Optimization Design And Degradation Modeling For Accelerated Degradation Tests With Random Initial Degradation Values

With the improvement of industrial manufacturing and the continuous development of science and technology,the failure of products can be traced back to their potential performance degradation process.Therefore,an accelerated degradation test(ADT)method is proposed,aiming at obtaining life data and richer information,compared with the former one under conditions higher than normal stress.Some products' physical properties will change themselves in the degenerative process.Similarly,inconsistencies between samples are magnified under limited sample conditions,which start a process from an undefined initial value,so subsequent life expectancy is not accurate.To improve the model evaluation accuracy of long-life and highly reliable products with uncertain initial degenerative values.Based on the existing research on random effects,thesis conducts the following research on degradation model modeling and scheme optimization design:(1)The random effect of performance degradation is considered for the random process model,and the Wiener process degradation modeling method considering the initial degradation value is proposed.Firstly,the Wiener process with random effect is studied and analyzed.The initial degradation degree plays as a random variable assumed to be normally distributed.Secondly,not to lose the generality of the results,the acceleration model with different parameters is normalized to the standard linear form,and then the Wiener process degradation model considering the random initial value will be proven.(2)Given the problem that accuracy deviation of test scheme caused by individual differences in the optimal design of accelerated degradation test,a scheme optimization design method of Wiener process with random effect under single/multiple stresses should be proposed.Firstly,the degradation degree of different products is assumed to have a random initial degradation value.Secondly,based on the statistical data,the uniform orthogonal design theory was used to determine the multi-stress combination form and combined with the stress combination mode,the multi-stress combination acceleration model was given.Then,the optimization objective function was established from the D-optimization criterion to minimize the variance of the model parameter estimation.The three optimization algorithms each got the optimal stress level,the optimal sample number,and the test detection time.Finally,a conclusion verified the rationality of the proposed model in experimental optimization design.(3)For the degradation trail model,the random uncertainty of the initial degradation value is taken into account.Assumes that the degradation of the initial value is the normal distribution and lognormal distribution of cases,through the analysis of the degradation model for the optimal model selection criteria and under different degradation trail models,get the sum of squares for error(SSE).The degradation trajectory model with the minimum SSE value is selected as the optimal performance degradation model,which calculates pseudo failure life for reliability assessment.The calculation accuracy and evaluation effect are greatly improved.