Research On Straightening Stroke Of Shaft Parts Based On Machine Learning

Shaft parts are one of the most common mechanical components used in industrial production and in everyday life.Their main purpose is to carry loads and transmit torques in different mechanisms.As a rule,shaft components require different processes to meet specific hardness,strength,surface roughness and other performance requirements after machining.Especially after heat treatment,shaft parts are subjected to thermal and organisational stresses and are prone to bending resulting in certain deviations between the central axis in the actual situation and the central axis in the ideal state,which seriously affects the performance and life of shaft products,and in more serious cases can even lead to major safety accidents.Therefore,the straightening of bent and deformed shaft parts is an important part of the shaft production process and a necessary part of improving the quality of shaft parts.This paper aims to improve the productivity of the shaft straightening process and reduce the scrap rate.Based on the analysis of the shortcomings of the existing shaft straightening process,the machine learning algorithm is used to predict the straightening stroke based on the actual sampled shaft straightening process data,which effectively reduces the number of straightening times of the straightening machine in a single operation and improves the productivity of shaft straightening.The main research elements of this paper are as follows:(1)The bending deformation characteristics of existing shaft components and the working principle of straightening are studied,then the maximum deformation position of shaft components during straightening is theoretically calculated and finite element simulated using the method of fitting the load deflection of a simply supported beam,followed by the theoretical analysis and optimisation of the straightening stroke of shaft components,and then the straightening stroke correction coefficient is proposed.(2)On the basis of the working parameters of the straightening machine,a shaft part straightening model was established and Ls-Dyna software was used to simulate the straightening of shaft parts of different diameters.By analysing the results of the simulation data,the straightening stroke correction coefficients were determined and then an optimised calculation method for the straightening stroke was proposed.(3)Based on the sampled straightening process data,an optimised BP neural network was used to predict the straightening stroke of the shaft parts,and an error analysis was carried out.Two typical optimization methods,genetic algorithm and particle swarm algorithm,were selected to compare the accuracy of the straightening prediction stroke using GA-BP algorithm and PSO-BP algorithm respectively.In order to further improve the accuracy of the straightening prediction stroke,an optimised LSTM algorithm is used to predict the straightening stroke based on the BP neural network,and the shaft parts are straightened by the straightening machine to verify the feasibility of the method,thus effectively reducing the number of straightening times and improving the straightening efficiency.