solve takes the following form z a t with T( x, t) T (2.29) o Boundary Condition: t 0 : ( z, t) 0, t 0 and z 0 : o ( t) z In our measurement and simulation configuration the surface solicitation is a constant square heat pulse of duration. The boundary condition at t > 0 (shown in equation below) takes the following expression where the constant heat flux density applied during the pulse: ( t) if t and ( t) 0 if t (2.30) o q o o The solution of such system can be obtained by using the Laplace transform. Other resolution approach can be found in literature. The solution for the surface temperature is reported shown in equation 2.31 below. If 2q0 t t : (0, t) and if b 2q0 ( t t ) t : (0, t) (2.31) b Combining this solution for the case of the heat pulse solicitation over the surface of the studied material with the effusivity approach proposed by Balageas in Characterization and non-destructive testing of carbon-epoxy composites by a pulsed photo thermal method the depth of the defect can be determined using the relation reported shown in equation below. z a t (2.32) def 0.95 min ( b n,min ) Where a is the thermal diffusivity, b the thermal effusivity (calculated using equation 2.31), tmin the time when the effusivity curve is minimum, zdef the depth of the defect in m and bn,min the normalized minimum effusivity. 82

97 Figure 2.58 presents results obtained with numerical simulation and measurement data for a step heating duration of 300 s. Depth obtained for the parallelepiped is in the range of 1.2 to 1.4 cm for numerical data, compared to 1.1 to 1.3 cm for measurement data. Natural convection observed for experimental data coupled with limited performances of the uncooled camera used, drive to more inhomogeneous depth map. As expected, direct calculation approach is more sensitive to these perturbations than it could be by an inverse approach using a regularization scheme. Figure 2.58 Square defect depth maps in meter: (a) simulation and (b) measurement (Dumoulin et al. 2002). Detection of the second defect from experimental data is not as obvious due to its poor thermal signature masked by the natural convection development observed during measurement and aggregate thermal behavior. Nevertheless, the interest of such approach is that by establishing a map of computed depth we can also make a segmentation to obtain the defect localization and, at the same time, to determine one of its dimensional characteristics Conclusions The parallelepiped defect located at 1.3 cm under the surface was easily detected even with a short heating phase (test were made down to 60s). The thermal behavior of the structure was greatly affected by the defect presence around it. The nature of the samples (porosity, heterogeneity) did not seem to affect the ability of the technique to non destructive control of such road material. Nevertheless, natural convection observed during experiments has a thermal signature that hides the one of the pyramidal defect. The proposed semi-infinite model with the constant heat pulse solution coupled with normal effusivity approach from 83

98 Balageas drives to a correct estimation of the square defect depth. The use of Principal Component Thermography method is useful to quickly detect the defect location using its spatial signature on empirical orthogonal functions map. In the future, we plan to implement more refined models using a segmentation of EOF maps in order to associate the most adapted model to pixels of the infrared image. Finally, this study has also shown the ability of un-cooled camera with less thermal sensitivity to detect defects, although care must be taken in the reduction of measurement noise NDE of Composites Delamination by Infrared Thermography (Songling et al. 2003) This paper briefly presents the principle of infrared thermography. And delamination defects of honeycomb aluminum composites which were inspected by infrared thermography technique. Testing results shows that infrared thermography is a rapid, effective nondestructive evaluation method for delamination defects of composite materials. The optimal testing time is different for different defects. And the testing speed was higher than two screens per minute for all delamination defects of honeycomb aluminum composites. Rapid automatic heating modes, automatic synchronous scanning of heating source and infrared detector defect identification system will be the important developing aspects of infrared thermography testing technique Introduction Composite materials are composed of two or more different materials to get higher integrated performances. It is difficult to accurately control the technical parameters in manufacturing and leads to instability and larger dispersing of the quality of composite materials. Dead loads, mechanical damage, fatigue and over heat could induce destruction in the use of composite materials too. So it is necessary to nondestructively evaluate composite in the process of manufacturing and using. NDE techniques of composites usually include ultrasonic method, acoustic emission testing, X-ray penetrating testing, surface penetrating method and infrared imaging. These methods have different characteristics. Acoustic emission testing needs to displace composites in a certain pressure environment. It is unpractical for most finished products and 84

99 its testing results only give qualitative data about strength. Surface penetrating could only find surface opening defects and could not find delaminating defects of composites. Ultrasonic method and X-ray penetrating testing are two important techniques for composites NDE [3]. Ultrasonic method, especially ultrasonic C-scan, could find delaminating, loosening, porous and most other dangerous defects of composites reliably [4-6], but its testing efficiency is low. X-ray penetrating testing could find 1%-2% variation of depth, but it is not sensitive to delaminating defects, which usually appear in composites. Infrared thermography, which is untouched, large area scanning and fast, will be a fast method for composites defects testing Principle of Infrared Thermography The basic principle of infrared thermography is that discontinuous defects of inspected objects will influence the heat conductivity of composites and induce local temperature differences [7]. If we detect these temperature gradients, defects inside objects will be concluded. Infrared thermography method can be classified as follow: 1) Active infrared method: Its characteristic is that conducting heat into detected objects by an external heat source and defects will be detected by judging the heat radiation distribution of object surface using infrared imaging equipment. 2) Passive infrared method: It uses the heat radiation of objects and has no external heat source. Active infrared thermography is widely used in engineering application. When even heat stream is conducted into a plate surface, three-dimension heat conduction could be simplified to two-dimension one according to the isotropy of plates. The law of two-dimension Fourier heat conduction is: 2 2 T T T K z ( y) K z ( z) C 2 p (2.33) 2 z y t where K z, K y are heat conducting coefficients of heat stream direction and its perpendicular direction, respectively. is the density of inspected material and C p is its specific heat. Figure 2.59 is the sketch map of heat conduction in a plate. Since the area of side boundary is much less than the areas of upper surface and undersurface, we can suppose that the boundary condition is T y 0 at y = 0 and y = D positions. 85

