You are on page 1of 15

The 2011 World Congress on Advances in Structural Engineering and Mechanics (ASEM'11+) Seoul, Korea, 18-22 September, 2011

Shaking Table Specimen Interface Design in Substructure Testing *Trung Vien Phan1), Van Thuan Nguyen2), Nasser Mohammad Khanlou3) and Uwe E. Dorka4)
1)

Steel and Composite Section, University of Kassel, Germany phanvien@uni-kassel.de,2)thuan@uni-kassel.de,3)khanlou@uni-kassel.de 4) uwe.dorka@uni-kassel.de

1), 2), 3), 4)

ABSTRACT Substructure testing using shaking tables requires careful considerations regarding the interaction between the substructure specimen and the table. The design of the interface between them is of particular importance since it has to recreate accurately the interface between the numerical structure and the specimen, and also allow high-resolution measurements of the emerging coupling forces. These play a major role in the substructure algorithm, which needs them as a feed-back to calculate the motion of the table using a step-wise time integration scheme. This paper summarizes the experience gained in designing such tests for three projects, namely: CEA-UNIKA5), SUBSHAKE6) and E-FAST7) project. Various arrangements of the interface between the specimen and the table were made, especially with respect to the measurements of the coupling forces between specimen and numerical model. The experience gained with these arrangements is reported. Conclusions are drawn regarding the requirement for stiffness and the measurements of forces in such interfaces.

Doctoral student Dr. -Ing. 4) Prof. Dr. -Ing. 5) CEA-UNIKA Project: Common action between University of Kassel and the Commissariat lnergie Atomique, agreement No. SAV 33 156, 2006-2009. 6) SUBSHAKE Project: Development of substructure test in real time for hydraulic shaking tables, under contract Do 360/22-1,2, German Research Foundation (DFG), 2006 2010. 7) E-FAST Project: Design study of a European Facility for Advanced Seismic Testing, EC Grant number. 212109, 2008 2011.
2)

1), 3)

1. INTRODUCTION Substructure testing is an advanced testing method in which most structures can be tested in a dynamic manner without testing the entire structural system. To achieve this, the system is divided into a numerical part, the one that can be modeled correctly on a computer, and an experimental part, where the dynamic properties are unknown or are difficult to model. Thus it needs to be tested as a real physical model. The substructure algorithm is based on a time integration scheme. A number of such schemes have been developed (Thewalt 1987, Nakashima 1990, Shing 1991, Combescure 1997, Pegon 2000), but in the studies reported here, the algorithm developed by Dorka (Dorka 1990) is used because of its superior performance and versatility in Real Time Substructure Tests (RTST), even for aerospace applications (Bayer 2005). Based on the general time discrete integration (Zienkiewicz 1977), Dorka developed a substructure algorithm using implicit integration with digital feedback (Dorka 1990). The digital feedback mechanism is described in Fig. 1a while the flow chart of the algorithm is shown in Fig. 1b (Dorka 1990, 1998, 2002, 2011). The displacement vector, u i +1 , of the numerical model at the next step is described as a linear control equation Eq. (1).
i u i +1 = u0+1 + G f ri +1 + f c

i +1

(1)

i where: uo+1 is a vector of explicit displacements that are known at the beginning each step, G is

the gain matrix, f ri +1 is the vector of nonlinear numerical forces and, f ci +1 is the vector of coupling forces that are measured on the specimen. In this digital feed back algorithm (see Fig. 1), the non-linear numerical forces, f r , and the currently measured coupling forces, f c , are fed back at the sub steps, which are equally distributed over the time step (Fig. 1a). At the end of each step, the equilibrium error is calculated and the error force is identified. The error force is compensated at the beginning of the next time step (Fig. 1b). Dorka proposed the PID error force compensation (Dorka 1990, 1991, 1998) and it was successfully applied in many substructure tests (Dorka 1991, 1998, 2002, 2006, 2007, Bayer 2005, Nguyen 2011). To allow the compensation adapting automatically to changing testing environments, Nguyen and Dorka (Nguyen 2007, 2009) introduced an adaptive force compensation based on data model and online system identification. The adaptive force compensation is currently tested using a non-linear Tuned Mass Damper (TMD) with the friction device UHYDE-fbr (Dorka 1995, US Patent number 5456047) and the hydraulic shaking table at University of Kassel (UNIKA).

