臺灣為螺絲螺帽緊固件產業的第三大出口國，但臺灣緊固件產業的品質檢驗大多仍舊依賴人工抽檢。在工業4.0的趨勢下，螺絲緊固件產業逐漸導入現代化的設備做為產品的檢測，如光學篩選設備；但混料、重複計次和品質不良篩選未篩出等問題依舊無法完全解決。近幾年國內開始有廠商將解決方案由產品產出後篩選，提早至產品製程中檢測，例如引進使用光學尺及主滑台安裝壓電感測器的製程檢測機等。希望在製程中即時反應異常鍛造，保護模具亦提升製程品質。但目前上述檢測機制無法感測到微小的模具損壞所產生的磨損，主因為模具磨損屬於工業製程中的不規則因素，無法通過單一成品量測來測量，要判斷模具磨耗通常需要長時間的觀察記錄，故目前的市售檢測機制無法有效檢測出模具的損耗進而預防保養。
本研究採用壓電感測器收集大量冷鍛製程資訊，探討沖棒模具於鍛造過程所承受壓力與沖棒模具之生命週期的關聯，以實現在線檢測沖棒模具故障的預警系統。本研究最後提出了一種演算法，乃透過本研究提出之模型提取冷鍛製程中電壓訊號的10個特徵值並結合單一決策樹。且透過多種不同實驗方式比較驗證，包括類神經網路(ANN-6, ANN-66)，支持向量機(SVM)等分類方法。最後的比較結果，本研究提出之演算法均優於其他方法，在超過400萬組數據中其分類辨視精確度達93%以上。

The study is to create an on-line tool to assist in the detection of the wear and tear failure process of the punch. Four groups of sensed data, were collected from the complete cold forging process, which is from a new punch just installed, and working to wear failure, in the actual production process of the industrial six-mode cold forging machine. There were a total of more than four million data items collected. We propose an algorithm, based on the single decision tree for constantly evaluating the values of 10 parameters to predict the failure of the punch, which is empirically recognized as the appearance of the overvoltage zone. These parameters are specifically tailored to the characteristics of the punching process. The proposed algorithm is compared with other machine learning methods including neural networks and support vector machine learning method in classifying the punching process in the normal zone and overvoltage zone. The proposed algorithm is the best and its accuracy rate is above 93%.

1. 精湛光學科技股份有限公司, (1992), https://www.ccm3s.com/ last retrieved August 30, 2019.
2. 峻欣企業股份有限公司, (2000), https://www.junesin.com.tw/ last retrieved August 30, 2019.
3. Mączka, T., Żabiński, T. (2012), "Platform for Intelligent manufacturing systems with elements of knowledge discovery", Manufacturing System, pp. 183–204.
4. Żabiński, T., Mączka, T., Kluska, J., Kusy, M., Hajduｋ, Z., Prucnal, S., (2014), "Failures prediction in the cold forging process using machine learning methods.", Artificial Intelligence and Soft Computing, pp. 622-633.
5. Żabiński, T., Mączka, T., Kluska, J., Kusy, M., Gierlak , P., Hanus, R., Prucnal, S., Sęp, J., (2015) , "CNC milling tool head imbalance prediction using computational intelligence methods", Artificial Intelligence and Soft Computing, pp. 503-514.
6. Żabiński, T., Mączka, T., Kluska, J., (2017), "Industrial platform for rapid prototyping of intelligent diagnostic systems", Trends in Advanced Intelligent Control, Optimization and Automation, pp. 712-721.
7. Gierlak, P., Szuster, M., (2016), "The manipulator tool fault diagnostics based on vibration analysis in the frequency domain", Applied Mechanics & Materials, Vol. 817, p234-244.
8. Breiman, L., Friedman, J.H., Olshen, R.A., Stone, C.J., (1984), Classification and regression trees, https://doi.org/10.1201/9781315139470/ last retrieved August 30, 2019.
9. Bejou, D., Wray, B., Ingram ,T.N., (1996), "Determinants of relationship quality: an artificial neural network analysis", Journal of Business Research, Vol.36, p137-143.
10. Sch¨olkopf, B., Smola, A.J., Williamson, R.C., Bartlett, P.L., (2000), " New support vector algorithms", Neural Computation, Vol.12, p1207–1245