Implementation On Real-Time Hybrid Simulation By GPU Computing

Real-time hybrid simulation(RTHS)is a testing method that combines numerical simulations with physical experiments.In RTHS,the numerical simulation should be completed using a small-time step,and the minimum time step depends on the time cost of the simulation.Therefore,how to improve the speed of real-time simulation of numerical substructures is a key issue that limits the development of large-scale RTHS.To overcome this limitation,a graphics processing unit(GPU)was used to solve the numerical model,which was more suitable for large-matrix calculations due to the large number of computing units.The specific research work in this dissertation includes the following aspects.1.Research on GPU-based dynamic analysis.GPU-based numerical integration algorithms were programmed with three programming languages,Matlab,Python,and C++.The performance of GPU numerical computation and the accuracy were verified by the comparation between GPU and CPU solution.2.A shaking table RTHS system based on Matlab and GPU was proposed,which includes GPU accelerated computation by the GPU toolbox in Matlab software and the connection between substructures.The comparison of simulation and testing results demonstrated that the testing framework can implement RTHS testing successfully.3.A Python-GPU based RTHS framework was proposed to improve the communication efficiency between numerical substructure and physical substructure.The feasibility of applying Python-GPU to RTHS was verified by simulation and experiment.The computational speed of applying GPU to solve numerical substructure was further improved compared with Matlab-GPU.4.Using GPU to solve numerical substructure with different damping models and integration algorithms,the feasibility and performance of GPU application to each damping model and integration algorithm were verified.The simulation results and the time costs were compared,which were verified that GPU is suitable for different damping models and numerical integration algorithms,and the acceleration performance of GPU for different damping models and integration algorithms were given.