Abstract:The control of shaking table mostly adopts three-parameter control as the underlying basic control algorithm,and there are many three-parameter control parameters,the parameters tuning is time-consuming and laborious.A three-parameter control parameter tuning algorithm of shaker based on LSTM (long short-term memory network) was proposed.The test acceleration input and output data of the shaking table system were divided into training set,test set and verification set,and an LSTM deep network was established and trained to simulate the system model of the shaking table.For the LSTM deep network model,a new three-parameter control link was introduced,and the gradient descent method was used to perform offline tuning of the control parameters.Finally,the parameters tuning were combined with the original parameters of the control system for real machine verification.The simulation results show that the proposed tuning method can achieve better results than manual parameter adjustment,and the tuning process is completed offline through the system model,without the need for real machine operation,which has the advantages of high efficiency and good effect.