COMPARATIVE ANALYSIS OF DRILLS FOR COMPOSITE LAMINATES

COMPARATIVE ANALYSIS OF DRILLS FOR COMPOSITE LAMINATES Luís Miguel P. Durão 1*, Daniel J. S. Gonçalves 2, João Manuel R. S. Tavares 2, Victor Hugo C. de Albuquerque 2, 3, A. Torres Marques 2 1 Instituto Superior de Engenharia do Porto (ISEP) / Centro de Investigação e Desenvolvimento em Engenharia Mecânica (CIDEM) - R. Dr. António Bernardino de Almeida, 431 4200-072 PORTO, PORTUGAL 2 Faculdade de Engenharia da Universidade do Porto (FEUP) / Instituto de Engenharia Mecânica e Gestão Industrial (INEGI) - R. Dr. Roberto Frias, 400 4200-465 PORTO, PORTUGAL 3 Universidade de Fortaleza (UNIFOR), Centro de Ciências Tecnológicas (CCT), Núcleo de Pesquisas Tecnológicas (NPT), BRASIL *lmd@eu.ipp.pt Abstract The characteristics of carbon fibre reinforced plastics allow a very broad range of uses. Drilling is often necessary to assemble different components, but this can lead to various forms of damage, such as delamination which is the most severe. However, a reduced thrust force can decrease the risk of delamination. In this work, two variables of the drilling process were compared: tool material and geometry, as well as the effect of feed rate and cutting speed. The parameters that were analysed include: thrust force, delamination extension and mechanical strength through open-hole tensile test, bearing test and flexural test on drilled plates. The present work shows that a proper combination of all the factors involved in drilling operations, like tool material, tool geometry and cutting parameters, such as feed rate or cutting speed, can lead to the reduction of delamination damage and, consequently, to the enhancement of the mechanical properties of laminated parts in complex structures, evaluated by open-hole, bearing or flexural tests. 1

1 Introduction 1.1 Drilling of Composite Materials Nowadays, composites are one of the most promising groups of materials for various industries. The main characteristics of composites, such as low density and high strength to weight ratio, make them ideal where high stiffness, high strength and low weight are required. Consequently, the usage of composite materials has grown in recent years, as can be seen in their intensive use in the new Airbus A380 or Boeing 787 airplanes, where 50% of the weight of its primary aircraft structure will be made of composite [1]. Composite materials are not only found in the aeronautical field, but also in other industries, such as the automotive, railway, energy, sports equipment and domestic appliances industries as well as in other applications where low weight and good mechanical characteristics are required. Although composite parts are produced to near-net shape, finishing operations like drilling, for assembly, are usually required. These operations can be carried out with conventional tools and machinery with the proper adjustments. However, due to the lack of homogeneity of the composites, this operation can lead to several damages. The most frequent and noticeable damages are around the edges of the machined hole, usually at the exit side of the drill, due to the drilling process. The most common damages are delamination, fibre-pull-out and thermal damages [2]. Of these, delamination is considered the most serious, as it can lead to a reduction of the mechanical properties of the laminate. Persson et al. [3, 4] demonstrated that a reduction of around 10% in static testing of pin load specimens and fatigue testing can occur in comparison to parts drilled using dedicated tools. In another work, Durão et al. [5] verified an increase of 4% on bearing test results, just by selecting the appropriate tool geometry. The reduction or even the elimination of such damage is of vital importance to the industries that manufacture parts in composite materials. However, increased knowledge of the damage 2

