Research On Engine Fault Diagnosis Using The Fusion Technology Of Sound Intensity And Neural Network

Fault diagnosis crosscuts multiple disciplines. With many recent technologicaladvancements and improvements, this technology is playing a more and more importantrole in modern manufacturing industry. Therefore, studying the theory and the applicationof intelligent engine fault diagnosis will have a significant practical impact.This paper presents my research on the development of an intelligent engine faultdiagnosis application. This application is based on the analysis of the sound intensitycharacteristics for various faults. It uses a modular neural network for recognizing thesound intensity characteristics. The research makes the following contributions.First, based on the sound intensity detection and virtual instrument technology, anautomatic sound intensity collection and analysis system for engines have been developed.Second, the sound intensity characteristics of engine faults are extracted by frequencyand time domain analysis method. A detailed study is carried out on the near field soundintensity characteristics of typical engine faults.Third, with the deep study on engine sound intensity knowledge fusion technology,information from different knowledge sources is processed and utilized together, and afault diagnostic knowledge database of engine is built with UML.Fourth, a modular neural network has been built for engine fault diagnosis. From theaspects of work, training, and study, the fusion diagnosis of diagnostic network and soundintensity knowledge are generated.Fifth, based on the frame of engine's fault diagnostic model, via the analysis ofdiagnostic model tasks' classification and mapping, an intelligent engine fault diagnosticmodel was confirmed by information flow integrated method.Sixth, through the analysis of each operation parameter in engine fault diagnosticmodel, an efficiency evaluation system on fault diagnostic model of engine is built, whichincludes efficiency, precision and reliability evaluations.Seventh, based on the engine fault diagnostic model, engine fault diagnosticapplication with the fusion technology of sound intensity and neural network is detailedlydescribed.Our application realizes engine working status non-touch collection, online diagnosisand learning. Through several experiments, it is proved that the engine fault diagnosticsystem (EFDS) can effectively identify the working status of engine it studies, find out andconfirm fault area and character and further find out the reasons that generate certain faults.The theory and technology used in the study of engine fault intelligent diagnostic systemprovide a new way to build a better intelligent EFDS.