基於機器學習的馬達驅動V型皮帶的故障檢測控制器設計與實現

Design and Implementation of the AI-equipped Controller for Detecting Motor-driven V-belt Failure

丁寧(Ting, Ning)

指導教授 ： 許健平

國立清華大學/工學院/智慧生產與智能馬達電控產業碩士專班/碩士(2021年)

摘要

V型皮帶通常應用於工業驅動設備中，在所有傳動皮帶類型中最為廣泛使用。目前透過定期檢查與維護去發現V型皮帶上的故障與失效問題，因此早期發現故障成為一個重要的議題，目的在於避免停機以及降低維護成本。本篇論文研究了沖床中馬達驅動V型皮帶的失效問題。透過計算原始資料的基本統計量，對皮帶不同張力下的特性進行有效的分類與識別。利用特徵選擇方法找出馬達參數與此皮帶故障預測問題的關聯性，取得其中影響性較大的特徵並使用Isolation Forest (iForest)訓練預測模型。除此之外，也設計並實作出一個配置人工智慧的控制器，此控制器在可程式化邏輯控制器 (PLC)運行訓練好的預測模型去做V型皮帶故障的即時預測。在控制器的設計上需要考慮到PLC有限的運算效能以及模型本身的預測效能。實驗結果顯示，本篇論文使用的方法相比於其它異常檢測演算法具有更少的計算複雜度，使得其能夠在不佔用過多資源且不影響PLC原有的控制程序下進行異常檢測的功能。

關鍵字

預測性維護 ； 機器學習 ； 馬達驅動V型皮帶失效 ； 不平衡資料 ； 異常檢測 ； 人工智慧控制器 ； 可程式化邏輯控制器 (PLC)

並列摘要

V-belt is the most widely used of all transmission belt types and is usually applied in industrial drive equipment. Regular inspection and maintenance are generally used to discover the faults in the V-belt, and early detection of faults becomes necessary to avoid downtime and reduce maintenance costs. This thesis investigated the prognosis of a motor-driven V-belt failure problem in a press machine. Calculating basic statistics of raw data is used to classify and identify the characteristics under different tension of the belt. The features that have a greater impact on the diagnosis of V-belt faults are selected, and Isolation Forest (iForest) is used to train a predictive model. We also design and implement an AI-equipped controller which runs a trained model on the programmable logic controller (PLC) to perform a real-time prediction of V-belt failure. In the design of the AI-equipped PLC, we consider the trade-off between the execution time of the PLC and the performance of the anomaly detection. The experimental results show that our proposed method has less computational complexity than other anomaly detection algorithms, which allowing it to run more efficiently on the PLC. Furthermore, the AI-equipped PLC can perform anomaly detection functions without affecting the original control process and occupying excessive resources.

並列關鍵字

Predictive Maintenance ； Machine Learning ； Motor-driven V-belt Failure ； Imbalanced Data ； Anomaly Detection ； AI-equipped Controller ； Programmable Logic Controller (PLC)

參考文獻

[1] W. J. Lee, H. Wu, H. Yun, H. Kim, M. B. Jun, and J. W. Sutherland, “Predictive maintenance of machine tool systems using artificial intelligence techniques applied to machine

condition data,” Procedia Cirp, vol. 80, pp. 506–511, 2019.

[2] K. A. Kaiser and N. Z. Gebraeel, “Predictive maintenance management using sensor-based

degradation models,” IEEE Transactions on Systems, Man, and Cybernetics-Part A: Systems and Humans, vol. 39, no. 4, pp. 840–849, 2009.

[3] S. D. Coleman, Chris and E. Deuel, “Predictive maintenance and the smart factory,” tech.