PROCEDURES FOR IDENTIFICATION OF OFFSHORE PLATFORM STRUCTURAL DAMAGES Paula F. Viero and Ney Roitman COPPE/UFRJ - Civil Engineering Department E-mail: paula@labest.coc.ufrj.br Abstract The present work reports on the evaluation of the performance of some identification methods applied on a small scale hydroelastic model of a fixed offshore platform designed and constructed according to the Similitude Theory. Experimental tests were carried on the model in order to verify the behavior of the structure due to and to deck mass changes and in order to evaluate the feasibility of the application of the used methods. NOMENCLATURE (<i>x)j j mode shape of the und structure (x) (< >p)j : j mode shape of the d structure (p) J^X,L ' J component of the und structure mode shape (^) corresponding to the mode shape L J^P,L J component of the d structure mode shape (< >p) corresponding to the mode shape L fsj : natural frequency of the model obtained using the Similitude Theory (Hz) fp : natural frequency of the prototype (Hz) N : number of included coordinates L max number of experimentally measured mode shapes INTRODUCTION The increasing depth of water in offshore oil exploitation made it hard to perform the visual inspection of structural s. These problems led to the development of simpler monitoring techniques for identification by
266 Offshore Engineering inspecting at changes in the offshore structure modal characteristics. This idea is being developed since the early 1970s^. Recently, the number of technical references on detection/location and fault diagnostics using experimental and theoretical modal analysis has been increased. One believes that a periodic evaluation of existing methods and procedures is desirable so that useful advantages and limitations of these methods can be emphasized. In such a manner, this paper evaluates the performances of some detection/location methods, using eigenvectors, on a small scale hydroelastic model of a fixed platform designed and constructed using the Similitude Theory. The following methods were used: Modal Assurance Criterion (MAC/, Coordinate Modal Assurance Criterion (COMAC), Modal Scale Factor (MSF/ and Change in Modal Vector Perpendicular to Predominant Modal Direction^. The modal characteristics of the perfect model were taken as a reference for comparison with the results obtained from the same model when some joint was imposed. In both cases the modal characteristics were obtained through white noise type random excitation. The results were analyzed by using the 'Least Squares Complex Exponential' (Prony Method/ and the main dynamic parameters relating to the first three flexural and to the first torsional vibration modes were determined. A preliminary analysis of these results concluded that the MAC and MSF methods showed sensitivity to. Although the COMAC method showed sensitivity to, it was unable to clearly identify the location. Changes in Modal Vector Perpendicular to Predominant Modal Direction seems to be of great importance to identify in offshore platforms. Tests carried on the model considering only deck mass changes showed differences between the behavior of the structure due to and to deck mass changes. Results obtained from the tests carried on the model in order to evaluate the feasibility of the application of the used methods showed that the tests could be applied on the prototype. MODEL DESCRIPTION The model constructed for observation of the physical behavior in air was a 1:85 scaled model of a fixed platform with four legs in 300 m depth water designed by Petrobras-Brazilian State Oil Company. A general view of the model constructed with ABS, polystyrene and polyurethane tubes is shown in Figure 1, and the main geometric and physical
Offshore Engineering 267 characteristics of the prototype and of the model are given in Table 1. Figure 2 shows thefrontalview and one typical section of the model. Figure 1 - General view of the model Table 1 - Prototype and Model Main Characteristics Characteristics Height of tower Base dimensions Top dimensions Jacket weight Deck weight Duct weight Risers Prototype Mod el 313.50m 3.69 m 88.59 x 90.59 m 1 04 mx 1.06 m 25.00 x 25.00m 0.29 mx 0.29 m 85 645.38 kn 0.505 kn 90 000 kn 0.364 kn 37 751.35 kn 0.011 kn Group of 36 ducts The physical similitude conditions necessary for modeling an offshore structure have been presented previously in referenced These conditions lead to the scale factors presented in Table 2.