Time integration algorithm:

i +1

=u

i +1 0

+G f

i +1 r

+ fc

i +1

Calculate explicit displacement at beginning of step j=0

u io+1

is a linear control equation with constant gain G

ui+1 Displacement ui ui-1 u0i-1 i-1 i u0i j=1 2 k-1 k

Apply displacement at each substep j i i +1 j u = u0 (1 ) + u0 ( ) + G( f r + fc ) k k G(fr + fc) u0


i+1

Calculate restoring force on nonlinear numerical substructure fr Error force compensation j=

Measure coupling force on experimental substructure fc No

i+1 step

j = j+1 Yes - Calculate velocity and acceleration at the end of step - Calculate error force of the equilibrium equation (b) Flow chart of substructure algorithm

(a) Linear control mechanism

Fig. 1. Substructure algorithm with digital feedback and error force compensation (Dorka 2002, 2011)

In order to perform RTST, actuators are used to impose the computed movement on the experimental substructure and load cells are used to measure the coupling forces. Therefore, an important issue in substructure testing is to design a proper interface between the two substructures. It requires not only reliable measurements of the coupling forces but also a faithful representation of the actual interface between substructure (specimen) and numerical structure. In the flow chart (Fig. 1), it can be seen that the role of the interface between table and specimen is to transmit the calculated displacement and to measure the coupling forces as exactly as possible. To transmit the calculated displacement exactly, the displacement control system should be able to provide accurate response. The stiffness and damping properties of the coupling should be well represented between specimen and table. In order to measure correctly the coupling force, load cells with high resolution and low noise are required. Due to inaccuracies in the transformations, the controllers and mechanical limitations of the actuators, positioning errors will occur especially in continuous RTSTs. Although they can be minimized by adaptive controllers (Stoten 2001, Wallace 2005, Nguyen 2008), but they cannot be completely avoided. They appear as an overshoot or undershoot with a certain noise level that depends on the quality of the testing equipment. In addition, measurement errors in the coupling force and also regular errors (incorrect amplification or insufficient resolution) may enter the algorithm. Except for the positioning errors of the actuators, they can be avoided completely by a proper test setup. In real applications, TMDs are mounted directly on the structure (Fig. 2). In a substructure test, the structure is modeled numerically and the TMD is placed on a shaking table. Load cells

must be installed between the shaking table and the specimen to measure the coupling forces and this will introduce some inaccuracies into the interface. In order to minimize this interference and make the behavior as close to the real structure as possible, the stiffness of the connection should be high in this case and the measurement of the coupling forces must have high accuracy.

(a) Bridge Britzer Damm, Berlin

(b) TMD fixed below the bridge

Fig. 2. A typical TMD application for a bridge (GERB Engineering GmbH)

2. INTERFACE DESIGNS 2.1. Interface design in the test setup of the CEA-UNIKA project CEA-UNIKA was a project supported by UNIKA and CEA. The complete structure consists of a two-storey steel frame and two TMDs located at the second floor (Fig. 3). Each TMD can vibrate only in one direction. The steel frame was designed for tests with excitations in the two horizontal directions (Dorka 2006).

Eigenfrequencies 1st 2nd 3rd

Hz 3.75 4.5 12

1: AZALEE shaking table; 2: Frame of two stories; 3: Added mass at first floor; 4: Rigid columns; 5: TMD 2; 6: TMD 1

Fig. 3. Reference tests of the full structure including the steel frame and the TMDs using the AZALEE shaking table in CEA (Dorka 2006) Fig. 4 shows the interface in the reference tests, in which the TMDs are placed on the steel frame. Fig. 5 shows instead the interface in the substructure test, in which the TMDs are placed on two distributed shaking tables. Each interface includes seven single load cells. They are installed in order to connect the steel frame and the TMD so that they can measure the coupling forces. Three load cells are installed in the horizontal directions X and Y to transmit the horizontal movement and to measure the horizontal coupling forces (labels 2 and 5 in Fig. 4), and four are installed in the vertical direction (label 4 in Fig. 4). The substructure tests were performed with a series of changing parameters such as time step, t, in 10 ms or 20 ms, number of sub-steps, k, from 2 to 5 and with or without PID error force compensation, P ranges from 0.85 to 1.0.

