A Usage Life Improvement Method Based On Material Ingredient Parameters And Process Parameters Optimization

Usage life improvement is effective and necessary method to achieve the design life of a product.This improvement can be realized in product design or in produce process.Traditional usage life improvement methods are based on quantities of life tests and improving on the failure factors.This seems to be impractical for the long-life components.To solve this problem,new life improvement methods are needed.Thus,relating to degradation test data analysis,this paper forms a new life development method based on material ingredient and process parameter optimization.When designing and producing,this method suggests to monitor the material ingredient parameters and process parameters to get representative products and then analysis their usage life parameters by applying degradation test design and analysis methods.After that,a nonlinear mapping relationship model can be generated between material ingredient parameters,process parameters and usage life parameter and optimization can be achieved based on this model.The main research of this paper are as below.(1).Study the process of this method and use a case to show how this method can be conducted easily and effectively.This method should be conducted following this 4 steps: failure mechanism and crucial parameter recognition,usage life parameter prediction based on degradation test and analysis,relationship model setting between ingredient,process parameter and usage life parameter,usage life optimization based on relationship model.(2).Study the design and train process of Back Propagation neural networks(BPNN)with different output sets particularly.Those output sets can be some life related parameters like performance at a given time,parameters of degradation model,pseudo life,parameters of life distribution and reliable life.Due to the advantages and weaknesses of those BPNNs with different outputs,they should be used in proper situation.(3)Study the design and train process of multiple regression models with different output sets particularly.Similarly to BPNN,those output sets can be some life related parameters like parameters of degradation model,pseudo life,parameters of life distribution and reliable life.A case is studied to compare the fitting result of BPNN and multiple regression model.