268 Offshore Engineering 0.294 Damages: H5 - Face A D6, D7, D9 - FaceB FaceB 0.483 Face A Dimension in meters 5th Seccion Figure 2 - Frontal view, typical cross section of the model- location Kg, Table 2 - Model Scale Factors Modeled physical parameters: Scale factor Length, L KL Outside diameter, D KL Wall thickness, d KL\(Kc)-' Axial stiffness, EA Kp f KL Damping factor, 1 Bending stiffness, EJ Kp f. KL Immersed or dry weight Kp f. KL Related fundamental parameters: Cross-section area, A Moment of inertia, I Modulus of elasticity, E Frequency, f Time, T Acceleration (ac) KL' KL' Kp f. KL (KL)-" (KL)'" 1 are respectively geometric, elastic modulus and fluid density scales.
Offshore Engineering 269 The design, construction and assessment of the first natural frequencies of the model were carried out in parallel with an adjustment through an integrated theoretical numerical experimental analysis. Details of this development can be found in reference. The natural frequencies of the model obtained using the Similitude Theory (f j = V85 xfp), numerically and experimentally (last column) are presented in Table 3. One can see the good correlation obtained. More details about the model design and construction can be found in reference. Table 3 - Natural Frequencies (Hz) of the Model Vibration Modes 1" Flexural- X 1" Flexural -Y l^torsional 2 Flexural - Y 2 Flexural - X 3 Flexural - Y 3 Flexural - X Similitude Theory 2.40 2.43 6.56 9.16 9.04 Numerical Model 2.38 2.43 6.99 9.40 8.88 17.34 16.82 Constructed Model 2.55 2.60 6.96 8.50 8.22 14.67 14.57 These values weren't identified until the 40 analyzed mode DAMAGE DETECTION TECHNIQUES In this item, some of the proposed methods to detect/locate in offshore structures are described. Modal Assurance Criterion (MAC) The Modal Assurance Criterion (MAC/ correlation technique takes, two sets of eigenvectors from either test and finite element analysis or test and test. In this work, the MAC was used to compare the eigenvector from the perfect structure to the d structure, in order to identify the. The MAC is defined as: N MAC(p,x) = N I(+p) Lj=i J (1)
270 Offshore Engineering This procedure results in a matrix - order m x nip - where is the number of und modes and nip is the number of d modes. The leading diagonal of the MAC matrix will indicate the similitude (or not) between two sets of eigenvectors. If the correlated mode shapes are similar, the leading diagonal of the MAC matrix will be approximately unity. If they are different, the leading diagonal will be less than unity and major or equal to zero. Although it is indicated that there is a disparity between the two sets of mode shapes, it does not show explicitly where the is located in the structure. Obviously, it is of greater importance to know the position, or at least, the region of the (s). In order to locate the, the Coordinate Modal Assurance Criterion (COMAC) has been developed. Coordinate Modal Assurance Criterion (COMAC) The Coordinate Modal Assurance Criterion (COMAC/ calculates a correlation factor for the und and d experimental coordinates in all mode shapes for a specific DOF (Degree of Freedom) j: COMAC(j) = Z(jx,: LL=I faf / 2. j+p LL=1 (2) This complete procedure results in a list of COMAC values of magnitude between zero and the unity that can be analyzed by the same criterion as MAC. Modal Scale Factor (MSF) The Modal Scale Factor (MSF) represents the 'slope' of the best straight line through the points for a pair of modes shapes, in this case, the und and the d mode shapes. This quantity is defined as: N (3) It should be noted that this parameter gives no indication as to the quality of the fit of the points to the straight line, simply its slope.
