臺灣博碩士論文加值系統

傳統的設備故障往往都是成品在加工過程中出現問題或是人員發現設備異常或聽到噪音時才被發現，如今科技日新月異的發展上，工業4.0與和人工智慧已是智慧工廠發展重點，工廠的運作過程中，設備的資訊傳遞尤其重要，透過資訊傳遞與資訊分析，可以即時回傳設備訊息和即早發現潛在的故障。因此，本論文開發設備分級系統，以抽水馬達和氣壓管的設備分級實驗為例，使用震動感測器與麥克風收取檢測設備之震動、音訊資料，將收集資料儲存於MySQL資料庫以便資料保存與切割，並用訓練於時間卷積網絡、長短期記憶等深度學習模型，以此檢測設備在使用時的狀態，透過深度學習模型辨識的結果，比較設備分級中震動與音訊檢測不同方法比較其準確性，同時在震動與音訊檢測中時間卷積網絡、長短期記憶的訓練模型比較其準確性，通過以上對比出設備分級系統在訊號與深度學習模型上的優勢。在本論文的具AI氣壓管狀態檢測之標號牌自動分類工作站中，將故障分級系統整合至的操作系統中，用Node-RED設計UI介面，以便能即時顯示設備狀態，操作人員能在第一時間了解設備的狀態並進行排查，減少設備發生更嚴重的故障、減少不良率產生、避免人員傷亡。

Traditional equipment failures are often only detected when problems arise during the production process, when personnel notice abnormalities, or when unusual noises are heard. However, with the rapid development of technology, Industry 4.0 and artificial intelligence have become key focuses in developing smart factories. In the operation of factories, the transmission of equipment information is significant. Through information transmission and analysis, it is possible to instantly relay equipment status and detect potential failures early on.
Therefore, this thesis develops an equipment grading system, taking the grading experiments of water pumps and pneumatic pipes as examples. Vibration sensors and microphones collect vibration and audio data from the equipment being tested. The collected data is stored in a MySQL database for data preservation and segmentation. Deep learning models such as TCN (Temporal Convolutional Network) and LSTM (Long Short-Term Memory) are then trained to detect the operational status of the equipment. By comparing the accuracy of different methods for detecting equipment conditions based on vibration and audio data, and by comparing the accuracy of the TCN and LSTM models in vibration and audio detection, the advantages of the equipment grading system in terms of signal detection and deep learning models are highlighted.
In the AI labeling workstation developed in this thesis, the fault grading system is integrated into the operating system. A UI interface is designed using Node-RED, allowing real-time display of equipment status. This enables operators to understand the equipment status at the earliest opportunity, conduct troubleshooting, reduce the occurrence of more severe failures, decrease defect rates, and prevent casualties.

摘要 …………………………………………………………………………………………………………………………………………i
Abstract ……………………………………………………………………………………………………………………ii
誌謝 …………………………………………………………………………………………………………………………………………iv
目錄 …………………………………………………………………………………………………………………………………………v
表目錄 …………………………………………………………………………………………………………………………………………vii
圖目錄 …………………………………………………………………………………………………………………………………………viii
符號說明 …………………………………………………………………………………………………………………………………………xi
第一章 緒論 ……………………………………………………………………………………………………………………1
1.1 研究動機 ……………………………………………………………………………………………………………………1
1.2 研究目的 ……………………………………………………………………………………………………………………1
1.3 論文架構 ……………………………………………………………………………………………………………………2
第二章 文獻探討 ……………………………………………………………………………………………………………………3
2.1 震動故障分類相關文獻 …………………………………………………………………………3
2.2 音訊故障分類相關文獻 …………………………………………………………………………5
2.3 其餘感測相關文獻 ………………………………………………………………………………………………7
第三章 系統架構 ……………………………………………………………………………………………………………………9
3.1 設備分級系統架構 ………………………………………………………………………………………………9
3.2 震動檢測系統流程 ………………………………………………………………………………………………10
3.3 音訊檢測系統流程 ………………………………………………………………………………………………11
3.4 模型介紹 ……………………………………………………………………………………………………………………12
3.4.1 LSTM(Long Short-Term Memory) ……………………………………………………13
3.4.2 TCN(Temporal Convolution Network) ………………………………14
3.5 混淆矩陣 ……………………………………………………………………………………………………………………16
3.6 YOLO ……………………………………………………………………………………………………………………17
3.6.1 YOLOv4 ……………………………………………………………………………………………………………………18
第四章 線切割機之水循環系統檢測 …………………………………………………………………………19
4.1 模擬馬達震動辨識實驗 …………………………………………………………………………19
