Research On Fault Diagnosis Expert System Of Gun Control Device Based On Bayesian Network

With the advancement of science and technology,modern warfare has put forward new requirements for the informatization and intelligence of weapons and equipment.As the center of the tank weapon system,the gun control box directly determines the performance of the weapon system.Due to the complex working environment of the tank and the high intensity of work,the failure of the gun control box is unavoidable.Therefore,the development of a fault diagnosis system for the gun control box is an inevitable requirement for military modernization.The gun control box is a complex electrical control device.Because its internal components are interrelated and coupled,different fault symptoms result in different fault modes and the same fault mode may also be caused by multiple fault mechanisms.This uncertainty between failures makes it difficult to locate faults in the gun cabinet.And the gun control box will cause sudden failure of parts or subsystems because of environmental changes or human factors.This will reduce the control accuracy of the gun control box and even cause the gun control box to stop working.Therefore,it is necessary to accurately determine the fault location and replace the faulty component in time when the fault turns out.This article uses the Bayesian network(BN)constructs fault diagnosis method to solve the problems faced by the fault diagnosis of the gun control box.The researches content is as follows:First,this paper analyzes the failure mode and failure mechanism of the gun control box,and finds that the difficulty in the fault diagnosis system of the gun control box mainly lies in the uncertainty of the needle during fault diagnosis.By analyzing and comparing several existing uncertainty theories,it is proposed that use Bayesian network as a fault diagnosis model to solve the uncertainty of fault inference,then use the combination of evidence theory combined with analytic hierarchy process to solve the uncertainty of expert's prior knowledge.Secondly,the basic theory of Bayesian network and the modeling method of Bayesian network are introduced.Because of the feature of the gun control box such as lack of fault samples and so on,a Bayesian network modeling method for gun control box based on fault tree model is put forward.This method builds a fault tree model based on the fault mechanism of the gun control box,and then transforms the fault tree into a Bayesian network.It solves the problems of slow reasoning,difficult to express polymorphic logic,and inability to express uncertainty of the gun control box.Third,in order to quantify the fuzzy knowledge expressed by multiple experts as failure probability,this paper proposes to use Dempster-Shafer evidence theory(D-S evidence theory)to solve the problem of uncertainty caused by differences in knowledge between experts and limitations of experts themselves.Then,according to Analytic Hierarchy Process(AHP),thesis constructs a comparison matrix and a knowledge matrix quantify expert knowledge,and gives a specific calculation method.Finally,according to the above-mentioned modeling method constructs the fault diagnosis system of the gun control box,describes the process of fault diagnosis and construct the fault tree model and Bayesian network of the gun control box.Then,using the ±15V power signal anomaly as an example,the expert knowledge was quantified using the DS/AHP method,and the Bayesian network was used for diagnosis reasoning.