Verification of Hybrid Simulation by Ali. Ozdagli, Wang Xi, Ge Ou, Bo Li, Guoshan Xu Shirley Dyke, Jian Zhang and Bin Wu Project funded by National Science Foundation - CMMI Grant #1011534 National Science Foundation of China – Project #90715036 Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu Presentation Outline Introduction Background and Motivation Experimental Setup Modeling of the System RTHS Comparison HS Efforts Conclusion 2 Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu Introduction Need for Testing Global performance of new systems Nonlinear response Options Shake-table: Scaled Structural Testing Hybrid Simulation (HS) 3 Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu Background “Comparison of Real-Time Hybrid Testing with Shake Table Tests for an MR Damper Controlled Structure” by Lin et al. (2009) “The results show a close correlation between the shake table tests and the real-time hybrid simulation.” “There is clearly a difference between the hybrid tests and shake table tests.” 4 Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu Background “Development of a Versatile Hybrid Testing System for Seismic Experimentation” by Shao et al. (2012) 5 Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu Motivation How do we know? RTHS and Numerical Simulations represent the real structural behavior? Gain acceptance in community Compare the RTHS to the real structure responses Numerical Simulation RTHS Shake Table 6 Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu Challenges Accurate modeling of the target structure System Identification Semi-active controllable nonlinear damper Hard to model rate dependent dynamics Damper-structure interaction 7 Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu Objective Verification of RTHS methodology using shake table tests on mid-scale structure Research Program Phase 1: Numerical Modeling and Simulation Phase 2: Shake Table Tests Phase 3: RTHS Testing 8 Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu Phase 1: Numerical Simulation Test Structure Base Dimension: 1.84 m by 2.04 m Story height: 1.2 m Material: Structural Steel 9 Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu MCK Update Method 1 Floor - mm 0 st -50 -100 0 5 10 0 -10 -20 15 0 20 60 40 80 100 120 100 120 2 More detailsFloor were given in 20 ‘Modeling of Distributed Real-time Hybrid Simulation’ 0 accessible from http://nees.org/resources/6641/ 0 -20 Model is awarded by NEES as the-40 best simulation model. nd -50 -100 0 5 10 15 0 Floor 3 0 -100 0 5 MCK Update Method 10 Frequency - Hz rd -50 Experimental Data 20 40 60 80 40 3 Floor - mm Magnitude - dB 10 2 Magnitude - dB 20 Floor - mm Magnitude - dB Floor 1 15 Experiment Simulation 20 0 -20 -40 0 20 40 60 Time - s 80 100 120 10 Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu MR damper numerical model alpha_a alpha_b c0_a c0_b k0 gamma A x0 k1 c1_a c1_b 15.65 57.16 1.00 9.76 11.08 23.44 155.32 0.0 0.009 19.15 139.96 Numerical model for volt = 1.7 V, freq = 2.9Hz 300 Damper force (lbf) 200 100 0 -100 Test data Simulink model ODE model -200 -300 -0.2 -0.15 -0.1 -0.05 0 0.05 Damper disp (inch) 0.1 0.15 0.2 Numerical model for volt = 1.7 V, freq = 2.9Hz 300 Lord MR damper RD-1005-03 Damper force (lbf) 200 100 0 -100 -200 -300 -4 Test data Simulink model ODE model -3 -2 -1 0 1 Damper vel (inch/s) 2 3 4 5 11 Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu Phase 2: Shake Table Tests Location: Harbin Institute of Technology Size: 3m×4m (shaking direction) Peak acceleration: ±1.33g Peak velocity: ±600 mm/s Stroke: ±125 mm Maximum payload: 12t Force capacity: 200kN Maximum overturning moment: 30 t-m Frequency bandwidth: 0 - 30 Hz Conducted uncontrolled, passive off, passive on and semi-active control cases 12 Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu Comparison – Shake Table vs Simulation 5000 5 0 -5 -10 64 66 68 70 72 74 76 78 80 Pure Simulation Experimental Data 2 Pure Simulation Experimental Data Acc 3rd floor-mm/s Disp 3rd floor-mm 10 0 -5000 64 82 66 68 70 72 74 Time-s Time-s 78 80 82 2 5000 Acc 3rd floor-mm/s Disp 3rd floor-mm 10 76 5 0 -5 -10 69 70 71 72 Time-s 73 74 0 -5000 69 70 71 72 73 74 Time-s 