100 The input heat stream is given by: T Q in K z (0) (2.34) z z 0 Figure 2.59 A sketch map of heat conduction (Songling et al. 2003). The upper boundary condition could be described by the third boundary condition based on Newton cooling formula. K z T z z h T T ) (2.35) ( h a where a is heat exchanging coefficient of the plate and environment, T h is the temperature of upper surface of the plate, T a is the temperature of environment. According to formula (1) and the boundary conditions above, heat stream densities inside the plate could be gotten and defects distribution inside the plate could be concluded according to the temperatures of the plate surface Experiments Infrared thermography system Figure 2.60 shows the infrared thermography system used in experiments with resolution C. 86

101 Figure 2.60 The block map of infrared testing system (Songling et al. 2003). Specimen Specimen was a honeycomb aluminum composite which was largely used in aeroengineering. The cover skin of the specimen was 0.5 mm, honeycomb was hexagon whose sides were 5 mm, and the length of honeycomb was 25 mm, the thickness of honeycomb was 0.1 mm. four kinds of artificial defects were made to simulate delaminating phenomena between cover skin and honeycomb. These defects were made by 16 mm diameter, 20 μm depth polyethylene slices filling between the cover skin and honeycombs. They were discarding glue film and 2 mm honeycomb, filling between the cover skin and glue film, filling between glue film and honeycomb, pressing honeycomb 2 mm respectively, and were numbered No.1, No.2, No.3, No.4, from left to right. The distances between these defects were 100mm. Testing Process The defects side of the specimen was heated equably by a 1000-watt electric rug for 30 seconds. And then the specimen was placed 1.5 meters to infrared detector. Infrared emission coefficient was set to Testing sensitivity and temperature limits was adjusted according to the temperature of specimen surface and its temperature range. The two parameters were adjusted along with the falling of specimen temperature and corresponding infrared images were noted. 87

102 Results and Discussions Error! Reference source not found. and Figure 2.61 to Figure 2.62 show the testing results. The sampling time in Error! Reference source not found.means the time from topping heating to sampling image. Testing sensitivity of defects lies in temperature gradient of defects ΔTM and relative temperature gradient of defects ΔTM/Tb, where Tb is the mean temperature of specimen. Defects will appear more obviously if ΔTM and ΔTM/Tb grows large. Table 2.3 Testing results (Songling et al. 2003). Defect type No.1 No.2 No.3 No.4 Sampling time (minute) Sensitivity SN ( o C) ΔTM ( o C) ΔTM/Tb

103 Figure 2.61 Defect temperature gradient curves (Songling et al. 2003). Figure 2.62 Defect relative temperature gradient curves (Songling et al. 2003). 89

104 Figure 2.61 and Figure 2.62 are curves of temperature gradient and relative temperature difference of defects. Figure 2.61 shows that the best testing time of No.1, No.2 and No.4 a defects is 11 minutes, and that of No.3 defect is 14 minutes. The best testing time of defects is slightly longer from relative temperature gradient in Figure Figure 2.63 is the infrared image of the specimen collected by PC. A line of hole defects appears apparently on the bottom of the image. These defects could be detected while the specimen was not heated and placed in the air for half a minute. A slight pressing mark produced in the production of the specimen was also shown in Figure 2.63 (a). The upper part of the Figure 2.63 (b) detected another large natural defect. (a) The infrared image of 0.5 minutes (b) The infrared image of 10.8 minutes Figure 2.63 Infrared images of honeycomb aluminum composites (Songling et al. 2003) Conclusions Experiments of delamination defects of honeycomb aluminum composites showed that infrared thermography technique could inspect these defects effectively. Testing results showed that infrared thermography was a rapid, effective nondestructive evaluation method for delamination defects of composite materials. And it is easy to produce temperature gradient in testing process. The position and size of defects display directly on the infrared image and testing image could store in computer for further disposal. The optimal testing time for different defects was different. And the testing speed was higher than two screens per minute for all delamination defects of honeycomb aluminum composites. Rapid automatic heating modes, automatic synchronous scanning of heating source and infrared detector, and intelligentized defect identification system will be the important developing aspects of infrared thermography testing technique. 90

