Researches On Uncertain Reasoning Of Multiple Hierarchical Fault Diagnosis Based On FMAGDM

A high performance fault diagnosis strategy for complicated equipments is of vital importance to ensure the equipments working in a more reliable and efficient condition. In fault diagnosis field, a widely used inference method is the rule-based reasoning method based on the fault tree. But, the complicated equipment often has complex function relations and intriguing breakdown nesting between the equipment components and the modules, it is very difficult to achieve the fault tree based on components framework. Hence, a multiple hierarchical hybrid diagnosis method based on fault categories is proposed in this dissertation.For the retrieving conflict resolution in rule-based reasoning, a general method is by sequentially weighted of the antecedents. The drawback of this method lies in that it has no sufficiency consideration to the synthesis influences of uncertainty factors in reasoning processing. The priority factor of rules was introduced in this dissertation. In this way, the search match sequence can be determined according to the priority factor of each rule. Meanwhile, the more important thing is to establish a method of making choice of priority factor.Since the antecedent and the meaning of the rule itself are uncertain to a certain extent, a method to deal with the uncertainty in reasoning process should be used, and such reasoning is also an inexact reasoning.In this dissertation, the inference strategy of the fault diagnosis for complicated equipments is studied. And, a method to make choice of priority factor and the uncertainty in reasoning processing that are the key factors to improve the diagnosis performance is presented.The contributions of this dissertation are summarized as follows:1. Study on fault diagnosis theory for complicated equipments, a multiple hierarchical uncertainty diagnosis method based on fault categories is proposed. The reasoning strategy is based on fault feature and the relations of all faults that can be gained without considering the function relations and location relations of equipment components. Therefore, it is calculated for complicated equipment fault diagnosis. Its basic idea is to disassemble the reasoning flow into multiple hierarchies according to fault feature and fault category that can be gained, each hierarchy contains a number of fault categories. Each fault category in different hierarchy contains a set of diagnosis rules, and the search sequences are decided by the priority factors of these rule.2. Study on the method of quantifying fuzzy linguistic values to use fuzzy sets. The opinions assigned by experts also are fuzzy linguistic values such as high confidence, lower fault probability, and so on. In order to quantize these values better, an algorithm of fuzzy numbers is adopted. The fuzzy number is one of the widely used fuzzy sets. As general fuzzy number, it is difficult to operate and acquire in applications. But it is easy to especial fuzzy numbers such as triangular fuzzy, trapezoidal fuzzy number. The trapezoidal fuzzy number is selected for the demonstration in this dissertation.3. Propose a method of determining diagnostic priority of each rule in a specified search level based on fuzzy multi-attribute group decision making. A simple method is using only the confidence of rule, but which is not the sole factors obviously. In order to gain higher efficiency, in general, such as the confidence of rule, the weights and confidences of antecedents, and the fault probability ought to be considered all. So, we propose the idea that the priority factor is determined synthetically by the confidence of rule, the weights and confidences of antecedents, and the fault probability etc. In addition, different experts may have different subjective opinions about the importance of each criterion, the concrete stages used to determine priority factor by fuzzy multi-attribute group decision making (FMAGDM) are studied.4. According to the inference strategy above, developed an efficient intelligent instrument fault diagnosis expert system. The knowledge base of the system is built in SQL Server database, and the expert system is developed in VS.Net. The experimental results of theoretical studies support the methodologies proposed in this dissertation.