Method For Identifying Dynamic Parameters Of Machine Tool Feed System Based On Reinforcement Learning

Virtual machine tool is a machine tool simulation model established by computer simulation technology.This technology can help users quickly develop and test related machining strategies on the simulation platform,and improve the machining strategies according to the simulation results,reducing manufacturing costs and production time.The servo feed system contains the drive chain from the motor to the worktable,which is one of the important components of CNC machine tools.Therefore,the establishment of an accurate servo feed system model is the premise of building a virtual machine tool.The servo feed system model is generally a complex nonlinear system,and some parameters cannot be obtained by simple calculation and measurement,so modeling the servo feed system and identifying some of the parameters is an important research direction.This article takes the single feed axis of the CNC machine tool servo feed system as the research object.Through the dynamic analysis of the servo feed system,the relationship equation between the motor angular displacement and the table displacement is derived.Based on this,the simulation software is used to establish the machine tool servo The dynamic model of the mechanical part of the feed system.The sensitivity analysis method was used to determine the degree of influence of the parameters to be identified in the dynamic model on the simulation results of the dynamic model.An experimental method for step-by-step identification of parameters was designed.A parameter identification algorithm based on single-step reinforcement learning and a parameter identification algorithm based on multi-step reinforcement learning are proposed,and a distributed computing framework is used to identify the parameters to be identified in the dynamic model.Comparing the parameter identification algorithm based on reinforcement learning with the traditional intelligent optimization algorithm genetic algorithm and particle swarm optimization algorithm,the results show that the identification result based on the parameter identification algorithm based on reinforcement learning is more accurate and the algorithm is more stable.Through the parameter identification of the dynamic model,the simulation accuracy of the model is further improved,and the accuracy of the parameter identification result is verified.