mechanisms and control of the conditions that lead to its onset and propagation are required to attain such results. Focusing on delamination, two different mechanisms are normally referred to: peel-up and push-down, Figures 1a and 1b, respectively. The former is a consequence of the drill entering the upper plies of the plate and the cutting edge of the drill abrades the laminate. The material spirals up along the flute before being effectively cut. This action can be avoided with the use of low feeds [6]. On the other hand, the latter mechanism is a result of the indentation effect caused by the quasi-stationary drill chisel edge, acting on the uncut plies of the laminate. The cause of this damage mechanism is the compressive thrust force, measurable during the drilling process, exerted by the drill on the uncut plies of the plate. As the uncut thickness approaches zero the mechanical resistance of the plate decreases. At some point, this load exceeds the interlaminar fracture toughness of the laminate and the plies tend to be pushed away from the plate, causing the separation of two adjacent plies of the laminate and, consequently, delamination takes place [6, 7]. In general, delamination can be reduced by minimizing the thrust forces exerted by the drill chisel edge. The push-down delamination is more difficult to avoid as, at some point during the drilling process, a situation of one uncut ply will necessarily occur. Numerous works had been published on the subject of delamination reduction. For example, Person et al. [3, 4] developed a patented method, known as NOVATOR, an orbital drilling method, in which the hole is machined both axially and radially by rotating the cutting tool around its own axis as well as eccentrically while feeding through the laminate. Based on experimental analysis, Piquet et al. [8] suggested that an optimal drill should have a greater number of cutting edges, from three to six, a point angle of 118º and a reduced chisel edge. Additionally, a variable feed rate could help to prevent delamination. Stone and Krishnamurthy [9] studied the implementation of a neural network thrust force controller that searches for the best feed rate 3

every three spindle revolutions in order to minimize delamination. Davim and Reis [10] presented an experimental work on the effect of tool geometry and drilling parameters on delamination. These authors found that the influence of feed rate and cutting speed was dependent on drill geometry. A study conducted by Shyha et al. [11], demonstrated a greater importance of feed rate when compared to cutting speed. Hocheng and Tsao [12] evaluated the performance of special drill bits and concluded that drill geometry could influence the thrust force and, consequently, the threshold of feed rate to avoid delamination. Thus, core drill, candle stick drill, saw drill and step drill can be operated at larger feed rates than twist drill, meaning shorter cycle times, without delamination damage. The influence of tool geometry on delamination was also evaluated in [13, 14, 15], showing the benefit of using a step drill. The advantage of pilot hole drilling on delamination reduction by cancelling the chisel edge effect during final diameter machining has been studied by different authors [16-18]. In [19] the authors studied the effect of the cutting variables on thrust force, torque, hole quality and chip. Drill diameter variation was also considered in this study. In the end it was noted that circularity issues are more pronounced when feed rate increases. A comprehensive review of the steps towards delamination free drilling of composite materials can be found in [20]. In addition, a literature survey on the machinability of composite laminates, in terms of tool material and geometry or cutting parameters, as well as hole quality issues, was compiled by Abrão et al. [21]. A recent advance was given by Schulze et al [22] minimizing damage by directing the cutting forces towards the centre of the workpiece. In this work, the effect of two variables on the drilling process of composite materials was studied: the tool material and the tool geometry. Two tool materials tungsten carbide (WC) and polycrystalline diamond (PCD) with the same twist drill geometry were evaluated; and three different tool geometries of the WC drills twist, Brad and step, Figure 2, were assessed. 4

The parameters considered in this study included thrust force monitoring during drilling, the assessment of the delamination extension from enhanced digital radiographies that are segmented by an artificial neuronal network and, finally, mechanical testing of drilled coupons to evaluate any reduction of the mechanical properties. The mechanical tests carried out were the Standard Test Method for Open-Hole Tensile Strength of Polymer Matrix Composite Laminates ASTM D5766M-07 [23], the Standard Test Method for Bearing Response of Polymer Matrix Composite Laminates ASTM D5961M-08 [24] and the Fibre-reinforced plastic composites -- Determination of flexural properties ISO 14125:1998, Method A (three point) [25]. The latter test is not commonly referred to in published literature for evaluating hole machining quality. The usefulness of this test in terms of drilling damage still needs further analysis. This test is frequently used to evaluate composite materials in terms of tensile, compressive and shear stresses acting simultaneously. To evaluate the relative importance of each experimental factor, statistical techniques, in particular the analysis of variance (ANOVA), were applied. This work shows that a proper combination of all the factors involved in drilling operations, like tool material, tool geometry and cutting parameters, such as feed rate or cutting speed, can lead to the reduction of delamination damage and, consequently, to the enhancement of the mechanical properties of laminated parts in complex structures. In the next section, the materials and methods adopted are introduced; then, the experimental results are presented and discussed; finally, in the fourth and last section of the paper, the conclusions of the work done are pointed out. 5

