機械系統性能衰退預測與故障診斷之研究

A Study of Performance Degradation Predictions and Fault Diagnoses for Mechanical Systems

機械在運轉情況下會發生漸進式的故障，初期出現異常徵兆時，若未及時處理可能造成後續嚴重故障，為及早偵測故障問題，宜建立預兆式維護作業。本文分析機械系統故障種類、故障問題，建立機台振動總量估算法及性能劣化管理方法，利用迴歸分析建立性能衰退預測模式，從機台振動變化預測機台性能劣化趨勢。此外，探討故障診斷層級，建立振動訊號量測診斷平台，發展振動時域訊號特徵擷取及故障診斷方法，提供機台性能出現劣化訊息時，進行性能評估及故障識別。實驗結果顯示，指數模型和量測數據具高關聯性，適合用於預估性能衰退；透過本文建立的相關軟硬體技術，可用於發展系統預兆式維護之基礎。

Gradually induced faults may occur while machinery is running for along time. This kind of fault may cause serious damage if the potentialfaults aren’t remedied in time. To avoid a serious breakdown, prognosticmaintenance is commonly adopted to treat this kind of problem. In thispaper, the general types of faults of mechanical systems are first introduced.The predicted model of performance degradation is proposed based on thetrend of the over-all values of vibration. The model was constructed usinga regression analysis method in cooperation with measured vibration signals.The techniques of vibration-based diagnosing are also reported foridentifying the fault sources when machinery is at the degraded condition.Fault identification is done by comparing time-domain signal characteristicswith the set of fault patterns. A rotational test device is designed toimplement signal measuring, performance degradation prediction and faultdiagnosis. The experimental results show that the exponential model andthe measured data have a high correlation, indicating that they can be usedto predict the performance of a machine. Moreover, the studied results canbe used as the basis for developing prognostic maintenance for a system.