1: Second floor of the frame; 2: Load cell for measuring coupling force of TMD 1; 3: TMD 1; 4: Load cell for checking other coupling forces; 5: Load cell for measuring coupling force of TMD 2; 6: TMD 2

Fig. 4. Interface between the steel frame and the two TMDs in the test setup in CEA (Dorka 2006)

Fig. 5. Interface between the shaking table and the TMDs in the test setup in CEA (Dorka 2006) Fig. 6 shows the comparison between the substructure test and the reference test under an earthquake load. In Fig. 6a, it can be seen that there are large pulses in the coupling force during the substructure test. These are also present in the reference test, but they are smaller. The pulses were generated by play in the ball bearings that connect that load cells to the table and specimen. The ball bearings are needed to avoid forces perpendicular to the load cells. The reason behind the smaller pulses in the reference test is a smaller stiffness of the interface there due to the elasticity of the frame. Since these are high-frequency pulses, they hardly excite the structural modes and thus have a negligible influence on the structural response. Also, they do not de-

stabilize this RTST algorithm which de e T m, emonstrates its robustn s ness against such vibr rations that may also arise in sh haking table due to oth reasons. es her

Fig. 6. Compar rison of the coupling fo orces (a) and the displa acements on the 2nd-flo (b) n oor be etween a sub bstructure te of the TM 2 and those of the reference t est MD e test, under Kobe K earthq quake, time s step: 10ms, k = 2 (Dork 2007). ka

ign est f oject. 2.2. Interface desi in the te setup of the SUBSHAKE pro In th project, t referenc tests usin a 3-storey steel fram and a TM (Fig. 7) have been his the ce ng y me MD done in IZIIS (Ins n stitute of E Earthquake Engineerin and Eng ng gineering Se eismology in Skopje, Republi of Maced ic donia) while the substru e ucture tests with the TMD as subs s structure an the steel nd frame as numerical model (Fig 8) were p l g. performed at UNIKA. T project was supported by the a The t German Research Foundation (DFG) under contrac Do 360/2 n n ct 22-1,2. Usin the contr ng rollable bidirectional friction device UH HYDE-fbr, d different non n-linear cou upling betw ween the TM and the MD ame was ac chieved. The frame, sy e ymmetric in one directi and asy n ion ymmetric in the other, n steel fra will allo investiga ow ating the eff of torsi in the st frame on substructu testing. ffect ion teel ure The interfaces in both ca ases, the re eference and the subst d tructure tes are sim sts, milar. Each interface includes t three single load cells ( (labeled as 3 in Fig. 9 and as FT TX1, FTY2 and FTY3 in Fig. 10) in ord to meas der sure the ho orizontal co oupling forc ces. Also, t there are four swivel fo of D n g. supports to create the same connection o the TMD to the steel frame in IZIIS (Fig 7) or the shaking table in U g UNIKA (Fig 8) and av g. void horizo ontal forces being pick up by th vertical ked he supports. The design of the swivel sup pports with spherical bearings a h allowed adj justing the distance between th base fram and the s e he me shaking table or the ste frame, re eel espectively.

EigenE frequencies 1st 2nd 3rd 4th 5th 6th 7th

Hz H 1.5 2.1 15 2.6 65 3,6 68 5.1 15 6.1 7.6 61

1: Doubl beam as m le mass; 2: Load cells; 3: Co d onnection pla ates; 4: 3-sto orey steel fram 5: Accel me; lerometers; 6 6: Shaking ta able; 7: Strain gages; 8: M 400kg as live load ( first and s n Mass a (in second storie es); 9: Air pressure valve e