105 3 INFRARED FIELD TESTING This chapter presents the infrared thermography based debond data for several railroad bridges with GFRP wrapped piles and pile caps. The chapter also includes laboratory and field testing conducted on rehabilitated railroad ties. 3.1 Field Testing and Evaluation of Timber Railroad Bridge Components Using Infrared Thermography Introduction Many of the timber railroad bridges in the US are over 50 to 100 years old and the degradation is visible in some of its components like piles, joints and pile caps. Fiber Reinforced Polymer (FRP) composite wraps are now widely being used in rehabilitating such deteriorating timber bridge components. The timber components are being wrapped with these composite materials to increase their strength, stability, load carrying capacity and serviceability. The structural integrity of such components relies on proper bonding between wraps and the underlying timber components. As the bridges age and the axle loads of freight cars increase, the need for rapid bridge inspection and strength evaluation methods increases as well. Detection of these subsurface debonds can be useful for taking remedial action such as injecting resin in defective areas. Section 3.1 presents the field experiments conducted on the timber railroad bridges located in Moorefield, West Virginia. Some of the components of the bridges (timber piles and pile caps) were previously rehabilitated with Glass Fiber Reinforced Polymer (GFRP) composite wraps. Infrared Thermography tests were conducted to detect any debonds present at the interface between the GFRP composite wrap and the underlying timber component. Discussions on the infrared testing equipment, location of the bridges, field setup, infrared tests and results, and difficulties encountered during the field study are included in the following sections. 91

106 3.1.2 Infrared Testing Equipment The infrared testing equipment used in the field is described below. Infrared cameras Two different infrared camera models were used in this study. These cameras are described below. (a) ThermaCAM S60 Infrared Camera The digital infrared camera that was used for field testing is ThermaCAM S60 by FLIR Systems. Figure 3.1(a) shows the picture of the Infrared camera. Figure 3.1(a) Picture of the ThermaCAM S60 infrared camera. The camera measures the infrared radiation emitted from an object and converts it to an equivalent temperature value in accordance with the Stefan-Boltzmann law. The thermal images that the camera produces can be directly recorded on a computer. The camera is a lightweight one with a built in 24 o lens. It can detect infrared radiation in the spectral range of 7.5 to 13 microns. The different temperature ranges that the camera offers are -40 to +120 o C (-40 to +248 o F) and 0 to +500 o C (+32 to +932 o F). The current experiments were all conducted with the first temperature range (-40 to +120 o C) since it gives better resolution for temperatures under 100 o C. Thermal sensitivity of this infrared camera at 30 o C is as low as 0.06 o C. The imaging performance for the camera has a spatial resolution of 1.3 mrad and the infrared image capture rate can go as high as 60 frames per second. It is possible to capture and store images on a removable flash card or using a laptop computer. The laptop controlled 92

107 mode also provides a more user friendly interface to control the data acquisition using the infrared camera. The images that the camera produces can be analyzed either in the field by using the real-time measurement markers built into the camera software, or in a computer using FLIR Systems ThermaCAM Researcher software (FLIR 2004). The software is used along with the camera and helps to acquire live infrared images in the laptop computer through the camera interface. The analysis of the images can be made with the various analysis tools like isotherm, spotmeter, area and line. The temperature corresponding to any point can be obtained by using the spot temperature measurement option offered by the software. The area feature provides average temperature over a small area and has the advantage of minimizing the random noise associated with the various pixels. (b) InfraCAM SD Infrared Camera Figure 3.1(b) shows the picture of the InfraCAM SD infrared camera manufactured by FLIR systems. The camera measures the infrared radiation emitted from an object and converts it to an equivalent temperature value in accordance with the Stefan-Boltzmann law. The thermal images that the camera produces are directly saved on a SD memory card which stores thousands of images in standard radiometric JPEG format. With the new SD card, you are no longer tethered to your computer. Also, this camera comes with a much lower price tag (~$3500) which makes it lot more affordable than the $50,000 price tag for high-end infrared cameras (e.g., ThermaCAM S60). Figure 3.1(b) Picture of the InfraCAM SD infrared camera. 93

108 The InfraCAM SD infrared camera is the lightest infrared thermal imaging camera that is currently commercially available and weighs just 1.21 pounds. The camera is Dust and Splash proof. The camera has a built in 24 o lens and meets IP 54 standards and withstands harsh industrial environments. It can detect infrared radiation in the spectral range of 7.5 to 13 microns. The camera is capable of making temperature measurements in the range of - 10 C to +350 C (+14 F to +662 F). The thermal images of InfraCAM SD are clearly displayed on the large 3.5 color LCD with 240 by 240 pixels (the actual detector is 120 by 120 pixels). The minimum focus distance of the infrared camera is 0.3m. Thermal sensitivity of this infrared camera is 0.1 o C. It is possible to capture and store images on a removable flash card. The images that the camera produces can be analyzed either in the field by using the real-time spot measurement marker built into the camera software, or in a computer using FLIR Systems Quickreport. The spot temperature measurement option offered by the software enables temperature measurement corresponding to any point. The area feature provides average temperature over a small area and has the advantage of minimizing the random noise associated with the various pixels. Heating source The shop heater shown in Figure 3.2 was the main heating source used in the field. This heater could be operated in two settings 750 W and 1500 W. To impart high energy in a short time the 1500 W with maximum level setting was used during the field testing. The heater should be placed at 12 (0.3 m) distance from the region to be tested to allow for uniform heating. Figure 3.2 Shop heater. 94

