Ultra-high-cycle Fatigue Life Prediction Of Metallic Mate Rials Based On Machine Learning

Along with the pursuit of modern machinery and equipment on high parameters,long life,intelligent,how to design and manufacturing point of view,to ensure the long-cycle safety operation of key components,is a major equipment independent research and development and safety in service facing the bottleneck problem.The current rapid development of society requires the service life of mechanical equipment is longer than before,that is,its fatigue life under cyclic load needs to exceed ten million cycles or even higher.Ultra-high cycle fatigue usually refers to the fatigue behavior of the material when the fatigue cycle reaches 10~7 and above.Accurate prediction of ultra-high cycle fatigue life of metallic materials plays an important role in ensuring safe,reliable and long-life use of equipment.At present,researchers at home and abroad have proposed a number of ultra-high cycle fatigue life prediction models,but most of these models are based on fatigue damage accumulation theory and damage mechanism,and there are many model parameters,large errors and low efficiency and other deficiencies.Therefore,in order to improve the accuracy and efficiency of ultra-high circumferential fatigue life prediction of metal materials,this topic introduces GB,RF and SVM algorithms,and tries to propose a fatigue life prediction method based on machine learning.The main research content is as follows with the following aspects.The paper first analyzes the effects of three factors on the fatigue life of metals,namely,the properties of the material itself,the condition of the component and the service environment,and then selects the yield strength,tensile strength,modulus of elasticity,density,elongation at break and geometry of the material as the input parameters of the model by compiling and summarizing the current research status of physically driven and data-driven fatigue life prediction models at home and abroad.Data were collected from published literature in China and abroad and subjected to outlier rejection,normalization and batch normalization.The model was designed in terms of activation function,weight initialization,learning rate,number of nodes and avoidance of overfitting,and the GB algorithm,RF algorithm and SVM algorithm were selected to construct the model.The prediction performance of the model is evaluated using mean square error and network model decision coefficient,and their advantages and application scenarios are discussed.Ultrasonic fatigue experiments are conducted on GCr15 bearing steel to obtain life data,and the prediction results are compared with those of the GB algorithm model and the RF algorithm model to investigate the generalization ability of the model.The prediction results show that the model constructed based on the GB algorithm has a high prediction accuracy with only one value exceeding the triple error band in the prediction of fatigue life of GCr15 bearing steel.In order to improve the accuracy and efficiency of fatigue life prediction of metallic materials,a new ultra-high cycle fatigue life prediction model based on machine learning is proposed in this paper.This model has the advantages of simple input parameters and fast response speed compared with traditional methods,which is of great significance and application prospect for predicting the fatigue life of metal materials and ensuring the long-cycle safety service of critical components.