Offshore Engineering 271 Change in Modal Vector Perpendicular to Predominant Modal Direction This procedure is based on the fact that the in offshore structures can be detected by analyzing the changes in the modal vector perpendicular component to the predominant modal direction measured only on the deck of the platform to the first flexural and first torsional modes''. The modal vector component in the Y direction of thefirstflexuralmode in the X direction, the modal vector component in the X direction of thefirstflexural mode in the Y direction and the modal vector perpendicular component of the first torsional mode present substantial changes when s occur. These changes are probably caused by the decrease of the platform stiffness, inducing platform eccentricity mainly on the deck. TEST PROGRAM Preliminary tests were carried out on the model vibrating in air. Frequency Response Functions (FRFs) were then obtained for the three first global flexural modes in the X direction (see Fig. 2) and for thefirsttorsional mode of the small scale model under two distinct conditions: with and without structural integrity. For the latter condition joint s were intentionally imposed to the model. The s were chosen according to a fatigue analysis that indicated the members with major probabilities of occurrence after ten years of platform operation. Figure 2 indicates these d tubular members with one of their end joints entirely cut out. The is classified according to the locations of the d members along the height of the tower and where the d members are located (face or horizontal section) as follows: - DJ : the d member is located on one face of the platform (diagonal member) - HJ : the d member is located on a cross section (horizontal member). where J = 5, 6, 7, 9 indicates the level along the tower. The tests were performed separately for each one of the above s cases. After each test the d members, in the level considered for investigation, were repaired by gluing firmly and the integrity of the whole model was thoroughly verified before the next test on the model with new s imposed. The FRFs were obtained from forced random excitation of the white noise type. The dynamic force was measured by a piezoelectric load-cell and
272 Offshore Engineering the structure response was measured by micro-accelerometers of the same type. Figure 3 indicates schematically the sensors location. Two different instrumentation were used according to the applied method: - instrumentation along the height of the tower on one of their legs (MAC, COMAC and MSF); - instrumentation on the deck (Change in Modal Vector). --Shaker Load-cell " Accelerometer D1X Instrumentation on the deck Instrumentation along the height of the tower Figure 3 - Sensors location Figure 4 shows a sketch of the equipment used for data acquisition and processing of the dynamic signals. In order to verify changes in the behavior of the structure due to deck mass changes some experimental tests were carried on the model. Two different typical situations of deck mass changes were considered: - situation 1: 16.7 of the total mass of the deck was evenly removed; - situation 2: 14 of the total mass of the deck was removed according to the sketch showed in figure 5. TEST RESULTS AND ANALYSIS The exciting force and acceleration response signals were processed by using a two channels spectrum analyzer HP 35660A model, with an average of
Offshore Engineering 273 50 samples and a frequency resolution of ± 0.08 Hz. The coherence function between force and response signals was always verified. T I Spectrum Analyzer, jh Power Amplifier Output - White Noise HP-IB Board Accelerorneter Figure 4 - Data acquisition and processing system of the dynamic signals Removed mass Remained mass Deck Figure 5 - Change of the deck mass in situation 2 The FRFs processed by the spectrum analyzer were automatically transferred to a microcomputer using a software developed for this purpose. Afterwards, these FRFs were analyzed using the 'Least Squares Complex Exponential' (Prony Method/ implemented by a software package developed for this purpose^, obtaining the modal characteristics of the structure.
274 Offshore Engineering Damage Identification Table 3 shows the elements on the leading diagonal of the MAC matrix for the small scale model when s are imposed. The values presented in the second column are obtained for the model on similar conditions (without s). The major changes of the MAC values are enhanced. Mode 1" Flexural 2 Flexural 3 Flexural l^torsional Table 3 - MAC H5 Diagonal Values D6 D7 0.86 D9 0.97 One can observe from Table 3 that the MAC values for thefirstflexural and for thefirsttorsional modes are almost insensitive to the imposed. For the second and third flexural modes the MAC values are, in most of the cases, significant when it is compared to the MAC values on the similar conditions. The MAC values are able to indicate the presence of in most of the cases when the second and third flexural modes are analyzed together. Existence of is not shown, only for case H5. COMAC values are shown in table 4. The underlined values indicate the level of the imposed, as can be seen in Figure 2. The nodes presented in the following tables are shown in Figure 3. It can be seen (from this table) that the COMAC values presented changes for cases D6, D7 and D9. Although COMAC values showed sensitivity to, they can only indicated the region of the. They were unable to identify the exact position. Table 4 - COMAC Values Node 1 2 3 4 5 6 7 8 9 10 0.97 H5 D6 0.97 D7 0.89 0.88 0.86 0.87 0.94 0.95 0.93 D9 0.97 0.95 0.94 0.95