109 3.1.3 Description of the Bridges and Infrared Field Testing Results The four timber railroad bridges are located on the South Branch Valley Railroad (SBVR). The railroad is owned and operated by the West Virginia State DOT-State Railway Authority (SRA). The line provides freight and passenger service to the state s eastern panhandle. The names and characteristics of the timber railroad bridges that were tested are shown in Table 3.1. All four bridges are located near Moorefield, West Virginia (Figure 3.3). The piles and pile caps in these timber bridges were rehabilitated using Glass Fiber Reinforced Polymer (GFRP) composite wraps in previous years. Table 3.1 Timber bridge names and numbers. Bridge Name New Bridge # Old Bridge # Mile Post Length (feet) Height (feet) Fort Runs Dumpling Runs Lilly Pond Durgon Figure 3.3 Location of Moorefield, WV. 95

110 The infrared field testing was conducted on June 16, The test started at 10:00 A.M. and continued up to 4:00 P.M. During this period, infrared images of the piles and pile caps of the four bridges were acquired. The ambient temperature varied from 21 C (70 F) at 10:00 A.M. to 26 C (79 F) at 4:00 P.M. The maximum ambient temperature occurred between 2:00 P.M and 4:00 P.M., during which the temperature remained constant at 26 C (79 F). The heat source used during infrared testing of the pile and pile caps of this bridge was the 1500W quartz shop heater (Figure 3.2). When the quartz shop heater was used as the heat source, a distance of at least 12 (0.3 m) between the heater and the pile or pile cap was maintained in order to achieve uniform heating. Very close distance can lead to overheating of the region directly underneath the heating rods, which can introduce unwanted features in the infrared image. The heating duration was about 1 to 2 minutes so as to achieve temperatures of about 50 C. Then the heater was removed and the infrared image was acquired during the cooling of the wrapped member. The S60 infrared camera (Figure 3.1 (a)) and Windows XP laptop computer was used for acquiring infrared images. Fort Runs Bridge The picture of the Fort Runs Bridge (new #36.7, old #568) is shown in Figure 3.4. Two piles and one pile cap of this bridge were tested during this field trip. The bridge had a stream running under it but it was quite easy to take the readings with infrared camera, as the pile and pile caps are located at the edge of the bridge and there was no need to enter into the flowing water. Figure 3.5 shows a basic field testing set up. Due to the small space under the bridge, the infrared camera could not be mounted on a tripod and was held by hand. For this bridge two piles and one pile cap were tested. 96

111 Figure 3.4 Photograph of Fort Runs bridge (new #36.7, old #568). Figure 3.5 Infrared setup in the field. 97

112 Figure 3.6(a) shows photograph of the first pile which is wrapped with GFRP and Figure 3.6(b) shows its infrared image. Before acquiring the infrared image, the pile was heated with shop heater. After heating the pile for about one minute, the heater was removed and the infrared image was acquired. (a) (b) Figure 3.6 (a) Photograph of the first pile with a metal bolt, and (b) Infrared image showing the metal bolts as cold spot. The infrared image in Figure 3.6(b) shows some areas with higher temperature i.e., brighter spots, towards the top and bottom of the image. These hot spots indicate existence of air gaps or voids between the GFRP wrap and the underlying timber pile (known as a debonds ). The temperature of the hot spot is 40.4 C whereas the temperature of the adjacent defect-free area is about 30 C. The pile was also tap tested with a wooden stick and it also revealed hollow air gaps at the same locations where brighter spots were observed. A black spot at the center of the infrared thermal image (Figure 3.6(b)) indicates a much cooler region compared to the surrounding area. This is because of a metal bolt which can be clearly seen in the photograph in Figure 3.6(a). Since metals are good conductors of heat compared to timber, the heat incident on the metal is carried to its full length very quickly with no temperature build up on the surface. Thus, metals show up as cold spots in the infrared image compared to surrounding areas. Figure 3.7(a) shows the digital photograph of the second pile and Figure 3.7(b) shows the corresponding infrared image. Since the area underneath the bridge was very muddy, the 98

113 area surrounding the pile had to be cleaned. The surface temperature at the bottom of the pile was very cool at nearly 6 C (this area is not shown in the infrared image). The pile was heated for about a minute and the infrared image in Figure 3.7(b) was acquired. The temperature of the brighter region towards the bottom of the infrared image (Figure 3.7(b)) is 31 C whereas the temperature of the central and top part is around 29 C. The brighter region (hot spot) in the infrared image indicates presence of debonds, which is clearly evident as the bulging of the wrap at the bottom area of the photograph shown in Figure 3.7(a). This area also shows a crack in the GFRP wrap. (a) Figure 3.7 (a) Photograph of the second pile and (b) infrared image. (b) Figure 3.8(a) shows the photograph of the pile cap and Figure 3.8(b) shows the infrared image. The temperature of the hot spots (debond locations) was around 46 C whereas the temperature of the adjacent area was about 34 C. This pile cap was also tested by tapping and the tap testing also corroborated the results from the Infrared Thermography testing. Figure 3.8(b) also indicates a cold spot (~24 C) at the center, which corresponds to the metallic nail see in Figure 3.8(a). (a) Figure 3.8 (a) Photograph of the pile cap and (b) infrared image. (b) 99

