Study On Key Technologies To Centrifugal Compressor System Vibration Autonomous Diagnosis

The present dissertation focuses on the study of centrifugal compressor fault vibrationautonomous detection and diagnosis method.The main content is about the signalcharacteristic identification,abnormity detection,diagnosis rule automatic abstract,diagnosis method etc. On the basis, the centrifugal compressor fault diagnosis system wasbuilt up. The usefulness of research method and the diagnosis system was examinedpractically.The chapter 2 is about the automatic identification of signal characteristic usingmodern signal processing method. It includes the identification of the orbital's featureusing image processing and Wavelet NN and the identification of small vibration signalfeature using Hilbert-Huang Transform.The chapter 3 is about the problem that the fault model of rotation machine and thefeature of the fault is not one to one. It exists the characteristic of non-linear reflectionbetween the fault model of rotation machine and the feature of the fault. But the existingmethod of fault diagnosis is difficult to satisfy the need of the dynamic fault diagnosis forrotational machine. For this purpose, we established a classified method of rotationalmachine based on Kohonen neural network for the fault classification. It was checked bycompressor vibration signals.The fault detection approaches of centrifugal compressors have been investigated inchapter 4. As to the vibration signals of centrifugal compressors in abnormal condition arevery complex, fault detection approach based on immune mechanisms is proposed. Thekey techniques of proposed fault detection approach, such as description of equipmentcondition, matching function of detectors, algorithm to generate detectors and optimizationof detectors are lucubrated. The effectiveness of the method is demonstrated by some faultcases of the test bed.In chapter 5 ,on the problem of redundancy information in compressor intelligencediagnosis, it puts forward a knowledge reduction arithmetic based on rough sets .And it was proved that the knowledge architecture for diagnosis can be reduced greatly bythe arithmetic. In chapter 6, under the synthesis of the method mentioned before, the vibrationintelligence diagnosis system of centrifugal compressor was built up .The system hasbeen used practically. In this chapter the architecture of the diagnosis system and thediagnosis expert code have been studied. Finally, through the analysis of examples,theresult shows that the intelligent diagnosis method is suitable for centrifugal compressorvibration fault diagnosis.In this dissertation, several intelligent diagnosis method of centrifugal compressor havebeen studied. The presented methods in every chapter have been checked by simulationand real signal. At the last chapter, the effect of the diagnosis system has been checkedthrough several examples.