Research On 6-DOF Spindle-coupled Road Simulator And Its Control Strategies

Fatigue durability testing is crucial in evaluating the reliability of full vehicle and parts in the vehicle industry.The servo-hydraulic road simulators are the critical equipments for the indoor durability test.The road simulator can replicate in-service conditions of the vehicle with high fidelity indoor,which not only improves the testing quality,reduces the cost but also shortens the product development cycle.Based on the project “Spindle-coupled Road Simulator” that Harbin Institute of Technology(HIT)developed for China Automotive Technology and Research Center(CATARC),the structure optimization of the hydraulically driven 6-RSS parallel mechanism and the multi-axle control strategy are researched.The topology and working principle of the spindle-coupled road simulator are first analyzed and the complete kinematics and dynamics model are established,and the speed and force Jacobian matrix between the hinge point and the end effector are obtained.The Jacobian matrix is normalized using a scaling matrix to eliminate the inhomogeneity limitation,and the performance indices evaluating force isotropy,force transmission,velocity transmission and kinematic coupling based on the normalized Jacobian matrix are developed.The influence of key parameters of simulator on four indices are illustrated.This paper proposes an improved root optimization algorithm(ROA)successfully applied to the simulator optimization design,which improves the dynamic performance of the simulator.The ROA combines the advantages of evolutionary algorithm and swarm intelligence algorithm,and the levy flight curve is inserted to further improve the global exploration ability.Finally,a physical prototype of the spindle-coupled road simulator was developed based on the optimized size parameters.The position and attitude feedback in the DOF position control of the spindle-coupled road simulator requires real-time calculation of the forward kinematics solution.Instead of the relatively slow solution of the Newton-Raphson method,an extended Kalman filter based method is proposed to improve efficiency.Regarding the problem that the dynamic coupling caused by the non-coincidence of the center of mass of the concentrated moving mass and the control point during the simulator design reduces the position control accuracy,the inverted decoupling network is applied to solve it.The recursive extended least square(RELS)method is used to identify the elements in the decoupling network,and the zero-magnitude error technology control(ZMETC)is used to obtain the steady-state inverse,which is compensated by a finite impulse response(FIR)filter to enhance the decoupling capability.By analyzing the characteristics of static and dynamic loading of traditional position-based impedance control,a composite control combining generalized PI impedance control(reducing the steady-state error),the feed-forward controller(improving the dynamic performance),the feed-forward interference force compensator and interference observer(suppressing the interference force caused by the uncertainty of the kinematic coupling of the specimen itself and the time-varying system parameters when the test piece moves in a large range)is proposed to tackle its deficiency.Limited by the bandwidth of the force/position hybrid control system and the strong coupling nonlinearity,it is difficult for the simulator to achieve accurate reproduction of the target road spectrum,thereby iterative control is further applied.The slow convergence speed and large number of iterations of the traditional offline iterative learning control(OILC)tend to cause pre-damage of the specimen and deteriorate the fatigue life evaluation,three acceleration methods are compared based on the mechanism decelerating the iteration.To speed up the convergence of OILC,the complex domain optimization theory is studied to update the impedance matrix,such as the conjugate gradient algorithm(CGILC)or the Quasi-Newton algorithm(QNILC)over the complex space.An auxiliary learning loop is inserted into the iteration process to attain an optimal gain to ensure the monotonic convergence.Furthermore,an estimation based iterative learning control(EBILC)based on QNILC is proposed,which estimates the optimal gain through the identified model,thus omitting the learning loop and greatly reducing the number of iterations.Finally,an independent suspension loading test system based on the physical prototype of spindle-coupled road simulator was built.Using this system as an experimental platform,the proposed position control,compound force control and iterative control strategies have been experimentally studied.The experimental results have verified the correctness and feasibility of the proposed control strategy.