Fig. 7. T 3-story steel frame with the TMD on the shaking tab at IZIIS The e ble

1: Hyd draulic cylind (X-direct der tion); 2: Shak king table (1.8m x 1.8m); 3: Air press sure valve; 4: Springs; 4 5: Fr riction device UHYDE-fb 6: Displa fbr; acement trans sducers; 7: L Load cells; 8: Rails; 9: Hy : ydraulic cylin nder (Y-dire ection); 10: S Swivel suppo orts; 11: Base frame of the TMD e Fig. 8. The TMD with base f . D frame on its s supports as a substructure on the shaki table at UNIKA e ing U

The TMD (Fig 9a) consi of a m g. ists mass, four diagonal spr d rings in tan ndem and the friction t device U UHYDE-fbr, which all lows contro olling the fr riction force between t mass an the base e the nd frame. T mass m The moves on top of perpend p dicular rails which are installed o the base frame. The s, e on f springs and the fric ction device UHYDE-f are plac between the mass a the base frame. In e fbr ced n and e this con nfiguration, the TMD has two horiz zontal DOF with diffe Fs erent freque encies. Fig. 9a shows a view of the TMD whil Fig. 9b sh e le hows the de etail of the s swivel supp ports.

(a) D Detail of the TMD

(b) Swivel supp ) ports with tw spherical bearings on wo b both en nds

1: The b base frame of TMD; 2: The swivel su f upports with two spherica bearings at the ends; 3: Load cells t al t : frame and TM 4: Diag MD; gonal springs; 5: Bearings s between f

Fig. 9. Details of the TMD (l . left) and con nnection be etween the T TMD and th shaking table with he t swiv supports (right) vel s

ATX, ATY: Acceleromet A ters; DTX, D DTY: Displac cement transd ducers; FTX, FTY: Load cells

Fig. 10. Ins strumentatio on the TM on MD In th project, tests were performed under sine sweep and earthquak excitation and the his e d ke ns pressure in the frict e tion device was either zero (linear TMD) or c r constant (ela astic-plastic coupling). c . In the su ubstructure tests, the n number of su ub-steps, the most important param meter in the algorithm, e , varied b between 2 and 6. The PID error compensa e r ation was t tested with a variation of the P n paramet ranging from 0.8 to 1.0 while the adaptiv force com ter o ve mpensation was tested with order nu rangi from 3 t 10 and th forgetting factor in the range b ing to he g n between 0.9 to 0.99. 90 Fig. 1 shows a comparison of the cou 11 n upling force and displa e acements between a su ubstructure and the reference te in the SU est UBSHAKE project. E

Fig. 11. A comparison of the coupling forces (a, b, c) and the displacements (d, e, g) between substructure test Sub02 (time step t = 10ms, number of sub steps k = 3, PID compensation with P = 0.9) and reference test No7 under earthquake Petrovec 1979 excitation, with air pressure p = 0.

In the SUBSHAKE project, the pulses due to play in the bearings was reduced (compare Fig. 11a and Fig.6) but not entirely avoided although the swivel supports were pre-stressed. The coupling force shows a better match than in the CEA-UNIKA tests (Fig.6), but the accuracy is still not satisfactory. The displacement response (Fig. 11d) does not match so well, which in this case is due to the higher complexity of the steel frame: The numerical model used in these RTSTs did not match as well as the model in the CEA-UNIKA tests. 2.3. Interface design in the test setup of the E-FAST project The E-FAST project is a design study for a new European testing facility, and is performed in collaboration between five leading European institutions in the field of earthquake engineering. Some major goals of the new European facility include high performance, large capacity, great flexibility, strong integration with other facilities and advanced networking capabilities not only in Europe but also worldwide. In this context, the test setup at UNIKA has been used to study real time substructure testing with shaking tables, to combine shaking tables with other on-site facilities and to perform distributed testing. The concept of the test setup for reference and substructure tests is given in Fig. 12. In the reference tests, the leaf spring between actuator and table (Fig. 12a) is unlocked. Table and leaf spring serve as the first DOF and the TMD on top of the table is the second DOF. The 2nd DOF system has two eigenfrequencies 1.875 Hz and 3.025 Hz. When the leaf spring is locked (Fig. 12b) a substructure test can be performed with just TMD modeling numerically the table and the leaf spring.

