Fault Diagnosis Strategy For Complex Systems Based On Multisource Heterogeneous Information Under Epistemic Uncertainty

With the development and innovation of modern industrial technology,the use of large-scale complex equipment is also increasing.At the same time,in order to meet the requirements of high reliability and high efficiency,redundancy technology is often used in complex systems,which not only improves system performance,but also raises some new challenges in fault diagnosis.Aiming at the fault characteristics of complex systems,a new dynamic diagnosis strategy is developed based on multi-source heterogeneous information,which can consider epistemic uncertainty,diagnosis decision-making algorithm and sensor information fusion.The proposed strategy can locate the faults quickly,reduce the diagnostic cost and offer a new idea for fault diagnosis in complex system under epistemic uncertainty.First of all,aiming at the fault characteristics of complex systems,a dynamic fault tree is used to model the complex systems,and interval numbers are used to describe the failure rate of components in these systems.A dynamic fault tree is mapped into a dynamic evidential network to calculate reliability parameters such as diagnostic importance factor and Birnbaum importance measure,which can deal with common cause failure and propose a solution for the dynamic fault tree with interval distribution parameters of components.In addition,language sets,intuitionistic fuzzy sets,expert evaluation and D-S evidence theory are used to obtain the test cost of components expressed with intuitionistic fuzzy numbers.A multi-source heterogeneous diagnostic decision table is built based on the diagnosis importance factor,Birnbaum importance measure and test cost.Secondly,for the problem of fault diagnosis algorithm,this paper mainly solves the following three key problems: normalization of multi-source heterogeneous data,determination of attribute weight and decision-making algorithm.A new normalization method is proposed to deal with the problem of different dimensions among heterogeneous data.The attribute weight is obtained by using the fuzzy analytic hierarchy process and the entropy weight method to determine the weight of each attribute more accurately.At the same time,the fault diagnosis sequence is obtained base on the improved VIKOR algorithm,and an illustrative example is given to prove the practicability and effectiveness of the proposed method in this paper.Finally,a sensor model oriented to dynamic evidence network is proposed to fuse sensor information in order to update fault information dynamically,which can optimize fault diagnosis process and improve diagnosis efficiency.The diagnostic importance factor of components is used as a measure to determine the optimal sensors placement when the number and type of sensors are certain.On this basis,the credibility of the sensor information is considered.The system reliability parameters are dynamically updated by fusing the current sensor information and current diagnostic result,and the updated optimal diagnosis sequence is obtained through the fault diagnosis decision-making algorithm.A hydraulic height adjustment system of a car dumper is applied to the proposed method.When this system fails,the optimal fault diagnosis sequence is obtained by using the proposed method,which can improve the efficiency of fault diagnosis.