114 Dumpling Runs Bridge The photograph of the Dumpling Runs Bridge (new #37.4, old #570) is shown in Figure 3.9. Five piles and one pile cap of this bridge were tested during this field trip. Figure 3.9 Photograph of Dumpling Runs Bridge (new #37.4, old #570). Figure 3.10(a) shows the photograph of the first pile that is wrapped with GFRP and Figure 3.10(b) shows the infrared image of the same pile. The infrared image shows a uniform temperature distribution (~24.4 C to 25.2 C) over the pile s surface, which indicates that the wrap is in good condition with no debonds. Figure 3.11(a) shows photograph of the second pile that is wrapped in GFRP and Figure 3.11(b) shows the infrared image of the same pile. The infrared image in Figure 3.11(b) shows some areas with higher temperature (i.e., brighter spots or hot spots). These hot spots indicate existence of air gaps or voids between the GFRP wrap and the underlying timber pile. The temperature of the hot spots was about 38 C whereas the temperature of the adjacent defect-free area was about 27 C. The pile was also tap tested, which revealed hollow air gaps at the same locations where the hot spots were observed. 100

115 (a) Figure 3.10 (a) Photograph of the first pile and (b) infrared image. (b) (a) Figure 3.11 (a) Photograph of the second pile and (b) infrared image. (b) Figure 3.12(a) shows a photograph of the third pile that was wrapped in GFRP and Figure 3.12(b) shows the corresponding infrared image. The infrared image in Figure 3.12(b) shows some areas with higher temperature (i.e., brighter spots) located primarily in the top half of the image. These hot spots indicate existence of air gaps or voids between the GFRP wrap and the underlying timber pile. The temperature of the hot spots was 32.5 C whereas the temperature of the adjacent defect-free area was about 28.5 C. 101

116 (a) Figure 3.12 (a) Photograph of the third pile and (b) infrared image (b) Figure 3.13(a) shows a photograph of the fourth pile that was wrapped in GFRP and Figure 3.13(b) shows the infrared image of the same pile. The infrared image in Figure 3.13(b) shows debonded areas with higher temperature (hot spots) towards the top of the image. The temperature of the hot spots was about 39.5 C whereas the temperature of the adjacent defect-free area was about 33.5 C. (a) Figure 3.13 (a) Photograph of the fourth pile and (b) infrared image. (b) Figure 3.14(a) shows photograph of the fifth pile that was wrapped in GFRP and Figure 3.14(b) shows the infrared image of the same pile. This infrared image shows some debonds (brighter region) towards the top of the infrared image (Figure 3.14(b)). The surface temperature over the debonded region was 37.5 C whereas the temperature of the bottom part of the image (defect-free areas) was around 33.2 C. 102

117 (a) Figure 3.14 (a) Photograph of the fifth pile and (b) infrared image. (b) It is important to note that the GFRP wrapped piles were in excellent condition during summer 2003 testing (Vasudevan 2004) as shown in the photograph and infrared image in Figure The prominent bright spot in the infrared image in Figure 3.15(b) is the tape over the strain gauge that can be seen clearly in Figure 3.15(a). (a) (b) Figure 3.15 (a) Photograph of the second pile and (b) infrared image. Two heating sources were used during infrared field testing of the GFRP wrapped timber components. One was the 1500W quartz shop heater (Figure 3.2) and the other was solar radiation. Solar radiation was used as a heat source when for such members; infrared images were acquired from a distance while the members were under solar radiation. Figure 3.16(a) and Figure 3.17(a) shows the photograph of the pile cap and Figure 3.16(b) and Figure 3.17(b) show the infrared image. These two sets represent two different regions of the same pile cap. In this case, the infrared images were acquired using solar radiation as the heat source, since the pile cap could not be accessed because of the stream of 103

118 water flowing underneath the bridge which made it very difficult to hold the heater at the desired location. Figure 3.16(b) shows a distinct cold spot (~26 C), which corresponds to the metallic bolt that can be clearly seen in Figure 3.16(a). In Figure 3.17(b), the temperature of the hot spots (debond locations) was around 31 C whereas the temperature of the adjacent area was about 29 C. This pile cap was also tested by tapping with a wooden stick and the tap testing corroborated the results from the Infrared Thermography testing. (a) Figure 3.16 (a) Photograph of the pile cap and (b) infrared image. (b) Figure 3.17 (a) Photograph of the pile cap and (b) infrared image. Lilly Pond Bridge The photograph of the Lilly Pond Bridge (new #42.4, old #574) is shown in Figure Four piles were tested during this field trip. Since water had flooded underneath the bridge (Figure 3.18, Figure 3.19(a), Figure 3.20(a), and Figure 3.21(a)), it was difficult to heat the wooden piles and pile caps. However, attempts were made to hold the heater from a distance and heat the test region in order to acquire the infrared images. 104