2 Materials and Methods 2.1 Composite Plates Production and Drilling For the experimental work, a batch of carbon/epoxy plates using a commercially available prepreg with a cross-ply stacking sequence of [(0/90) 6 ] s were produced. The plates were then cured under 300 kpa pressure and 130 ºC for one hour, followed by cooling. Final plate thickness was 4 mm. Similar plates with 2 mm thickness were produced for flexural testing. Then, the plates were cut into test coupons of 165x96 mm 2 for the drilling experiments and in coupons of other dimensions depending on the mechanical test specifications open-hole, bearing and flexural. The drilling operation was carried out in a 3.7 kw CNC machine. Two setups were used: In the first, the twist drill geometry remained unchanged and tool material was changed from WC to PCD. In the second, the tool material WC was kept constant while tool geometry was changed: initially, a twist drill with a point angle of 120º (Fig. 2a), then a Brad drill (Fig. 2b) and lastly a step drill (Fig. 2c) were used. Drill bit details, specifically on the step drill can be found in [26]. All the drills had a 6 mm diameter and the cutting parameters for all the tryouts were: three cutting speeds: 2800, 4200 and 6000 rpm, corresponding to a cutting speed of 53, 79 and 113 m/min; three feeds rates: 0.02, 0.06 and 0.12 mm/rev. An increase in feed is known to cause an increase in both thrust force and delamination [9-11, 20, 27]. As mentioned earlier, the influence of the cutting speed on thrust force and delamination is inferior to the influence of the feed rate. However, the influence of feed rate on the mechanical strength by destructive testing has not yet been confirmed. During drilling, the axial thrust forces were monitored with a dynamometer associated to an amplifier and a computer for data acquisition and processing. No back plates were used. The experimental setup is presented in Figure 3. 6

2.2 Delamination Evaluation After drilling, the delaminated region around the drilled hole was evaluated using enhanced digital radiography. In order to produce a contrast the plates were first immersed in diiodomethane for thirty minutes. Then the radiography images were acquired using a 60 kv, 300 khz 2100 X-Ray system associated with a digital acquisition system. The radiography images were processed and analysed to obtain the segmentation and posterior characterization of the regions of interest [28], Figure 4. The round white region at the centre, Figure 4a, b), is the hole and the darker border around the hole is the damaged area, Figure 4a, b). The segmentation of the input images, that is to say, the identification of hole, delaminated and non delaminated regions, was achieved by using an artificial neuronal network with the following topology: an input layer composed of three neurons, which corresponds to the pixel values of each image to be evaluated in RGB (red, green and blue) colour spaces, a hidden layer consisting of 7 neurons, which was defined by the heuristic rule proposed in [29], and finally an output layer with three neurons. The logistic function [30] was considered for neurons activation function. It presents three output values: -1 (one), 0 (zero) and 1 (one). Although the network was used in this work to segment gray level images (0 to 255 levels) and to classify three regions of interest, it can also segment colour images into a maximum of 27 classes. The backpropagation learning algorithm [31] was used to train the multilayer perceptron neuronal network, as it is one of the most used algorithms for this type of network training. It should be pointed out here that to segment a set of similar images, the training of the neuronal network only needs to be performed once. For the experimental work the neuronal network was trained using an average of 10 pixels manually selected from 6 training images for each of the 3 segmentation classes. Also, the training phase was stopped once an absolute error not higher than 0.01 or a number of 7

iterations equal to 2,500 epochs was reached. Further details of the neural network used here can be found in [32]. After the segmentation of each input image, Figure 4c), measurements of the hole and delaminated regions, such as diameters and areas, can be computed: the diameters by searching for the longest diagonal within the segmented regions, Figure 4d), and the areas by summing up the pixels within the segmented regions. The values obtained from the radiography images can be used to determine the delamination factor proposed by Chen [33] that is defined as the ratio of the maximum delaminated diameter (D max ) to the nominal hole diameter (D): F d Dmax. (1) D One limitation of Chen s criterion is related to situations when the delamination involved isn t round, but presents breaks and cracks. In such cases, the values of the delaminated area are more appropriated for the damage quantification. Based on this, Davim et al. [34] suggested a novel approach known as the adjusted delamination factor F da, F da Dmax Amax, (2) D A where A max is the area corresponding to the maximum delaminated diameter and A o the nominal hole area. In this new criterion, the first term is the conventional delamination factor and a second term is added, taking into account the contribution of the damaged area and the parameters α and β 8

