專家系統在轉子振動診斷之研究

Abstract

本論文主要之目的係在建構一專家系統，藉由此專家系統來執行轉子機械振動診斷 之研究。專家系統的優點是能將專業人員的知識轉化成電腦程式中的法則，以便代替專 業人員做判斷。本論文中建構之專家系統包含有人機界面、資料庫、推論機及振動訊號 處理程式等五個模組。而其建構方式是使用程式語言直接建立的方式，以符合轉子機械 診斷系統的要求。另外就振動訊號的分析上，採用頻譜(spectrum)分析及全息譜 (holospectra)分析兩種方法，針對固定頻率的振幅及其他相關資料輸入推論機中，利 用診斷法則進行分析。而在診斷法則的建立上採用文獻收集、整理的方式進行，透過對 文獻的整理歸納出各種故障的特徵，將其彙整為診斷法則。最後經由模擬的振動資料驗 證此專家系統程式的可行性。希望透過此專家系統的發展使轉子機械振動診斷的專家系 統早日實現。

In this thesis, an expert system to diagnose rotating machinery vibration problems is developed to achieve the goal of diagnosis automation. The underlying expert system includes the following three modules: (1) Main architecture of the expert system. (2) vibrational signal processing. (3) Rules for diagnosis. The main architecture of the expert system consists of manmachine interface, data-base, rule-base and inference machine. The inference machine performs three inference mechanisms, i.e., the symbol inference, certainty factor inference and fuzzy inference. The second module will execute the job of extracting the features from the specific data of vibration signal. The holospectra and spectra are adopted for feature extraction in vibrational signal analysis. The third module is the essence in our expert system where the diagnosis rules should be carefully built according to the research results reported in refereed literature and knowledge from the in-field expertise in that particular domain. The final expert system is tested by using the simulation of vibration data in rotating machinery. Furthermore, it is noted that the parameters and rules for diagnosis in the expert system should be updated based on the results from a series of experiments with the hope that it will become more and more intelligent and accurate.