Offshore Engineering 275 The results presented on tables 3 and 4 could be better if the number of measured points were greater. This fact would increase the accuracy of the mode shapes. Table 5 shows that the MSF values can indicate existence of in all cases. One can observe from this table that only the first flexural mode is almost insensitive to the imposed. Table 5-MSF Values Mode jffjexurat 2 Flexural 3 Flexural l" Torsional 1.01 099 1.04 H5 0.94 1.02 132 D6 1.04 0.87 1.23 D7 1.05 1.50 0.93 D9 0.78 1.34 The mass-normalized modal shape measured on the deck of the platform for thefirstflexuraland for thefirsttorsional modes are presented on tables 6 and 7, respectively. These tables also present the percent changes observed in the modal vector. Table 6 - Modal Vector Values on the deck (Kg"") -1" Flexural Mode Mode lflex. XDir. 1" Flex. YDir. Accel. D1X D2X D3Y D4Y D3Y D4Y 15.03 14.48 3.04 3.57 3.38 H5 14.48 14.77 6.07 7.48 3.7 2.0-146.1 - D6 14.89 13.42 5.59 4.21 3.96 1.0 7.3-38.5-20.7 These modal coordinates were not identified D7 14.80 14.13 4.10 5.16 5.14 4.35 1.5 2.4-69.7 44.0 32.6 D9 14.18 14.32 4.03 5.87 3.87 4.61 5.7 1.1-93.1 8.4 40.6 Table 7 - Modal Vector Values on the deck (Kg"') -1" Torsional Mode Accel D1X D2X D3Y D4Y 1.96 1.94 1.80 1.74 H5 2.68 4.08 4.05 3.91 36.7 110.3 125.0 138.4 D6 3.49 343 4.07 3.85 78.1 102.6 126.1 134.8 D7 3.25 3.52 3.81 3.63 65.8 81.4 111.7 121.3 D9 1.66 1.85 5.03 7.41 15.3 4.6 179.4 351.8 Table 6 shows that the modal components measured in the Y direction (perpendicular to predominant modal direction - X) for thefirstflexural mode in X direction present substantial changes in their values when s are
276 Offshore Engineering imposed. These parameters measured in predominant modal direction - X, present small changes. It can be seen from table 7 that the modal vectors of the first torsional mode measured in both directions present substantial changes when s were imposed, mainly in the Y direction. These changes can be seen in figure 6 that shows a comparison between the FRFs obtained from the und model and when some s were imposed. Deck Mass Changes The MAC and MSF values obtained when two different typical situations of deck mass changes were considered are presented on table 8. It can be seen that these values showed sensitivity not only to but also to deck mass changes. Mode l" Flexural 2 Flexural 3" Flexural 1" Torsional Table 8 - MAC and MSF Values - Deck mass change MAC Situation 1 0.97 0.97 0.95 Situation 2 L_ 1.01 099 1.04 MSF Situation 1 1.04 0.80 2.08 Situation 2 0.79 0.93 2.42 The COMAC values and the percent changes () observed in the modal vector measured on the deck of the platform for the first flexural and for the first torsional modes obtained from the tests with deck mass changes are presented on tables 9 and 10, respectively. Table 9 - COMAC Values - Deck mass change Node 1 2 3 4 5 6 7 8 9 10 0.97 Situation 1 0.95 0.93 0.94 0.95 0.91 0.85 0.81 0.85 Situation 2 0.95 0.90 0.89 0.88 0.90 0.89 0.89 0.83 0.85 0.87
Offshore Engineering 277 One can observe from table 9 and table 10 that the COMAC values in almost all DOF (degrees of freedom) and the modal components measured in the X direction (predominant modal direction) for thefirstflexural mode in X direction showed sensitivity to deck mass changes. Thus, the situation of deck mass changes can be identified by analyzing the COMAC values and the modal vectors measured on the deck of the platform for thefirstflexuralmode in the X direction. Table 10 - Percent Change () in the Modal Vector measured on the deck Deck mass change Mode 1 Flexural X Direction 1" Flexural Y Direction l^torsional Accel D1X D2X D3Y D4Y D3Y D4Y D1X D2X D3Y D4Y Situation 1 +13.4 +15.7-37.2 +8.7 +2.6-3.3 +31.1-14.8-11.4 These modal coordinates were not identified Situation 2 +9.0 +10.9-38.6-30.5-32.2 +19.4 +3.7 +16.3 +19.7 FINAL REMARKS The results shown in this paper lead to conclude that the MAC and MSF methods were sensitive not only to the imposed s but also to deck mass changes. The behavior of the structure due to and to deck mass changes can be distinguished by analyzing the COMAC values and the modal vectors measured on the deck of the platform. In the situation, the maximum changes of COMAC values occurred only in the vicinity region of the, while in the deck mass changes, the COMAC values showed sensitivity in almost alldof. For thefirstflexuralmode in the X direction, the modal vectors measured in the X direction did not present substantial changes in their values when some were imposed, while the same components in the same direction (X) showed sensitivity to deck mass changes. The COMAC values showed sensitivity to in most of the cases, but they were unable to clearly indicate the location. Here, the maximum changes of COMAC values occurs in the vicinity region of the, so it does not occur exactly at the defect position.