119 Figure 3.18 Photograph of Lilly Pond bridge (new #42.4, old #574). Figure 3.19(a) shows a photograph of the first pile and Figure 3.19(b) shows the corresponding infrared image. The infrared image in Figure 3.19(b) shows some areas with higher temperature i.e., brighter spots, towards the top and central region of the image. The temperature of the hot spots was 30.5 C whereas the temperature of the adjacent defect-free area was about 28.4 C. (a) Figure 3.19 (a) Photograph of the first pile and (b) infrared image. (b) 105

120 Figure 3.20(a) shows the photograph of the first, second and third piles together, and Figure 3.20(b) shows the corresponding infrared image. A close up photograph of the second and third piles is shown in Figure 3.21(a) and the corresponding infrared image is shown in Figure 3.21(b). Since the underneath of the bridge was very muddy, it was possible to heat only the third pile with the heater. This pile shows a uniform bright spot with temperature of 32 C whereas the second pile which was difficult to reach using the heater showed a temperature of 26.5 C. (a) Figure 3.20 (a) Photograph of the first, second and third piles and (b) infrared image. (b) (a) Figure 3.21 (a) Photograph of the second and third piles and (b) infrared image. (b) Figure 3.22(a) shows a photograph of the fourth pile and Figure 3.22(b) shows the infrared images of the same pile. The temperature of the brighter region in the infrared image of the fourth pile (Figure 3.22(b)) was 25.9 C whereas temperature of the bottom area that is covered by water was 23.5 C. 106

121 (a) Figure 3.22 (a) Photograph of the fourth pile and (b) infrared image. (b) Durgon Bridge The photographs of the Durgon Bridge (new #46.2, old #583) are shown in Error! Reference source not found.(a) and (b). Two piles were tested during this field trip. (a) (b) Figure 3.23 Photograph of Durgon Bridge (new #46.2, old #583). Error! Reference source not found., Figure 3.25 and Figure 3.26 show the piles and corresponding infrared images in the Durgon Bridge. Error! Reference source not found.(a) shows photograph of the first pile and Error! Reference source not found.(b) shows the infrared images of the same pile. The temperature of the brighter region in the infrared image of the first pile (Error! Reference source not found.(b)) was 19.5 C whereas temperature of an adjacent area was 17.9 C. The non-uniform temperature distribution in the infrared image in Error! Reference source not found.(b) is indicative of 107

122 the wrinkles formed in the GFRP wrap of this pile, which is clearly evident in Error! Reference source not found.(a). (a) Figure 3.24 (a) Photograph of the first pile and (b) infrared image. (b) Figure 3.25(a) shows photograph of the second pile and Figure 3.25(b) shows the corresponding infrared image. The temperature of the brighter region in the infrared image of this pile (Figure 3.25(b)) was 29.5 C whereas temperature of an adjacent area was 28.4 C. (a) (b) Figure 3.25 (a) Photograph of the second pile and (b) infrared image. Figure 3.26(a) and Figure 3.26 (b) show the photograph and the infrared image of the first and second pile combined. The combined infrared image is not as informative as the separate infrared images of the two piles presented in earlier figures. 108

123 (a) (b) Figure 3.26 (a) Photograph of the first and second pile and (b) infrared image Conclusions The field testing results showed that infrared thermography was successful in detecting the debonds between GFRP wrap and underlying timber piles and pile caps. It is obvious that infrared thermography can be a valuable tool for quality control and for subsequent inservice monitoring of the condition of the wrapped members. If debonding between the FRP fabric and the underlying timber member occurs during service, it can be detected using infrared thermography testing, and techniques such as injection of resin can be used to fill these subsurface voids to ensure continued structural integrity of the wrapped piles. This study has shown that some of the FRP wrapped piles showed increase in debonds and in some cases the wraps deteriorated severely at the bottom of the pile due to adverse impact by the river and flowing debris. While the severe deterioration was visible to the naked eye, the lower level of degradation in the form of subsurface debond formation could only be detected using infrared thermography. 3.2 Laboratory and Field Testing of FRP Composite Railroad Ties Introduction This section presents the results of laboratory and field testing conducted on GFRP composite railroad ties using infrared thermography. Old timber railroad ties were taken out of service and used as the core material with GFRP composite surrounding the ties in order to 109

124 make new "molded" composite railroad ties as a part of another project involving recycling/rehabilitation of old railroad ties. These molded composite ties were tested using infrared thermography just after completion of the manufacturing process in June 2009 in the Structures Laboratory at West Virginia University, Morgantown, WV. The infrared thermography tests were conducted to detect any debonds present at the interface between the GFRP composite wrap and the underlying timber component as a part of the quality control during the manufacturing process. Subsequently, the composite ties were installed in a railroad track in South Branch Valley Railroad (SBVR), Moorefield, West Virginia (Figure 3.3). These composite ties were then tested using infrared thermography on September 4, 2009, which was after about two months of service. The laboratory and field infrared test results and conclusions are presented in the following sections Laboratory Testing, Analysis and Results Several composite wooden ties were cast by another group of students for field installation. These ties were tested in this research study using infrared thermography just after casting with GFRP composite around them. The size of the cast composite ties was 102 x 8.75 x 7.5 (2591 mm x 222 mm x 191 mm) as shown in Figure While this figure also shows the wooden core, in reality the two ends were also capped with GFRP composite. The actual composite ties are shown in Figure Figure 3.27 Schematic of the composite tie. 110

