Research On Dynamic Reliability Assessment And Prediction System Based On Field Data

Reliability assessment and prediction is often achieved by analyzing the life data statistically. For a long time, the life data is obtained by (accelerated) life test at laboratory. However, as the development of science and technology, this kind life data has been more and more difficult to obtain. Therefore, it's an urgent and significant issue to excavate and utilize all valuable data and information instead of life test for reliability assessment. On the other hand, most domestic enterprises have not noticed the value of field failure information of products at present, which causes that this kind information lies idle, while the field life data extracted from field failures can reflect the actual and practical reliability of product and their amount is frequently large. For this reason, it's a new orientation and breakthrough of reliability assessment and prediction to research an applicable solution of collection of product field failure information and analysis of field data.To solve that problem, this paper first studied the analysis methods based on large or small samples and the reliability prediction theory. The study of analysis methods based on large sample involved maximum likelihood estimation, least square method, Fisher information matrix method, likelihood ratio method and their applications. As the main inference methods based on small sample, Bayes method was studied seriously and the collection, category and convertion theory of prior information was emphasized, while the numerical calculation method of posterior distribution and prior information infusion theory were involved. According to the engineering practice, single sample prediction was researched in reliability prediction theory. The simple life prediction method and Bayes method were applied to realize that function. On the basis of theory studies, Ddynamic Reliability Assessment and Prediction Studio was researched and developed. In this software, database was designed to store the output and failure information of product, and field life data could be generated according to this information and the current analysis date. The main inference and prediction algorithms were realized through computer programming, which is convenient for the field life data analysis.As verified by simulations, the software has friendly user interface and is easy to operate. It has nice information management function and fast running speed and its analysis is accurate. Using this software, enterprises can record the output information of product comfortably, store the field failure information in time and implement reliability analysis for product dynamicly. The successful development of this software solves the shortage of life data in reliability assessment and prediction work and the problems of reliability information management to some degree, which will provide a new tool for reliability work in enterprise.