Research On Intelligent Fault Diagnosis Methods For Reciprocating Compressor Based On Local Wave Time-Frequency Spectrum

It is self-evident that fault diagnosis is necessary for mechanical equipments, especially for reciprocating machines that play an important role in industries. Similar to the main process of ordinary equipments, fault diagnosis can be divided into three parts: achieving the message for diagnosis, extracting the diagnostic feature and assuring the fault node & classifying the fault. Above this basic idea, the research is carried out based on the "Research on Local Wave Method and its Engineering Application" (supported by Chinese National Nature Science foundation, Project No.50475155) and former correlative research, in which the reciprocating machine is set as the target. The research includes improving certain algorithm of Local Wave Method, the combination of Local Wave Method and fuzzy duality tree theory, Local Wave Method and system-level fault diagnosis model, and Local Wave Method and SCGM predicting model, while the methods are testified in real application. The main work of the dissertation is list as follows:1. Aiming at the defects of Local Wave Method and the main application of Bidimensional Local Wave Method in image analysis, the related research is processed. Based on the basic theory of Local Wave Method, the existing defects is analyzed firstly, including the influence imposed on strong signal by weak, the application method of Local Wave time-frequency spectrum, and etc. Three main types of classification used in attracting fault feature from Local Wave time-frequency spectrum are contrasted. The image diagnosis method is presented based on bidimensional Local Wave Method. The grey image of time-frequency spectrum is set as the study target, from which the character correlative to fault is attracted. The method is summarized that the analyzing method involving with Local Wave time-frequency spectrum is effective in identifying the fault from the vibration signals obtained from the reciprocating compressor surface, while the fault degree is confirmed. Therefore, the method is perfect for fault diagnosis of reciprocating compressor because of its utility in analyzing the special vibration character.2. The mechanism of the reciprocating is analyzed. On discussing the main fault source in the construction, the main vibration styles and the main character of the vibration signal are studied. Then the topology of the sampling position distribution is optimized. Finally, the importance of the containing character of the signal in on-line controlling and according fault diagnosis is analyzed.3. The diagnosis method based on fuzzy duality tree and the local wave time-frequency spectrum is presented. The spectrum is achieved from the locale vibration signal applying Local Wave Method, on which the state-feature table of the system is established. Then the feature group with the maximum of fault information is worked out, the fuzzy tree is established sequently. Applying the fuzzy duality tree, the component of possible fault which contributes the maximal information increment in transform paths is picked out. In this way, the fault source is protruded visually with the fuzzy tree.4. The special system-level diagnosis method aiming at reciprocating machine is presented based on Local Wave Method. Confirming the fault of the reciprocating machine is difficult because of the complexity of the construction and the variety of the fault feature. In the dissertation, the system-level diagnosis is proposed. Firstly, the basic theory and algorithm of the system-level is discussed; secondly, the feasibility is studied, and lastly, with the special improved PMC model applied, the fault diagnosis method is presented and testified in project.5. In view of the special character of the reciprocating machine, the compatible grey prediction model is proposed, with which, the system state is predicted in the field of non-linear time serial. The feasibility is studied firstly. Setting the energy spectrum of the vibration signal as the input, the compatible prediction model of SCGM is chosen in system prediction model establishing. The method provides a new orientation in system state prediction which is also testified in elementary prediction on the reciprocating machine The result illustrates that the presented model is effective in the state fluctuate prediction of reciprocating machine with great precision.