4.1.1 模型訓練 ……………………………………………………………………………………………………………………20
4.1.2 震動檢測結果 ………………………………………………………………………………………………22
4.2 線切割機介紹 ………………………………………………………………………………………………23
4.2.1 水循環系統介紹 ………………………………………………………………………………………………23
4.3 感測器安裝方式 ………………………………………………………………………………………………24
4.4 類別介紹 ……………………………………………………………………………………………………………………25
4.4.1 震動樣本 ……………………………………………………………………………………………………………………26
4.4.2 音訊樣本 ……………………………………………………………………………………………………………………27
4.5 震動模型訓練 ………………………………………………………………………………………………28
4.6 音訊模型訓練 ………………………………………………………………………………………………32
4.7 抽水馬達之檢測結果 …………………………………………………………………………37
第五章 具AI氣壓管狀態檢測之標號牌自動分類 工作站開發 …………38
5.1 系統架構 ……………………………………………………………………………………………………………………38
5.2 動作流程 ……………………………………………………………………………………………………………………40
5.2.1 視覺分類 ……………………………………………………………………………………………………………………40
5.2.2 視覺架構 ……………………………………………………………………………………………………………………41
5.2.3 視覺流程 ……………………………………………………………………………………………………………………41
5.3 操作系統架構 ………………………………………………………………………………………………43
5.4 出料站介紹 ………………………………………………………………………………………………45
5.5 入料站介紹 ………………………………………………………………………………………………46
5.5.1 感測器安裝方式 ………………………………………………………………………………………………46
5.6 樣本介紹 ……………………………………………………………………………………………………………………47
5.6.1 取樣流程 ……………………………………………………………………………………………………………………48
5.6.2 震動取樣 ……………………………………………………………………………………………………………………48
5.6.3 震動樣本 ……………………………………………………………………………………………………………………49
5.6.4 音訊取樣 ……………………………………………………………………………………………………………………50
5.6.5 音訊樣本 ……………………………………………………………………………………………………………………51
5.7 模型訓練 ……………………………………………………………………………………………………………………54
5.7.1 震動模型訓練 ………………………………………………………………………………………………54
5.7.2 音訊模型訓練 ………………………………………………………………………………………………58
5.8 實驗結果 ……………………………………………………………………………………………………………………61
第六章 結論暨未來展望 ………………………………………………………………………………………………62
6.1 實驗成果 ……………………………………………………………………………………………………………………62
6.2 結論 ……………………………………………………………………………………………………………………63
6.3 未來展望 ……………………………………………………………………………………………………………………64
參考文獻 …………………………………………………………………………………………………………………………………………65
附錄1、抽水馬達音頻FFT圖 ………………………………………………………………………………………………68
附錄2、抽水馬達音頻FFT圖_0到100Hz …………………………………………………………………………74
Extended Abstract ………………………………………………………………………………………………79

動機文獻：
[1]Deloitte Analytics Institute. (2017). Predictive Maintenance and the Smart Factory. Retrieved from https://www2.deloitte.com/us/en/pages/operations/articles/predictive-maintenance-and-the-smart-factory.html, (July 15, 2024).
震動故障分類相關文獻：
[2]Goyal, D., Dhami, S. S., & Pabla, B. S. (2020). Non-Contact Fault Diagnosis of Bearings in Machine Learning Environment. IEEE Sensors Journal, 20(9), 4816-4823.
[3]Wang, W., Zhou, G., Ma, G., Yan, X., Zhou, P., He, Z., & Ma, T. (2023). A novel competitive temporal convolutional network for remaining useful life prediction of rolling bearings. IEEE Transactions on Instrumentation and Measurement, 72, 1-12.
[4]Zhang, Y., Ren, Z., Feng, K., Yu, K., Ma, H., & Liu, Z. (2023). Transformer-enabled cross-domain diagnostics for complex rotating machinery with multiple sensors. IEEE/ASME Transactions on Mechatronics, 28(4), 2293-2304.
[5]Huang, M., Liu, Z., & Tao, Y. (2020). Mechanical fault diagnosis and prediction in IoT based on multi-source sensing data fusion. Simulation Modelling Practice and Theory, 102, 101981.
[6]Zhang, J., Song, X., Gao, L., Shen, W., & Chen, J. (2022). L2-Norm shapelet dictionary learning-based bearing-fault diagnosis in uncertain working conditions. IEEE Sensors Journal, 22(3), 2647-2657.
[7]Zhou, A. Y., & Barati Farimani, A. (2024). FaultFormer: Pretraining Transformers for adaptable bearing fault classification. IEEE Access, 12, 70719-70728.
[8]Dhandapani, R., Mitiche, I., McMeekin, S., & Morison, G. (2022). A novel bearing faults detection method using generalized Gaussian distribution refined composite multiscale dispersion entropy. IEEE Transactions on Instrumentation and Measurement, 71, 1-12.
