Equivalent Data Conversion And System Reliability Evaluation Method

In the process to estimate the reliability of large complex system, it is very common that the system is constructed by lots of components which may obey different probability distributions. Different kinds of data will be obtained because of different test methods and the components's life distribution. Because the variety of the components's data forms, it brings large troubles to estimate the system's reliability. So the reliability engineers have paid much more attention to data transformation, and wish to find high efficient formulas. Through the data tranfoimation we can translate different distribution data to uniform one, and then using the tanslated data to estimate the whole system's reliability, it will greatly reduce the difficulties.For data transformation, there existed different kinds of formuias. Based on the reliability and the reliability confidence lower bound, many data transform formulas had been built, Based on information entropy, data transform formulas can also be constructed, and based on the theory of probability distribution moments, data transform formulas had been estabilished too. Although there are lots of data transformation formulas, but how to measure the equaivalence of the different data, which transformation formulas is the best among these formulas and so on, but it is lack documents to discuss these problems. So the main tasks of this thesis is at first to define the concepts of equaivalence between data, based on these concepts, then to compare the different transformation formulas under exponential distribution, normal distribution, and Weibull distribution, some properties of the transformation formulas are also considered. For the shortcomings of the formulas, we build a new more efficient data transform one. Using the equivalent deta transform methed, and also as an appliacation of this method, we at last calculate a complex system's reliabilityThe main works of this thesis will be descripted more clearly in the following points.1) To estabilish the criterions to assess the translated data. In this thesis, based on the two critical characters,reliability and reliability confidence lower bound, we have difined some concepts such as Equivalent data in a point of confidence level, Equaivalent data in an interval confidence level,Equaivalent index, etc.2) Based on the criterions established, we compared the different kinds of dada transform formulas which have been mentioned above in exponential distribution, normal distribution and Weibull distribution, and we also researched the properties of some of the formulas, such as the existence of the solution, the robostness of solutions on the parameters, the optimium of the parameters,etc.3) For the existence of many shortcomings in the data transform formulas, such as the existence of solution depends on the selection of the parameters, the reliability point estimation is bigger than the one calculated by the original data and so on, we have built a new high effective one, and also researched some of its properties.4) The main reason why we pay so much more attention on the data's transform is that when large complex system need to estimate the reliability, we can use the formula to transform different kinds of probability distribution data to the same one, it simplify the process of estimating the system reliability, and at the same time, get a more accurate reliability lower confidence bound, and realized the purpose to estimate the complex system. As an application of this thesis, at the last chapter, we have researched the reliability lower confidence bound of a complex system which is constructed in series by a Success/Failure distribution component, an exponential distribution component, a norm distribution component and a Weibull distribution component.