Research On Detent Force Optimization Of Permanent Magnet Synchronous Linear Motor Based On Deep Neural Network

In recent years,with the continuous development of high-end manufacturing,the tubular permanent magnet synchronous linear motor(TPMSLM)has become increasingly widely used in high-end CNC machine tools due to it’s advantages such as high thrust,high precision,good dynamic performance,and compact structure.However,due to it’s structural characteristics,TPMSLM may generate detent force,which seriously affects the stability of motor operation and reduces the machining quality.Therefore,its of great significance to optimize the structural parameters of TPMSLM and reduce it’s detent force.A structural parameter optimization method using adaptive deep neural network combined with improved immune genetic algorithm is proposed in this paper,whose main steps include two aspects: first,an adaptive deep neural network is used to establish a fast calculation model from structural parameters to detent force;second,the improved immune genetic algorithm is used to obtain the optimal solution of the fast calculation model and obtain the optimal structural parameters.The specific research carried out includes:(1)Establishing a finite element model of TPMSLM,sensitivity analysis of key parameters,using a central combination design method to find the sensitivity of key parameters to detent force,selecting sensitive parameters,and optimizing them in subsequent steps;then establishing a horizontal factor table based on the structural parameter range of TPMSLM model,using ANSYS Maxwell software parameterized scanning technology to obtain sample data of sensitive structure parameters and detent force.(2)In order to address the limitations of parsing calculation methods,finite element analysis,and traditional machine learning algorithms in TPMSLM modeling,this paper introduces an adaptive deep neural network for constructing a detent force regression model for TPMSLM.The adaptive process utilizes a neural network architecture search algorithm based on genetic algorithm.The model is then compared with the K-nearest neighbors algorithm and the random forest algorithm to validate its superiority,providing a fast and reliable computational tool for subsequent detent force optimization.(3)Encoding genes and aggregating fitness analysis for immune genetic algorithm,and using elite retention strategy,adaptive similarity threshold,and local exploration optimization strategy;then using the improved immune genetic algorithm to obtain the global optimal solution of detent force calculation model,comparing it with traditional immune genetic algorithm,and verifying the superiority of the improved algorithm.A prototype experimental platform was constructed to measure the detent force of the prototype.A comparison was made between the detent force data of the original prototype and the optimized prototype.The standard deviation of the TPMSLM detent force before optimization was approximately 20.68,whereas after optimization,it reduced to approximately 10.88,resulting in a decrease of about 47.39% in the fluctuation of the motor’s detent force.The research findings indicate a significant reduction in the fluctuation of the motor’s detent force after optimization,demonstrating the applicability and effectiveness of the proposed parameter optimization method combining adaptive deep neural networks and improved immune genetic algorithm in TPMSLM detent force optimization research.