Integrated Intelligent Fault Diagnosis And Its Application

Based on two fault diagnosis instrument systems for rolling bearing and car, several diagnosis strategies using integrated intelligence have been brought forward. Applied methods that fault feature is extracted from high frequency resonance signal have also been presented. Many experiment results have shown the effectiveness of these methods.In chapter 1, firstly, the meaning of the research is introduced briefly, and the development, key problems, and application prospects on the fault diagnosis technology are stated in detail. Secondly, research progress, existing problems, and the development trend on feature extraction and intelligent diagnosis are summarized. Finally, structure of the dissertation and the main fruits of the research are presented.In chapter 2, energy separation algorithm (ESA) is discussed in detail. AMFM model of bearing vibration signal is constructed after analyzing the vibration mechanism. Feasibility of envelope analysis in high frequency resonance signal with the ESA is demonstrated. The applied of demodulation with ESA is advanced with which diagnosis precision has been improved, and real time diagnosis and accuracy ratio are satisfied on bearing production line.Automatic classification model using integration of Adaptive wavelet transform network and Self-Organizing Feature Map Network is introduced. The strong suits of wavelet and neural network are made use of in this model and bearing fault classification is realized automatically.The integrated method of time and frequency domain analysis and RBFNN is given for bearing fault diagnosis. Through selecting representative parameters, feature space vector is constructed so the input dimension is reduced and fault classification information retained. This method provides a reference for intellectualized bearing fault diagnosis.Fault diagnosis methods with fuzzy inference, rules inference, case-based inference and combination of fuzzy inference and rules inference are analyzed briefly. As these methods exist shortages, advanced diagnosis strategy is presented, which combines the fuzzy neural network instead of fuzzy inference to rules inference. The problem of accuracy ratio decline because of roughness of fuzzy synthesizing algorithm has been solved. An example of car diagnosis is given and validity of this method has been proved.A novel instrument system for bearing fault diagnosis, which extends the functions of present bearing vibration testers, is introduced in Index F1. This practical instrument is advanced in bearing industry in our country.