Research On Fault Diagnosis Method For Complex Systems Based On Dynamic Evidential Network

The application of high and new technology to the engineering system has significantly improved the performance of the system and greatly increased the complexity of the system at the same time,which gives severe challenges in fault diagnosis for complex system.These challenges are shown as follows.(1)Dynamic dependency of the failure components.(2)The diversity of components distribution parameter.(3)The problem of epistemic uncertainty caused by the lack of fault samples.It will be important to establish a fault diagnosis model based on the unique fault characteristics and develop a dynamic diagnosis strategy which can locate the fault component quickly and reduce the maintenance cost.First of all,aiming at the problem of the dynamic fault characteristics and epistemic uncertainty caused by using some redundant technologies in complex systems,this dissertation proposes a method of constructing fault models for complex systems under epistemic uncertainty.The fault model is established by using a dynamic fault tree to simulate its dynamic fault characteristics;Different methods are adopted for the diversity of components distribution parameter.For the components that follow the exponential distribution,the triangular fuzzy subsets and the interval-valued triangular fuzzy subsets are used to describe the failure rate,and an estimation method of the interval-valued triangular fuzzy weighted mean is established to express the interval failure rate of components.For the components that follow the Weibull distribution,the coefficient of variation method is employed to estimate the interval failure rate of the components.It can effectively solve the problems of fuzziness,epistemic uncertainty,and the diversity of component distribution parameter,and also can avoid the subjectivity of the component failure rate estimation.Secondly,aiming at the problem in the traditional dynamic fault tree analysis method,this dissertation proposes a method based a dynamic evidence network.By mapping the dynamic fault tree graph structure and numerical into the dynamic evidence network one by one,the dynamic logical structures in the system can be expressed in the dynamic evidence network.The D-S evidence theory is introduced in this method,which effectively deals with the problem of the failure rate of the bottom events expressed in an interval value.The conditional mass distribution table of the root nodes in the dynamic evidence network and the algorithm to map a dynamic fault tree into a dynamic evidence network are also given.The dynamic evidence network graphical structure provides an easy way to specify the dependencies.In addition,reliability,the posterior probability and the importance measure can be calculated with Netica software.Finally,an illustrative example is given to illustrate the efficiency of this method.Finally,multi-source fault information,including diagnostic importance factor,Birnbaum importance and heuristic information value,is comprehensively taken into account to develop an improved VIKOR diagnosis algorithm which can obtain the best fault search solution.Furthermore,an entropy weight method is used to determine the weight of attributes when the attributes are interval numbers.An illustrative example is given to illustrate how the proposed method can be used to perform the diagnosis strategy analysis for the braking system using multi-attribute decision making with interval numbers.The simulation results show the effectiveness of the proposed method.