Product Quality And Reliability Analysis And Improvement Through Designed Experiments

Both product quality and reliability characteristics are important considerations for all manufacturers in product and process design.Increasingly reliability is one of the most important dimensions of product quality as it can be described as quality over time.In many cases,reliability experiments do not use a completely randomized design given the constraint of time and cost.Currently analysis and improvement of product reliability with random effects is a hot research point in quality engineering.This dissertation focuses on the modeling of accelerated life test,reliability improvement through designed experiments,and simultaneous optimization of product quality and reliability characteristics.Firstly,to correctly incorporate both random effects and different failure mechanisms into the analysis of accelerated life test,this dissertation proposes a Weibull nonlinear mixed model with both random effects and nonconstant shape parameters.An illustrative example and a simulation study reveal that the proposed model calculates the experimental error and observational error terms properly,and it performs better than other models in percentile estimation.Secondly,this dissertation proposes a percentile re-parameterization model to help reliability engineers obtain a better lower percentile estimation of accelerated life tests.A log transformation is made with the Weibull distribution to a smallest extreme value distribution.The scale parameter is assumed to be nonconstant.The simulation results show that the re-parameterization model outperforms the traditional model in the estimation of lower percentiles,in terms of Relative Bias and Relative Root Mean Squared Error.Thirdly,this dissertation considers the improvement of product reliability with designed experiments when the test is actually not completely randomized.The smallest extreme value distribution is used to model the log-lifetime.Random effects and experimental factors are incorporated into the model through mean time to failure(MTTF).Product reliability can be improved through maximizing the MTTF.The illustrative example and simulation study show that the parameter estimation and significant factors identification of the random effects model is obviously better than the traditional model.Fourthly,this dissertation proposes new methods for the experiments involving both a quality characteristic and a reliability characteristic(lifetime)within a subsampling protocol.The quality characteristic analysis is based on the mean of the response.The reliability characteristic analysis is based on the two-stage method.The method then uses a desirability function approach to obtain a trade-off between the quality and reliability characteristics.These methods correctly model the experimental error and provide engineers an easy to implement approach for the simultaneous optimization of quality and reliability characteristics.Finally,for Type I censored lifetime,the two-stage method cannot incorporate the information from test stands with no failures.This dissertation provides the linear mixed model for quality characteristic and the nonlinear mixed model for Type I censored lifetime to incorporate the random effects into the analysis.Desirability function approach to obtain a trade-off between the quality and reliability characteristics is employed.The mixed models in this dissertation can incorporate the information from all censored test stands and random effects.It provides practitioners a more accurate approach to optimize both the quality and reliability characteristics with constraint randomization.The dissertation is of great significant for promoting research on reliability analysis and improvement theory.The dissertation achienements provides engineers with new methods to analyze accelerated life tests and improve product quality and reliability,which has important applicable value.