278 Offshore Engineering The Change in Modal Vector Perpendicular to Predominant Modal Direction seems to be of great importance to identify in offshore platforms, since with only a deck instrumentation it would be possible to detect s. The modal vectors measured in the Y direction for the first flexural mode in X direction presented substantial changes in their values when some were imposed. The modal vectors of the first torsional mode measured in both directions also present substantial changes when s were imposed. Tests carried on the model showed that this procedure could be applied on a prototype by using a hydraulic actuator with a stroke of 30 cm and a reaction mass of = 10 ton, located on the deck of the platform. It could be possible to excite the prototype up to the third flexural mode and measured the structure response with great accuracy. REFERENCES [1] Vandiver; J.K., "Detection of Structural Failure on Fixed Platforms by Measurement of Dynamic Response", Offshore Technology Conference, paper 2267, Houston, 1975. [2] Coppolino, R.N. and Rubin, S., "Detectability of Structural Failures in Offshore Platforms by Ambient Vibration Monitoring", Offshore Technology Conference, paper 3865, Houston 1980. [3] Shahrivar, F. and Bouwkamp, J.G., "Damage Detection in Offshore Platform Using Vibrations Information", International Offshore Mechanics and Arctic Engineering, 1984. [4] Idichandy, V.G., Ganapathy, C. and Rao, P.S., "Structural Integrity of Fixed Offshore Platforms", IABSE Colloquium, Bergamo, 1987. [5] Ewins, D.J., "Modal Testing: Theory and Practice", Research Studies Press Ltd., London, 1984. [6] Lieven, N.AJ. and Ewins, D.J., "Spatial Correlation of Mode Shapes, the Coordinate Modal Assurance Criterion (COMAC)", Proceedings, 6 IMAC, Vol. I, pp. 690-695, Kissimme, Florida, USA, 1988. [7] Mergeay, M., "Least Squares Complex Exponential Method and Global System Parameter Estimation Used by Modal Analysis", Proc. of the 8 Intl. Modal Analysis Sem., Leuven, 1983. [8] Viero, P.F., "Procedures for Identification of Offshore Platform Structural Damages", D.Sc. Dissertation (in Portuguese), COPPE/UFRJ, Rio de Janeiro, Brazil, 1996. [9] Cameiro, F.L.L.B., "Some Aspects of The Dimensional Analysis Applied to the Theory and Experimentation of Offshore Platforms", Offshore Engineering, Vol. 2, p. 542-558, 1980. [10]Rosa, L.F.L., "Modal Parameter Estimation using an Optimization Technique", D.Sc. Dissertation (in Portuguese), COPPE/UFRJ, Rio de Janeiro, Brazil, 1996.
Offshore Engineering 279 0.4 0.3 l^flexuralmode 0.2 0.1 2.5 5 7.5 Frequency (Hz) Und 10 12.5 Damage D7... Damage D9 0.08 0.06. 0.04 _ 0.02 _ 0.25 0.2 0.15 (N i 01 u. 0.05 _ Und r'flexuralmode i 5 7.5 Frequency (Hz) Damage D6 _. Damage H5 [D4Yl 1" torsional mode 12.5 0 _ Und 5 7.5 Frequency (Hz) Damage D7.. Damage D9 Figure 6 - FRF measured on the deck 12.5