are used as weights, their sum always being equal to 1 (one). This criterion was adopted here with the intention of making comparisons. 2.3 Mechanical Testing The mechanical tests carried out were the Open-hole tensile strength test, according to ASTM D5766M-07 [23], the Bearing test, according to ASTM D5961M-08 [24] and the Flexural test, according to ISO 14125-1998 [25]. These tests were used to assess the effect of delamination on the mechanical properties of the drilled plates. Test coupons of 250x36 mm 2, 135x36 mm 2 and 100x15 mm 2, respectively for the above mentioned tests, were cut and drilled under the same experimental conditions, as described in section 2.1. The tests were performed in a 100 kn Universal Testing machine equipped with the necessary accessories to run the different tests and connected to a computer for machine control and data acquisition. The test speeds were 5 mm/min for the Open-hole test, 2 mm/min for the Bearing test and 4 mm/min for the Flexural test. 3 Results and Discussion 3.1 Comparison of Tool Material The results presented for the thrust force are the maximum values acquired during drilling as, according to published analytical models [6], superior thrust forces correspond to a higher delamination probability. Due to the signal variations during drill rotation, the thrust force values were averaged over one spindle revolution. Also, to reduce the influence of outlier values, the final results used were the average of six experiments run under identical conditions. Independently of the tool geometry or material used, the thrust force was always superior with increased feed rates, Figure 5. This was an expected outcome, and it is in line with the 9

findings from previous works [9-20, 26]. Consequently, the differences observed on thrust force results are due to the tool variations, either in terms of the geometry or in terms of the material. From figure 5, it turns out that tungsten carbide tools have always smaller thrust forces than polycrystalline diamond tools for all feed rates used in the experimental work. This result was not expected as it is known that PCD drills, due to lower friction coefficient, should present lower thrust forces. Three reasons can be given to this result. First, PCD tool is a tipped tool, meaning that although the main angles are identical for both tools, rake angle will differ from one material to the other. As commercial tools were used, this situation was not possible to change. In spite of this fact, it represents the actual condition of available tooling. The second reason is related with the fact that only new tools were used for each run of six holes. So, this situation reduces the advantage of using PCD drills as new WC tools, with sharp cutting edge, are known to enable a clean cut. Greater series will change this result. Finally, it should be noted that the use of PCD drills allows for higher cutting speeds when compared to WC drills and in this work this acquainted advantage was not explored. The results comparing the maximum thrust force, the adjusted delamination factor, the open hole strength, the bearing strength and the flexural strength are presented in Figure 6 and in Table 1 for the twist drills made with the two tool materials under analysis. The maximum thrust force results are the average of the three feed rates considered in the experimental work. The values presented in Figure 6 and Table 1 show that for twist drill geometry, under the experimental conditions applied here, tungsten carbide appears to be the best material. Possible reasons for this are discussed above. One must remember that PCD tools are more wear resistant and in this work the number of holes drilled with each tool was small, only six holes, and consequently the long term effects were not taken into account. A second argument in favour of PCD tools is related to productivity. PCD drills allow higher cutting speeds and 10

feed rates, leading to greater output. However, this interesting parameter was not in the scope of this study. From the results presented in Figure 6 and Table 1 it is possible to affirm that WC tools present less thrust force during drilling. This result correlates with less damage extent, by the adjusted delamination factor and greater values of mechanical strength, as shown by the values of bearing and flexural strength. These results are coherent in the whole and so, it appears as, under appropriate conditions and for a reduced number of holes, WC tools are well suited. On the other hand, it could be the consequence of a combination of experimental parameters that are more adjusted for WC tool drilling than for PCD tool drilling. Past experience of the authors has been mainly focused on WC tools. It also is noteworthy to refer the excellent condition of the machining centre. After every tool assembly, eccentricity of the drill barrel was always checked. This good machine condition could have been an important contribution to good WC drilling results. 3.2 Comparison of Tool Geometry The results here presented were obtained following the same experimental sequence as the previous section. The intention was to evaluate the geometry of the tool, thus only one material, tungsten carbide (WC), was used with three different feed rates, 0.02, 0.06 and 0.12 mm/rev, as in the aforementioned section, and three cutting speeds, 53, 79 and 113 m/min or a spindle speed of 2800, 4200 and 6000 rpm. PCD tools do not have such a wide variety of geometries as WC tools do. So, and independently of the results from material evaluation, the comparison of drill geometries in tungsten carbide allows for more diversity. As mentioned before, cutting speed has less influence on the results and so the number of variable factors was kept low to avoid possible interacting effects that could mask the conclusions. 11