13 Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu Phase 3: RTHS MTS loading Frame @ HIT MTS Loading Frame, 2500kN, Internal LVDT Load cell, 15kN MTS Flex GT Controller Inner Loop Control Lord MR damper, 2kN Clamp for vertical loading Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu RTHS Setup Force Complete Structure Numerical substruc. Physical substruc. Damper Desired Displacement 4 Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu RTHS Result: Kobe displacement (mm) displacement (mm) 5 10 5 0 0 -5 Due to the limitation of Pump Velocity Limitation, Piston maximum moving speed 20 30 40 50 60 20 30 40 50 60 50mm/s time (sec) -10 -5 -15 00 10 10 70 70 80 80 time (sec) displacement (mm) displacement (mm) 5 10 Reference Reference Measured Measured 5 0 0 -5 -10 -5 -15 6 6 8 8 10 10 12 12 time (sec) time (sec) 14 14 16 16 18 18 Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu displacement (mm) displacement (mm) RTHS Result: Morgan 10 5 5 00 -5 -5 displacement (mm) displacement (mm) -10 0 10 10 20 20 30 30 40 40 time time(sec) (sec) 5050 6060 70 70 80 80 10 5 Reference Reference Measured 5 Measured 00 -5 -5 -10 6 88 10 10 12 12 time time(sec) (sec) 1414 16 16 18 18 Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu Phase 3: RTHS (Replace pics with IISL Actuator) Shore Western loading Frame @ IISL 2 kip Actuator Loading Frame High performance programmable DSP system plus high precision servohydraulic motion control system. Servo Valve Lord MR Damper Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu Compensation Performance 2 0.8 Desired Displacement Measured Displacement 1 Gain (abs) 0.6 -1 0.2 -2 0 5 10 15 20 25 Frequency (Hz) 30 35 40 0 5 10 15 20 25 Frequency (Hz) 30 35 40 0 -0.2 50 Phase (deg) 1st Floor Displacement - cm 0.4 0 -0.4 -0.6 -50 NRMS: 3.62% -0.8 66 68 70 72 Time - s 74 76 0 78 19 Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu Comparison – ST vs RTHS 5000 Experimental Data RTHS 5 Acc 3rd floor-mm/s2 Disp 3rd floor-mm 10 0 -5 -10 64 66 68 70 72 74 Time-s 76 78 80 66 68 70 72 74 Time-s 76 78 80 82 5000 Acc 3rd floor-mm/s2 Disp 3rd floor-mm 0 -5000 64 82 10 5 0 -5 -10 69 Experimental Data RTHS 70 71 72 Time-s 73 74 0 -5000 69 70 71 72 Time-s 73 20 74 Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu Remarks on RTHS To verify the RTHS methodology, shake table responses at HIT are compared to RTHS results at IISL. A new control oriented model updating method is implemented using mode shapes to derive MCK. MCK model based on fully identified results Accurate zero tracking A new compensation scheme, RIAC is implemented. High performance even in large noise/signal ratio condition Flexible to choose loop shaping function Experimental tuning is easy to perform 21 Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu Model updating with UKF m2 c2 kf 2 m1 c1 kf1 Physical BRB Numerical BRB r kd (1 )kz n 1 n z d d z z d z k , , , n, r, d Constrained Kalman filter -22- Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu Exact UKF CUKF Initial 100 0 -100 100 0 -100 -200 -200 -3 Exact UKF CUKF Initial 200 Restoring force (kN) Restoring force (kN) 200 -2 -1 0 1 Displacement (mm) Physical BRB 2 -2 -1 0 1 2 3 Displacement (mm) Numerical BRB -23- Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu Real-time hybrid test validations =30 Physical BRB =30 Numerical BRB Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu 200 200 Restoring force (kN) 300 100 0 -100 -200 -300 -4 Physical BRB Numerical BRB -3 -2 -1 0 1 2 3 Initial 300 CUKF 100 0 -100 -200 -300 -4 4 Displacement (mm) Restoring force (kN) Restoring force (kN) 300 Physical BRB Numerical BRB -3 -2 -1 0 1 2 3 4 Displacement (mm) UKF 200 100 0 -100 -200 -300 -4 Physical BRB Numerical BRB -3 -2 -1 0 1 Displacement (mm) 2 3 4 -25- Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu Finite element based sectional constitutive model • Section Yield Function M fy N h Aw hw tw hw h fy (1 2 )h h Af b t (2 1)2 ( N FaN ) 2 M FaM (s, Fa ) , 2 4 1 Ny MP N FaN (4 1) M FaM (s, Fa ) , Ny 2(2 1) Mp N 1 N y 2 1 N 1 N y 2 1 •Section Restoring Force Model (RFM) s k se e When (s, Fa ) 1, Section- elastic When (s, Fa ) 1 , Section- plastic s k se (e e