125 Figure 3.28 Wooden ties after encasing with GFRP composite. The thermal images were recorded using both ThermaCAM S60 and InfraCAM SD infrared cameras (shown in Figure 3.1(a) and 3.1(b)), which are capable of detecting the infrared energy in the spectral range of 7.5 to 13 microns. Tests Using ThermaCAM S60 Camera The first set of laboratory tests were conducted using the high end ThermaCAM S60 camera. The surface of the composite ties were heated to a temperature of 160 o F (71 o C) using a long heater which covered the entire top surface of the composite tie. Then the heat source was removed and the acquisition of infrared images commenced. Figure 3.29(a) shows the photograph of the first composite tie and Figure 3.29(b) shows the corresponding infrared image. The infrared image in Figure 3.29(b) shows an area with higher temperature, i.e., bright spot at a temperature of 68 C. This bright temperature is most likely due to non-uniform heating by the long heater. Additional infrared images were acquired as the composite tie was cooling. Figure 3.30(a) shows photograph of the same composite tie and Figure 3.30(b) shows the infrared image after some cooling. This infrared image shows the temperature of the bright spot as 56 C. Subsequent test results using a shop heater that is presented later will show that these bright spots are not indicative of any subsurface defects but rather a result of non-uniform heating by the long heater used for acquiring the particular images in Figure 3.29(a) and Figure 3.29(b). 111

126 (a) Figure 3.29 (a) Photograph of the first tie and (b) infrared image. (b) (a) Figure 3.30 (a) Photograph of the composite tie and (b) infrared image. (b) 112

127 Tests Using InfraCAM SD Camera For these tests, the surface of the tie was heated with the shop heater (Figure 3.2) for about a minute and the heater was removed before acquiring the infrared images. Figure 3.31(a) shows the photograph of the first composite tie and Figure 3.31(b) shows the corresponding infrared image. The infrared image in Figure 3.31(b) shows some areas with higher temperature, i.e., bright spot at the center of the image. The temperature of this bright spot is 41.3 C whereas the temperature of the adjacent defect-free areas typically ranges from 35.7 C to 38 C. The brighter region (hot spot) in the infrared image indicates presence of debonds (small air gaps between the composite material and the underlying wooden core). (a) Figure 3.31 (a) Photograph of the first tie and (b) infrared image. (b) Figure 3.32(a) shows photograph of the second tie that was molded with GFRP composite and Figure 3.32(b) shows the corresponding infrared image. The infrared image in Figure 3.32(b) shows some areas with higher temperature i.e., brighter spots, at the center of the image. These hot spots indicate existence of air gaps or voids between the GFRP material and the underlying wooden core. The temperature of the hot spots was around 37 C whereas the temperature of the adjacent defect-free area was around 35 C. The difference between the hot spots and the adjacent defect-free area is small (only 2 C) which indicates that there are only small voids and most of the tie is uniform, which shows that the molding process has worked very well. 113

128 (a) Figure 3.32 (a) Photograph of the second tie and (b) infrared image. (b) Field Testing of FRP Composite Railroad Ties Seven molded GFRP composite railroad ties were placed along the South Branch Valley Railroad (SBVR) in Moorefield, West Virginia. The SBVR line is owned and operated by the West Virginia State DOT-State Railway Authority (SRA) and is used to provide freight and passenger service to the state s eastern panhandle. The seven composite ties were placed at three locations. Each location had two or three ties, and the three locations were within 2000ft. The field infrared testing was conducted on September 4, The tests were conducted under solar heating between the hours of 11:00 A.M and 1:00 P.M., which is the ideal time for acquiring infrared images. The ambient air temperature varied from 20 C (68 F) at 11:00 A.M to 27 C (81 F) at 1:00 P.M. The maximum ambient air temperature occurred between 2:00P.M and 4:00P.M., during which the temperature remained constant at 26 C (79 F). The seven molded composite ties were placed in the SBVR line according to the following order: two at the southern end (Figure 3.33(a)), three in between (Figure 3.33(b)) and the remaining two at the northern end (Figure 3.33(c)). 114

129 (a) (b) (c) Figure 3.33 Location of the seven molded composite ties, (a) two at the southern end, (b) three in between, and (c) two at the northern end. Figure 3.34(a) shows the photograph of the outer part (the portion outside the rail track) of the first molded composite tie and Figure 3.34(b) shows the corresponding infrared image. The infrared image in Figure 3.34(b) shows areas in the central region with higher temperature, i.e., brighter spots, with 36 C to 38 C temperature. In contrast, the temperature at the top part of the infrared image was around 29 C. The brighter region (hot spots) in the central part of the infrared image indicates presence of some debonds (small air gaps) and cracking/deterioration in this portion of the tie, which is also evident from the photograph in Figure 3.34(a). (a) Figure 3.34 (a) Photograph of the first tie and (b) infrared image. (b) 115