[9]Mei, J., Zhu, M., Liu, W., Fu, M., & Tang, Q. (2024). Conditional variational encoder classifier for open set fault classification of rotating machinery vibration signals. IEEE Transactions on Industrial Informatics, 20(3), 3038-3049.
[10]Jiang, G., Nie, S., Xie, P., Li, Y., & Li, X. (2022). Multiscale one-class classification network for machine health monitoring. IEEE Sensors Journal, 22(13), 13043-13054.
音訊故障分類相關文獻：
[11]Tang, L., Tian, H., Huang, H., Shi, S., & Ji, Q. (2023). A survey of mechanical fault diagnosis based on audio signal analysis. Measurement, 220, 113294.
[12]Kim, S.-M., & Kim, Y. S. (2024). Enhancing sound-based anomaly detection using deep denoising autoencoder. IEEE Access, 12, 84323-84332.
[13]Alsumaidaee, Y. A. M., Paw, J. K. S., Yaw, C. T., Tiong, S. K., Chen, C. P., Yusaf, T., Benedict, F., Kadirgama, K., Hong, T. C., & Abdalla, A. N. (2023). Fault detection for medium voltage switchgear using a deep learning hybrid 1D-CNN-LSTM model. IEEE Access, 11, 97574-97589.
[14]Wang, Y., Liu, C., Wu, H., Sui, Q., Yang, C., & Gui, W. (2023). Revolutionizing flotation process working condition identification based on froth audio. IEEE Transactions on Instrumentation and Measurement, 72, 1-12.
[15]Ahn, H., & Yeo, I. (2021). Deep-learning-based approach to anomaly detection techniques for large acoustic data in machine operation. Sensors, 21(16), Article 5446.
[16]Tagawa, Y., Maskeliūnas, R., & Damaševičius, R. (2021). Acoustic anomaly detection of mechanical failures in noisy real-life factory environments. Electronics, 10(19), Article 2329.
[17]Chen, J., Zhang, F., & Li, Y. (2023). Sound-based real-time scrap falling state monitoring in the shearing process of wide heavy plate mill. IEEE Sensors Journal, 23(15), 17092-17102.
其他感測器故障分類相關文獻：
[18]Peng, H., Jiang, B., Mao, Z., & Liu, S. (2023). Local enhancing transformer with temporal convolutional attention mechanism for bearings remaining useful life prediction. IEEE Transactions on Instrumentation and Measurement, 72, 1-12.
[19]Tao, Y., Jun, Z., Zhi-hao, Z. et al. Fault detection of train mechanical parts using multi-mode aggregation feature enhanced convolution neural network. Int. J. Mach. Learn. & Cyber. 13, 1781–1794.
[20]Guan, X., Gao, W., Peng, H., Shu, N., & Gao, D. W. (2022). Image-based incipient fault classification of electrical substation equipment by transfer le arning of deep convolutional neural network. IEEE Canadian Journal of Electrical and Computer Engineering, 45(1), 1-8.
[21]Yan, J., Li, Q., & Duan, S. (2024). A simplified current feature extraction and deployment method for DC series arc fault detection. IEEE Transactions on Industrial Electronics, 71(1), 625-634.
[22]Sun, J., Li, C., Zheng, Z., Wang, K., & Li, Y. (2022). A generalized, fast and robust open-circuit fault diagnosis technique for star-connected symmetrical multiphase drives. IEEE Transactions on Energy Conversion, 37(3), 1921-1933.
[23]Xu, D., Jin, F., Zhou, F., Zhao, J., & Wang, W. (2024). Active fault diagnosis based on dual-observer for leakage detection and estimation with adaptive auxiliary signal mechanism in industrial gas networks. IEEE Transactions on Instrumentation and Measurement, 73, 1-11.
[24]Dan, H., Yue, W., Xiong, W., Liu, Y., Su, M., & Sun, Y. (2022). Open-switch and current sensor fault diagnosis strategy for matrix converter-based PMSM drive system. IEEE Transactions on Transportation Electrification, 8(1), 875-885.
LSTM相關文獻：
[25]Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural Computation, 9(8), 1735–1780.
TCN相關文獻：
[26]Bai, S., Kolter, J. Z., & Koltun, V. (2018). An empirical evaluation of generic convolutional and recurrent networks for sequence modeling. ArXiv, 1803.01271.
[27]Long, J., Shelhamer, E., & Darrell, T. (2015). Fully convolutional networks for semantic segmentation. ArXiv, 1411.4038.
[28]Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016). You only look once: Unified, real-time object detection. ArXiv, 1506.02640.
[29]Bochkovskiy, A., Wang, C.-Y., & Liao, H.-Y. M. (2020). YOLOv4: Optimal speed and accuracy of object detection. ArXiv, 2004.10934.