The results obtained for a spindle speed of 2800 rpm show clearly that the tool geometry has a large influence on all parameters used for comparison. As it can be observed in Figure 7 there are remarkable differences from one tool geometry to the other, independently of the outcome to be considered. The specific geometry of the step drill, dividing the drilling operation in two phases, provides a considerable reduction of the maximum thrust force. Although there was not a clear reduction on the delamination extension, plates drilled with step drill had returned the higher results in the mechanical tests. This is a promising result. The non-chisel edge Brad drill eliminates the extrusion effect exerted by the drill tip, promoting a diverse cutting mechanism with a large reduction in the thrust force. This drill tip tends to tension the fibres prior to cut, enabling a clean cut. From visual inspection, the holes drilled with this tool always seem smoother and the surface roughness of hole walls is lower [13]. Similar conclusions were taken from the remaining spindle speeds. From the results of this experimental stage it was possible to apply analysis of variance - ANOVA - on the maximum thrust force and on the delamination factor to estimate the relative influence of tool geometry, feed rate and spindle speed, see Tables 2 and 3, respectively. The feed rate reveals to be the most important factor acting on the maximum thrust force and on the delamination factor, with a percentage of contribution of 75% and 53%, respectively. The second most important factor is tool geometry with a percentage of contribution of 22% to the maximum thrust force and 35% to the delamination factor and finally the cutting speed contribution is always smaller, with a contribution of 2% to the maximum thrust force and 8% to the delamination factor. Experimental error was about 1% for the maximum thrust force and 4% to the delamination factor. Values from the F test on all the experimental factors show that feed rate and tool geometry are strong factors with physical and statistical significance on the maximum thrust force and on the delamination 12

factor. F-test values for cutting speed show that this factor has a moderate effect on both results evaluated. The correlation between the delamination factor or the adjusted delamination factor and the mechanical test results are represented in Figure 8. The correlation with Open-hole test results is represented in Figure 8a) and it shows that larger delamination causes a reduction in the mechanical strength of the plate around the connecting regions where screws, rivets or bolts are used for assembly purposes. Identical result is obtained with Bearing test results, see Figure 8b), leading to the same evidence. Moreover, data correlation for these data points was the highest of all. Finally, the correlation with flexural test results, represented in Figure 8c), does not seem to point out a conclusion so evident. Flexural stress values are greater as delamination is more extended as demonstrated by the linear correlation also represented in the same figure. However, data correlation was not so evident. 4 Conclusions Carbon fibre reinforced laminates were drilled in order to compare the performance of two different tool materials and three tool geometries on the maximum thrust force, delamination extension and mechanical strength. The experimental work involved three feed rates and three cutting speeds. It is possible to draw the following conclusions from the work presented: From the two tool materials compared, WC tools seems to be more favourable than PCD tools, as they are less expensive, cause less damage and turn out to be more economical for small series of holes. It is confirmed that low feed rates are appropriate for laminate drilling, as it reduces the axial thrust force and consequently, the delamination around the hole. Moreover, low feeds minimize the loss of mechanical strength of the drilled part. 13

ANOVA results show that feed rate is the most important factor for delamination reduction, followed by tool geometry. Both factors have physical and statistical significance as it is demonstrated by the results of the F-test. Therefore, an appropriate selection of feed rate and tool geometry is recommendable when drilling composite laminates. Bearing test, in spite of being more complex, should be used to characterize mechanical loss due to existing damage. Data correlation of test results with damage extension was the highest. Open-hole tests and flexural tests (3-point) could possibly represent an easier way to characterize mechanical loss due to delamination damage. Both test procedures are very simple to implement and perform. However further analyses with these tests, as well as 4- point flexural tests, are necessary to assess this possible relevance. Acknowledgement This work was partially granted by FCT - Fundação para a Ciência e a Tecnologia, Portugal, in the scope of the project Drilling of polymeric matrix composites structures with reference PTDC/EME-TME/66207/2006. The fourth author thanks National Council for Research and Development (CNPq) and Cearense Foundation for the Support of Scientific and Technological Development (FUNCAP) for providing financial support through a DCR grant (project number 35.0053/2011.1) to UNIFOR. References [1] Gilpin, A. Tools solutions for machining composites, Reinforced Plastics, August/September 2009, 30-34. 14

