Study On Expert System Of Fault Diagnosis For Engine Valve Abnormal Sound Based On Wavelet Analysis

As we all know, the fault diagnosis expert system is a program that combined the advanced fault diagnosis methods and the modern computer technologies.Because of its good exhibition,such as fluent running,faster diagnosis and powerful working,it is used widely in many fields.Today ,we can find lots of successful expert systems are applied to help people diagnose certain mechanical failures in the engineering of machine.In this article,the author attempt to design a fault diagnosis expert system and use it to diagnose the engine valve abnormal sound faults.The signal of engine valve abnormal sound,which based on acoustics, could be described as follow: y(t)= w(t)+s(t)+ v(t)In the above formulation,y(t) is the engine sound which include the fault signals,w(t) is the normal sound signals of the engine,s(t) is the pure abnormal sound signals of the valve and it is a mixed signals when there are several fault spots,v(t) is the noise of the background.One of the most important goals of this article is to separate the signals of the abnormal sound from the mixed ones,and then, draw a conclusion of the abnormal sound through the expert system according to the characteristic of s(t).1.Total Design of the SystemAs far as the structure of fault diagnoses expert system is concerned,the main parts of it include Data Acquisition System,Signal Processing and Analysis System, Knowledge Expression,Methods of Inference,Data Transfer and Knowledge Acquisition etc.Those contents are all discussed in the article ,and at the same time,they are also the main parts of the article.The Figure 1 is the total structure of the system that I designed. 2. Circumstance of Expert System ExploitureThe expert system of this article ,based on the windows XP, is designed through the Graphical User Interfaces(GUI) of Matlab language.Because of the flexible and powerful capability of data storage, Cell arrys are used widely in the designed system,especially inthe data storing and transfering process.PCC-818HD,AVL3058, Piezoelectricity crystal sensors and other instruments are used at the periphery of the system too.3. Data Acquisition and Signal ProcessingFor an excellent fault diagnosis expert system,it's a premise request to identify the fault's symptom correctly while it used in practice.In order to get the fault's information more accurately, an acceleration sensor has been used and through which the system could acquire vibratory data from the surface of the engine's valve cover.Because the electric charge that produced by the piezoelectricity crystal is so tiny that it is necessary to amplify and filter it before inputting it in the system,and the purpose of doing like that is to let the system recognise the signal efficiently.In this article,the signals processing are mainly depend on the Wavelet Toolbox of Matlab,for example,in order to determine the frequency domain of the abnormal sounds,we can (1)subdivide the omnibus signal using wavelet analysis(db3,five layer),and compare the details with the normal engine sound which is also subdivided by the same Wavelet,and(2)analyze the detail that composed the abnormal sound using Envelope Power Spectrum, Finally (3) separate the abnormal sound with digital filters and obtaine its character.4.Design of Expert SystemThe main parts of the fault diagnosis expert system are Knowledge Expression, Knowledge Base and Methods of Inference.In my expert system,the knowledges are expressed in production rules, that is to say, a relationship is set up between the faults-tree and the parse-tree(based on rules),which is organised with"if…then…"commands in the programe. By building the knowledge base,Matlab cell arrys are used and each row of the cell represent a fault records,in which different diagnosis parameters and different faults'name are stored in different columns.There are three methods of inference mentioned in this article(1)Forward Inference,in some times it also called Forward Chaining,which is searched or traversed from a problem to its solution and was used in the fault diagnosing process,(2) Backward Inference, it also called Backward Chaining ,a chain that is traversed from a hypothesis back to the facts that support the hypothesis and was used in the Knowledge Base building process,and(3) Multiple inference, which is a association of Forward Inference and Backward Inference and was used in the expert system self-studying process.5. Test Result AnalysisThe expert system was tested by diagnosing several valve abnormal sound of EQ61000 engine in the laboratory ,and the results showed that the signals we got from the valve cover contained the informations of the valve abnormal sound. The Data Acquisition System we designed for the system is fluent and efficient,the datas we got from it accorded with the demand that the system asked for. Wavelet analysis,which used in the fault signals processing,is proved to be a feasible method for picking up signals character of the faults, and it can be used to separate the abnormal signals too.As to the system,those are obvious that the Knowledge Expression is reasonable, the Knowledge Base is logical and the Inference is powerful.The conclusions that it figured out are correct when we use it to diagnose certain valve faults.Even though,there are many things that we can do to improve the functions of it,such as Multi-fault Diagnosis,Neural Network inference and so on .I'm of conviction that with the deeply and continuously study we devote in the feild, the system will be going to be changed more and more powerful in the future.