Eigenfrequencies 1st 2nd

Hz 1.875 3.025

(a) Test setup for reference test

(b) Test setup for substructure test (locking device at leaf spring)

Fig. 12. The test setups for reference test (a) in which the leaf spring is unlocked and for substructure test (b) in which the leaf spring is locked by a locking device.

In using this concept, there is no difference in the structure of substructure tests and reference tests. This allows focusing mainly on the accuracy of the substructure algorithm and on the control of the hydraulic shaking table. An advanced force measurement concept using multi-directional load cells was developed and applied. Four multi-directional load cells are placed between the TMD and the shaking table (Fig. 13). Each load cell measures two horizontal forces and a vertical load. The two horizontal forces can be used as coupling forces in substructure tests while the vertical force is mainly used for adjusting the distribution of static vertical loads when placing the TMD on the table. The coupling force Fc in substructure test is the sum of the measured forces by the four load cells in the horizontal y-direction. The measurement of the coupling forces using multi-directional load cells avoids the pulses in the coupling forces observed in the previous tests. It provides high stiffness within compact dimensions.

Fig. 13. Multi-directional load cells for substructure tests with non-linear TMD on the shaking table at UNIKA The load cells have been calibrated using a dynamic testing machine with a calibration load cell (Fig. 14).
Calibration of m ulti-load ce lls 12 M lti-lo d c ll fo e(k ) u a e rc N

0 -12 0 12

-12 Re fe re nce Force (k N)

(a) Test setup for calibration

(b) Comparison between new and calibration load cell

Fig. 14. Calibration of the multi-directional load cells

More than twenty reference tests with different types of excitation (sine, sine sweep and earthquake) and air pressure in the UHYDE-fbr as well as several substructure tests with and without compensation (PID, phase lag or adaptive force compensation) have been performed (Nguyen 2011). In Fig. 15, the comparison between the results of the tests Ref016 and Sub066 under earthquake excitation is shown.

1.875 Hz, 1st eigenfrequency

3.025 Hz, 2nd eigenfrequency

1.875 Hz

3.025 Hz

Fig.15. Comparison between reference test Ref016 (Kobe earthquake 1995 excitation, amplitude is 10% of the real record TAZ090; air control pressure p = 0) and substructure test Sub066 (time step t = 10ms, number of sub step k= 4, adaptive force compensation with nu = 7, = 0.99; phase lag compensation with nu = 5, = 0.99)

Because of the solid connection of the load cells in this interface, no pulses are observed and an acceptable match is reached for the coupling force of the RTST in comparison to its reference test (Fig. 15a). Since the interface in the RTST and reference test are exactly the same the small deviations in the response around the first eigenfrequency are related to other sources. These can now be studied in detail without serious inference by the interface.

3. CONCLUSIONS Besides using an advanced substructure algorithm, the design of a good interface between specimen and shaking table has proven to be very important in order to achieve meaningful and accurate results in substructure testing. The tests where spherical bearings have been used between specimen and shaking table exhibit large pulses in the coupling force due to play in the bearings. However, these pulses have negligible influence on the substructure test when the substructure algorithm developed by Dorka is used. It remains stable and overall results, like displacements may be reproduced with good accuracy. The design of adjustable swivel supports with spherical bearings allows adjusting the gap of the bearings but it is not a solution to reduce large pulses in the coupling forces. Therefore, any interface solution based on ball bearings is not recommended, since an acceptable accuracy cannot be achieved for the coupling force, which reduces the confidence in the test results. Within the E-FAST project, a new interface using multi-directional load cells between the specimen and the shaking table at UNIKA has been developed. This new kind of interface can provide up to six force components with high accuracy, has compact size and high stiffness in the couplers. No pulses were observed and the accuracy of the coupling force was greatly enhanced. REFERENCES Bayer, V., Dorka, U.E., Fllekrug, U. and Gschwilm J. (2005), On real-time pseudo-dynamic substructure testing: algorithm, numerical and experimental results. Aero. Sc. and Tech., Vol. 9, 223-232. Combescure, D. and Pegon, P. (1997), Alpha-operator splitting time integration technique for pseudodynamic testing error propagation analysis. Soil Dynamics and Earthquake Engineering, Elsvier, Vol. 16(7-8), 427-443. Dorka, U.E. (1990) Fast online earthquake simulation of friction damped systems. SFB151 Report No. 10, Ruhr-University Bochum, Germany. Dorka, U.E. and Heiland, D. (1991), Fast online earthquake simulation using a novel pc supported measurement and control concept, Proceedings of the 4th International Conference on Structural Dynamics, pages 636-645, Southampton. Dorka, U.E (1995), Friction device for protection of structural systems against dynamic actions, Patent number 5456047, United States Patent. Dorka, U. E., Fllekrug, U. (1998), Algorithmen fr real-time pseudo-dynamische Substrukturtests, Report of the DFG project SubPSD-Algorithmen, project number Do 360/7, University of Kaiserlautern, Germany. Dorka, U.E. (2002), Hybrid experimental - numerical simulation of vibrating structures, Proceedings of the International Workshop WAVE 2002, pages 183 191, Okayama, Japan. Dorka U.E., Quval, J.C., Nguyen, V. T. and Maoult, A. L. (2006), Real-time sub-structure testing on distributed shaking tables in CEA Saclay, 4th World Conference on Structural Control and Monitoring, San Diego, USA. Dorka U.E., Quval, J.C., Nguyen, V. T. and Maoult, A. L. (2007), Substructure testing on distributed shaking tables, 2nd International Conference on Advances in Experimental