[2] Wern CW, Ramulu M, Schukla A. Investigation of stresses in the orthogonal cutting of fiber-reinforced plastics, Experimental Mechanics 1994: 33-41. [3] Persson E, Eriksson I, Zackrisson L. Effects of hole machining defects on strength and fatigue life of composite laminates, Composites A 1997; 28: 141-51. [4] Persson E, Eriksson I, Hammersberg P. Propagation of hole machining defects in pinloaded composite laminates, J. of Composite Materials 1997; 31: 383-408. [5] Durão L M P, Magalhães A G, Marques A T, Baptista A M, Figueiredo M. Drilling of fibre reinforced plastic laminates, Materials Science Forum 2008; 587-588: 706-10. [6] Hocheng H, Dharan CKH. Delamination during drilling in composite laminates, J. of Engineering for Industry 1990; 112: 236-39. [7] Stuart, ML. International Encyclopaedia of Composites 1991; 2: 297. [8] Piquet R, Ferret B, Lachaud F, Swider P. Experimental analysis of drilling damage in thin carbon/epoxy plate using special drills, Composites A 2000; 31: 1107-15. [9] Stone R, Krishnamurthy K. A neural network thrust force controller to minimize delamination during drilling of graphite-epoxy composites, Int. J. of Machine Tools & Manufacture 1996; 36: 985-1003. [10] Davim J P, Reis P. Drilling carbon fibre reinforced plastics manufactured by autoclave experimental and statistical study, Materials & Design 2003; 24: 315-24. [11] Shyha IS, Aspinwall DK, Soo SL, Bradley S. Drill geometry and operating effects when cutting small diameter holes in CFRP, Int J Machine Tools & Manuf 2009; 49, 1008-14. [12] Hocheng H, Tsao CC. Effects of special drill bits on drilling-induced delamination of composite materials, Int. J. of Machine Tools & Manufacture 2006; 46: 1403-16. [13] Gonçalves DJS, Durão LMP, Tavares JMRS, de Albuquerque VHC, Marques AT. Evaluation of tools and cutting conditions on carbon fibre reinforced laminates, Materials Science Forum 2010; 638-642: 944-49. 15

[14] Durão LMP, Gonçalves DJS, Tavares JMRS, de Albuquerque VHC, Vieira AA, Marques AT. Drilling tool geometry evaluation for reinforced composite laminates, Composite Structures 2010, 92, 1545-1550. [15] Durão LMP, Gonçalves DJS, Tavares JMRS, de Albuquerque VHC, Marques AT, Baptista AM. Drilling of carbon fibre reinforced laminates a comparative analysis of five different drills on thrust force, roughness and delamination, Materials Science Forum 2010, 636-637, 206-213. [16] Tsao C C. The effect of pilot hole on delamination when core drilling composite materials, Int. J. of Machine Tools & Manufacture 2006; 46: 1653-61. [17] Won M S, Dharan CHK. Chisel edge and pilot hole effects in drilling composite laminates, Trans. of ASME J. of Manuf. Science & Engineering 2002: 124: 242-47. [18] Tsao CC, Hocheng H. The effect of chisel length and associated pilot hole on delamination when drilling composite materials, Int. J. Machine Tools & Manuf 2003; 43: 1087-92. [19] Zitoune R, Krishnaraj V, Collombet F. Study of drilling of composite material and aluminium stack, Composite Structures 2010, 92, 1246-1255. [20] Hocheng H, Tsao CC. The path towards delamination-free drilling of composite laminates, J. Materials Processing Technology 2005; 167: 251-64. [21] Abrão AM, Faria PE, Campos Rubio JC, Reis P, Davim JP. Drilling of fiber reinforced plastics: A review, Journal of Materials Processing Technology 2007; 186 (1-3): 1-7. [22] Schulze V, Becke C, Weidenmann K, Dietrich S. Machining strategies for hole making in composites with minimal workpiece damage by directing the process forces inwards, Journal of Materials Processing Technology 2011, 211, 329-338. [23] ASTM D5766M-07: Standard test method for Open-Hole tensile strength of polymer matrix composite materials, ASTM International. 16

