Estimation Of Storage Life With Physical Probability Methods

The Engineering technology indicates that the operating life and service time limitof various kinds of products are highly underestimated, which is often caused by theoversimple prediction technique of system reliability. The correct estimation of theresidual life of products is of great significance in practical engineering.Based on DM and DN distribution model, the thesis confirms the estimationformula of the residual life of products in accordance with the residual life of productsprobability density function of the initial moment, the residual life of productsprobability density function of τ moment, products reliability function of τ moment andthe interrelationship of mathematical expectation of continuous random variable.Making use of L Hospital rule, the thesis further simplifies the the estimation formula ofthe residual life of products that have longer operating life than average.Meanwhile, the thesis estimates the location parameter and shape parameter relatedin the estimation formula of the residual life of products by adopting the maximumlikelihood estimation and square estimation. Firstly, based on complete sample, usingmaximum likelihood estimation and square estimation, the thesis also estimates the twoparameters. Then based on fixed failure number sample and fixed time samplerespectively, the thesis estimates the two parameters through maximum likelihoodestimation.Finally, based on fatigue test experimental data of two groups of different types,the thesis calculates and settles by groups and estimates location parameter and shapeparameter by applying maximum likelihood and square estimation and acquires theresidual life estimation four groups datas at different moment and compares them withthe average value of the residual life of the test sample and the estimated residual lifedata based on reference distribution model. The result indicates that the residual lifeestimated based on DM and DN distribution model is much closer to the average valueof residual life of test sample. The accuracy prevails the estimated residual life based ontraditional distribution model and has a much higher accuracy in estimating the residuallife of products.