p ) 1 T ( ) k se e where e H s s p Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu Numerical example • Identification results 2050 Yield Axial Force [kN] 2000 1950 1900 True Value Indtified Value 1850 1800 1750 1700 0 5 10 15 20 25 30 215 210 205 True Value Indentified Value 200 195 190 185 0 35 0.031 5 10 Time [sec] 25 30 35 0.028 0.026 0.025 0.024 10 0 -10 -20 -30 2 3 Time(s) 4 5 6 5 10 15 20 3000 25 30 Back internal force Internal force path Correlation curve 2000 Before updating 100 50 0 -50 After updating 1000 0 After Updating and Kinematic -1000 -100 -150 1 0 Time [sec] Axial Force[kN] 20 True Value Indentified Value 0.027 150 Bending Moment(kN.m) Horizontal Disp(mm) 20 0.029 200 TR-Response WO-Update WI-Update 30 -40 0 15 0.03 Time [sec] • Model updating results 40 Kinematic hardening coefficient [1] Plastic Bending Moment [kN.m] 220 -200 -0.03 -0.02 -0.01 0 0.01 Section Curvature(1/m) TR-Resp WO-Update WI-Update 0.02 0.03 -2000 -3000 -300 -200 -100 0 100 200 300 Bending Moment[kN.m] Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu Experiment plan • Test scheme • Test setup and three cases of HS Traditional HS (Linear/Nonlinear) FE Model updating by HS Distributed Hybrid Simulation Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu Delay over-prediction Delay compensation based on over-prediction ①Calculate di+1 ②Predict with c ③Load with prediction ④Find force measurement Delay Compensation: Compensated Delay > System Delay Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu Implicit algorithms for RTHS Limitations: 1. Iteration 2. Intensive computation 3. Time delay Fixed Number of Iterations with Interpolation (Shing et al) Equivalent Force Control Method (Wu et al) Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu New implicit algorithm based on over-prediction (1) Over-prediction dp (2) Interpolation (4) Process 2 FEQ ,i 1 + Iterative calculation dc Actuator (3) KN K PD r d d ik1 Optimal force Process 1 ui (ai , vi , d i ) ui1(ai1,vi1, di1) Measured Measured Force Disp RN (dik1 ) K PD dik1 REk ,i 1 ++ + 1、Modified Newton’s Method applied results in good iteration performance. 2、System delay is compensated for based on over-prediction method. -31- Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu Single time step validation -5 5 k3/2 m3 m2 0.5 1.5 0.5 1 Shing方法 0.5 k3/2 0 2 4 6 迭代步 8 10 0 2 ke 10 seconds X: 426 Y: 0.001369 Shing方法 6 计算步 8 0 10 2 0.5 4 6 迭代步 8 10 -4 x 10 x 10 X: 105 Y: 0.0001493 X: 465 Y: 0.002562 2 1 1.5 EFCM 1 1 0.5 0.5 新方法 0 0 500 积分步 Shing EFCM 500 积分步 新方法 1000 11 0.005 10 -0.005 -0.015 0 5 10 15 时间 (s) 20 25 30 100 9 8 7 -0.01 500 积分步 x 10 0.01 0 0 0 -3 参考解 位移 (m) 位移 (m) 0 0 1000 0.015 Delay comptn error 2.5 耗时(s) k1/2 k2/2 4 -3 x 10 耗时(s) m1 新方法 1 2 1 x 10 EFCM -3 k2/2 1.5 耗时(s) k4/2 x 10 2.5 k5/2 耗时(s) k4/2 m4 3 1.5 耗时(s) k5/2 -5 -5 x 10 耗时(s) Test Shing EFCM 新方法 参考解 -32- 6 27.4 27.5 27.6 时间 (s) Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu Trans-pacific test between UCB and HIT 2 MX CX R ( X) MBX g Conducted in HIT, CHINA Physical model of structural resistance Performed in UCB, USA PLATEFORM: OpenFresco Express M 0.04 K 1.8 C = K =0.02 C 1.0m LEGEND Analytical model of structural energy dissipation and inertia M K Y Ug 1 X 33 Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu HIT,CHINA UCB, USA Data available @ http://peer.berkeley.edu/~aschell/DHS%20with%20HIT/ Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu Acknowledgements National Science Foundation - CMMI Grant #1011534 National Science Foundation of China – Project #90715036 HIT Lab Steve Mahin & Andreas Schellenberg @ UCB Tao Wang @ IEM Project data will be available @ NEES.org #1076 35 Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu China-US collaborative project on hybrid simulation Bin Wu, Professor Harbin Institute of Technology Yurong Guo, Professor Hunan University Shirley Dyke, Professor Purdue University Tao Wang , Assoc Professor Institute of Eng. Mechanics Jian Zhang , Associate Professor UCLA Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu THANK YOU! 37 Purdue University, West Lafayette, IN 47907 •Phone: (765) 494-7434 •Fax: (765) 494-0539 •E-mail: iisl@ecn.purdue.edu