Research On Straightening System Of Long Tube Parts Based On Machine Learning

The long-tube precision parts that are widely used in industrial products will have a certain degree of bending after processing,especially after heat treatment.Such product defects seriously affect the accuracy of use.At present,there are few pipe straightening equipments on the market,and the existing equipment systems have a low degree of intelligence and low straightening efficiency,which cannot meet the production needs.The key point of straightening of long tube parts is the combination selection of straightening stroke,straightening pressure point and straightening support point.The traditional straightening method mainly relies on the manual straightening experience and the expert system in the straightening system to select the straightening parameters.This mode leads to the unreasonable selection of straightening parameters,which seriously affects the straightening accuracy of the pipe fittings.In view of the above problems,this topic takes the straightening parameters and straightening stroke prediction in the straightening system as research objects,and introduces machine learning methods into the research process of the above problems.A new mathematical model is established based on the material and geometric characteristics of the long pipe parts,so that the pipe straightening system can intelligently identify the part parameters,and finally predict the alignment parameters to meet the requirements of straight production.The main research work of this paper is as follows:(1)The problem of straightness and straightness of long tube parts is studied,and an evaluation method is given.The theory of three-point reverse bending straightening during straightening is researched,and a simplified model of straightening process is given.At the same time,the bending conditions of elastoplastic mechanics are introduced to make elastic mechanics assumptions.(2)The deformation and stress in the process of pipe fitting straightening were studied.Based on this,the elastoplastic load and deflection model during the straightening process were further studied,and the mathematical relationship between the load and deflection was obtained.Finally,the straightening detection system,mechanical structure system and some hardware systems of the pipe straightening system are introduced and designed.(3)In the process of identifying material parameters,based on the load and deflection models,the neural network in machine learning is introduced.Yield limit and elastic modulus values in the straightening state were obtained through corresponding simulation experiments.Through comparative analysis,the RBF neural network was selected as the material parameter identification method in the process of tube straightening.(4)The prediction model and model accuracy in the process of pipe fitting straightening were studied,and various characteristic parameters were selected for regression operation.Finally,through regression analysis and variable testing,the elastic modulus,yield limit,original curvature,pipe size,part material,and support point distance were determined as the input of the machine learning model.(5)Aiming at the straightening stroke problem,predictions were made by using the LSSVM algorithm,the standard PSO-LSSVM algorithm and the improved PSO-LSSVM algorithm.By comparing the error with the expected value,the overall average prediction error value of the improved PSO-LSSVM algorithm is 2.42%,which can meet the requirements of the pipe straightening process.