Structural Engineering, December 4-6, 2007, Shanghai, China. Dorka U.E. and Nguyen, V. T. (2011), Advanced Substructure Algorithm With Digital Feedback And Its Applications, The 2011 International Conference on Earthquakes and Structures (ICEAS11), Seoul, Korea, September 18-23. Nakashima, M., Ishii, K. and Ando, K. (1990), Integration techniques for substructure pseudodynamic test, Proc. 4th National Conf. Earthquake Eng., Vol. 2, 515-524. Nguyen, V.T. and Dorka, U.E. (2007), Unbalanced force compensation in substructure testing based on online system identification, Proceedings of the second International Conference on Advances in Experimental Structural Engineering, Shanghai, China. Nguyen, V.T. and Dorka, U.E. (2008), Phase lag compensation in substructure testing based on online system identification, Proceedings of the 14th World Conference on Earthquake Engineering, October 12-17, 2008, Beijing, China. Nguyen, V.T. and Dorka, U.E. (2009), Adaptive compensations based on online system identification for real-time substructure testing, Proceedings of the 3AESE conference, San Francisco, USA. Nguyen, V.T. Dorka, U.E. Khanlou, N. M. and Phan T. V. (2011), Real-time substructure tests of non-linear tuned mass damper using shaking table, presented in the 2nd EFAST Workshop and 4th International Conference on Advances in Experimental Structural Engineering (4AESE), Ispra 29-30 June, 2011, Italy. Pegon, P. and Pinto, A.V. (2000), Pseudo-dynamic testing with substructuring applications at the ELSA Laboratory. Earthquake Engineering and Structural Dynamics, Wiley, Vol. 29(7), 905925. Roik, K. and Dorka, U. E. (1989), Fast online earthquake simulation of friction damped systems, SFB151 Report No. 15, Ruhr-University Bochum, Germany. Shing P.B. Vannan, M.T. and Cater, E. (1991), Implicit time integration for pseudodynamic tests, Earthquake Engineering and Structural Dynamics, Wiley, Vol. 20(6), 551-576. Stoten, D. P. and Gomez, E.G. (2001), Adaptive control of shaking tables using the minimal control synthesis algorithm, Phil. Trans. R. Soc. Lond. , A 359, 1697-1723. Thewalt, C.R. and Mahin, S.A. (1987), Hybrid solution techniques for generalized pseudodynamic testing, Report No. UCB/EERC-87/09, University of California Berkeley. Wallace, M. I., Wagg, D. J. and Neild, S. A. (2005), An adaptive polynomial based forward prediction algorithm for multi-actuator real-time dynamic substructuring, Proc. R. Soc. A., 461, 38073826. Zienkiewicz, O.C. (1977), The Finite Element Method, McGraw-Hill, Book Company.

You might also like