[24] ASTM D5961M-08, Standard test method for Bearing response of polymer matrix composite laminates, ASTM International. [25] ISO 14125:1998, Fibre-reinforced plastic composites - Determination of flexural properties, ISO. [26] Marques AT, Durão LMP, Magalhães AG, Silva JF, Tavares JMRS. Delamination analysis of carbon fibre reinforced laminates: evaluation of a special step drill, Composites Science & Technology 2009; 69: 2376-82. [27] Hocheng H, Puw Y, Yao KC. Experimental aspects of drilling of some fibre-reinforced plastics, Proc. Machining Comp. Mat. Symp., ASM Materials Week, 1992, 127-38. [28] de Albuquerque VHC, Tavares JMRS, Durão LMP. Evaluation of delamination damage on composite plates using an artificial neural network for the radiographic image analysis, Journal of Composite Materials 2010; 44: 1139-59. [29] Bodyanskiy Y, Kolodyazhniy V, Otto P. Neuro-fuzzy Kolmogorov s network for time series prediction and pattern classification. Springer LNCS 3698:91-202, 2005. [30] Elliott DL, Better A. Activation function for artificial neuronal networks, ISR Technical Report TR 93-8, Institute for Systems Research, Univ. of Maryland, 1993. [31] Singh V, Rao SM. Application of image processing and radial basis neuronal network techniques for ore sorting and ore classification, Minerals Eng. 2005; 18: 1412-20. [32] de Albuquerque VHC, Silva CC, Menezes TIS, Farias JP, Tavares JMRS. Automatic evaluation of nickel alloy secondary phases from SEM images, Microscopy Research and Technique 2011; 74 (1); 36-46. [33] Chen WC. Some experimental investigations in the drilling of carbon fibre-reinforced plastic (CFRP) composite laminates, Int J Machine Tools & Manuf 1997; 37: 1097-1108. 17

[34] Davim JP, Campos Rubio JC, Abrão AM, A novel approach based on digital image analysis to evaluate the delamination factor after drilling composite laminates, Composites Science and Technology 2007; 67 (9): 1939-1945. 18

FIGURE CAPTIONS Figure 1 Delamination mechanisms: a) peel-up; b) push-down. Figure 2 Drill geometries used: a) twist; b) Brad; c) step. Figure 3 Experimental setup. Figure 4 Sequential setup adopted to obtain the measurements from radiography images of the damage caused by the drilling operation: a) original image; b) after filtering; c) segmentation of interest region; d) measurement of maximum diameter and hole diagonal. Figure 5 Tool material and feed rate effects on maximum thrust force. Figure 6 Tool material effects on thrust force, delamination and mechanical properties. Figure 7 Tool geometry consequences on thrust force, delamination and mechanical properties. Figure 8 Correlations between delamination factor/ adjusted delamination factor and mechanical tests results: a) open-hole stress; b) bearing stress; c) flexural stress. 19

TABLE CAPTIONS Table 1 Comparison of tool material results. Table 2 ANOVA for maximum thrust force of tungsten carbide drills. Table 3 ANOVA for delamination factor of tungsten carbide drills. 20

TABLES Table 1 Drill Max thrust force Adjusted Open hole test Bearing Flexural material [N] Delamination strength [MPa] test strength factor (F da ) strength [MPa] [MPa] WC 105 1.146 602 565 2146 PCD 125 1.188 626 529 1979 21

Table 2 Factor Level S/N SS DF V %P F 1 2 3 Geometry -39.82-37.48-35.36 30 2 15 22 20 Feed rate -32.98-38.55-41.12 104 2 52 75 68 Spindle speed -38.39-37.29-36.97 3 2 2 2 2 Error 2 2 1 1 TOTAL 139 8 100 SS sum of squares; DF degree of freedom; V variance; %P percentage of contribution; F- result of F-test. 22

Table 3 Factor Level S/N SS DF V %P F 1 2 3 Geometry -1.44-2.61-1.67 2.3 2 1.2 35 8 Feed rate -1.07-2.09-2.55 3.4 2 1.7 53 12 Spindle speed -1.63-2.22-1.87 0.5 2 0.3 8 2 Error 0.3 2 4 TOTAL 6.5 8 0.1 100 SS sum of squares; DF degree of freedom; V variance; %P percentage of contribution; F- result of F-test. 23

FIGURES a) b) Figure 1 a) b) c) Figure 2 Figure 3 24

a) b) c) d) Figure 4 Figure 5 25

Figure 6 Figure 7 26

a) b) Figure 8 c) 27