130 Figure 3.35(a) shows photograph of the outer part of the second molded composite tie and Figure 3.35(b) shows the infrared image of the corresponding tie. The infrared image in Figure 3.35(b) shows one area with higher temperature, i.e., brighter spot, with a temperature of 56 C. The rest of the infrared image shows a uniform temperature, especially at the center. (a) Figure 3.35 (a) Photograph of the second tie and (b) infrared image. (b) (a) Figure 3.36 (a) Photograph of the third tie and (b) infrared image. (b) 116

131 Figure 3.36(a) shows the digital photograph of the outer part of the third tie and Figure 3.36(b) shows the corresponding infrared image. The infrared image shows a more or less uniform temperature ranging between 40 C and 42 C. The tie was cracked at the end which is clearly evident in the digital photograph as well as the infrared image. Figure 3.37(a) shows photograph of the fourth tie and Figure 3.37(b) shows the infrared image of the same tie. The infrared image showed temperatures varying between 62 C and 70 C. This tie was under the shadow cast by a tree, which explains the variation in the surface temperature. (a) Figure 3.37 (a) Photograph of the fourth tie and (b) infrared image. (b) Figure 3.38(a) shows a photograph of the outer part of the fifth tie and Figure 3.38(b) shows the infrared image of the same tie. The infrared image showed temperatures varying between 51 C and 63 C. This tie was also under the shadow cast by a tree, which explains the variation in the surface temperature. 117

132 (a) Figure 3.38 (a) Photograph of the fifth tie and (b) infrared image. (b) Figure 3.39(a) shows photograph of the outer part of the sixth tie and Figure 3.39(b) shows the infrared image of the same tie. The infrared image in Figure 3.39(b) shows some bright spots, i.e., areas with higher temperature (59 C). These hot spots indicate existence of air gaps or voids between the FRP mold and the underlying wooden railroad tie. (a) (b) Figure 3.39 (a) Photograph of the sixth tie and (b) infrared image. 118

133 Figure 3.40(a) shows a photograph of the seventh tie and Figure 3.40(b) shows the corresponding infrared image. The infrared image shows a fairly uniform temperature except for a brighter region (debonded area) towards the bottom of the infrared image with a temperature of 57 C. (a) Figure 3.40 (a) Photograph of the seventh tie and (b) infrared image. (b) Figure 3.41 shows the photographs of two wooden ties that were previously wrapped with GFRP fabric in the portions under the rails and then reinstalled in the field. The wrapped portions of these ties were tested using infrared thermography. Figure 3.42(a) shows the photograph of the first GFRP wrapped wooden tie and Figure 3.42(b) shows the corresponding infrared image. The temperature of the brighter region in the infrared image was 54 C whereas temperature of an adjacent area was 51 C. This indicates that there are debonds under the high temperature regions. The FRP wrapped wooden tie was also tap tested with a wooden stick and it also revealed hollow air gap at the same location where the brighter spot was observed. The black spot at the top of the infrared image (Figure 3.42(b)) indicates a much cooler region compared to the surrounding area. This is because of a metal plate which can be clearly seen in the photograph in Figure 3.42(a). Since metal plate is a good conductors of heat compared to wood, the heat incident on the metal plate is carried to its full length very quickly with no temperature build up on the surface. Thus, metal plates show up as cold spots in the infrared image compared to surrounding areas. 119

134 (a) (b) Figure 3.41 (a) Photograph of the first and second FRP wrapped wooden ties, (b) Close up photograph. (a) Figure 3.42 (a) Photograph of the first FRP wrapped wooden tie and (b) infrared image. (b) Figure 3.43(a) shows a photograph of the second wooden tie wrapped with GFRP composite and Figure 3.43(b) shows the infrared image of the corresponding tie. The temperature of the brighter region in the infrared image of this tie was 59 C whereas temperature of the adjacent area was 56 C. This GFRP wrapped wooden tie was also tested by tapping with a wooden stick and the tap testing also corroborated with the results from the Infrared Thermography testing. Figure 3.43(b) also indicates a cold spot (~38 C) at the top, which corresponds to the metallic plate and bolts see in Figure 3.43(a). The infrared image in Figure 3.43(b) also shows the temperature of the wooden area just outside the wrapped area 120

135 as being lower (51 o C), indicating that the temperature of the wood surface is lower than that of the GFRP wrapped surface. (a) (b) Figure 3.43 (a) Photograph of the second FRP wrapped wooden tie and (b) infrared image Conclusions During the laboratory experiments the GFRP molded composite ties were tested by heating them with a long heater which could heat the entire top surface simultaneously. There were surface temperature differences in the images that were a result of non-uniform heating by this long heater. Subsequent testing using a shop heater showed that there were only a few small size debonds scattered in the tie. The GFRP molded composite ties were placed in service in the field for about two months and then tested once again using infrared thermography. The field testing was conducted under solar heating, and the results showed that infrared thermography was successful in detecting some debonds in the GFRP molded ties. Additional infrared tests conducted on two GFRP "wrapped" wooden ties did not reveal any subsurface debonds, thus indicating that the wrapping process was conducted properly. It is obvious that infrared thermography can be a valuable tool for quality control and for subsequent in-service monitoring for both molded and wrapped composite ties. 121