| The purpose of lifetime data analysis is to quantitatively grasp the life characters of systems or components,as well as feedback the information obtained to the process of design, manufacture or repair and maintenance work,with a view to improving reliability.Much interest has been concentrated on life expectancy data analysis for engineering,medicine and the biological sciences.The statistical method of single life time test data is already mature. However,the mixed exponentially distributions frequently occur.From the viewpoint of mathematical statistics,lifetime data of the components may be considered as coming from different population;from the actual angle,engineers sum up these products expiration as the different expiration mechanism.As long ago as 1952,Davis indicated that most of the failures of systems were caused by personnel and equipments,and different cause corresponds to different distributions with different parameters according to corresponding frequency. There should be a turn in consistence with the realistic logic that using mixed distribution to describe component failure.Therefore,it is necessary to develop complex system.Based on the significance of the exponential distribution in life time data analysis this paper does an extensive research for mixture distribution,which concept can be summarized as follows:1.In order to expand mixture distributions,this paper proposes a multivariate mixed exponential distribution model.After studying the accelerated life test and currently available mixture distribution,this paper expands two-parameter exponential distribution into hybrid multiplex distribution.At the same time,MLE of the model is built.2.Aiming at the problem of mixture model,optimized technique was proposed to apply in the MLE of the parameters.We proposed a genetic algorithm simulated annealing algorithm to solve the problem of MLE.Example of calculation is presented to show its validity and efficiency.
|
PDF Full Text Request