Application And Research Of Mixed Expert System For Fault Diagnosis On Power Device Of Ship

The power device of ship is vital to ship's running .It is necessary and important to diagnose the fault about the power device of ship. There is many equipments in the power device of ship and they often go wrong. In order to assure ship's running, the manager and worker have to spend a great deal of time to maintain it, In addition, only those experts who specialize in diagnostics can dispose of some fault. Therefore the diagnostics and maintenance is very burdensome.Expert System is applied widely for the fault diagnosis. Expert System has many merits.It has the ability of heuristic illation ,and can explain for illation and append new knowledge in the knowledge database.But it also has obvious shortcomings, Such as , poor ablitity in KA(Knowledge Achieve), inefficient and incomprehensive.The artificial neural network has the ablitity of parallel processing,associative memory,distributed storage of knowledge and high robust etc.It also has perfect characteristics of self-organizing ,self-adaptive ,self-learning .It specializes in visualize ideation but is short of logic ideation. The mixed expert system combines artifical neural network with expert system organically.lt can improve work efficency.In this paper,the base principle and structure of KBS.(Knowledge Base System )is explained first ,and then discuss BP neuralwork ,RBF network on structure and atithmetic.in detail.The Mixed Diagnostic Expert System bases on Microsoft Windows, Microsoft VC++ 6.0 is choosed as program development tool, Microsoft SQL Server2000 is choosed as database development tool. It is composed of four parts .1. The Data Manage ModuleThe module includes the manage of measured parameter,such as,data appended ,deleted and combined query.2. KBS (Knowledge Base System) ModuleThe module includes the manage of produced rule ,for example the rule added ,deleted,logic checked and corresponding reason added ,deleted, edited.3. Neural Network ModuleThe module includes the arithmetic of sample data's self-learning, such as,RBFN(Redial Basis Function Network) arithmetic.4. Inferential ModuleThe module include kinds of KBS positive inferential arithmetic, Comparing with the Expert System basing on produced rule, the Mixed Expert System combines produced rule with RBF arithmetic. Therefore it can't only deal with uncertain information, but also can deal with certain information.