| 摘要: | 工業4.0旨在推動智慧化與自動化等概念，並導入物聯網、人工智慧、大數據等現代相關資訊技術，為此台灣政府也積極對於傳統產業進行輔導與轉型工作，現今台灣傳統產業人才斷層嚴重，關於精密零件產品品值以往皆由資深師傅判斷，且判斷依據不一，新一代技術人員無法得知加工刀具之磨耗狀況，導致產出產品精度不合格之不良品；因此為了提升傳產數位能量，經由數據導入智慧化與自動化是不可或缺的關鍵。 本研究將以個案公司-歐權科技為例，改善其成品良率不佳、製造現場回饋能力不足等問題，藉由個案公司刀具視覺檢測儀進行數據收集與判斷刀具磨耗程度，因此本研究將利用個案公司所提供之刀具視覺檢測儀來量測刀具磨耗與使用情形，並記錄刀具於加工機加工完後之刀長、刀徑等數據，而為了能夠預測刀具的使用壽命、刀具健康度，因此本研究特地鎖定於同一種加工料件材質－「轉塔」上使用三種加工刀具進行加工時的各種量測數據進行分析，即可透過數位化方式跳脫以往由加工機操作員以目視、觸摸方式進行判斷換刀依據。 本研究使用長短期記憶神經網路(Long Short-Term Memory ,LSTM)作為迴歸分析(Regression Analysis)模型來預測未來的刀長、刀徑變化，將歷史數據輸入至神經網路模型後，學習出刀長、刀徑隨著使用時間而磨耗的變化曲線，對接下來的刀長、刀徑進行預測，當預測長度低於設立的門檻值時，便可知道此把刀具將會於下一次使用中到達使用壽命。
Industry 4.0 aims to promote concepts such as intelligence and automation and introduce modern related information technologies such as the Internet of Things, artificial intelligence, and big data. For this reason, the Taiwanese government is also actively counseling and transforming traditional industries. Nowadays, there is a severe gap in talents in conventional industries in Taiwan. Regarding the product value of precision parts, the product value has been judged by senior masters in the past, and the judgment basis is different. The new generation of technicians cannot know the wear condition of the processing tools, resulting in the production of defective products with unqualified product accuracy; therefore, to transfer the production and improve the number Energy, intelligence and automation through data import is the indispensable key. This study will take the case company-AutoCam Technology as an example to assist the case company in improving the problem of poorly finished product yield and insufficient feedback capabilities at the manufacturing site. The case company’s tool visual inspection instrument collects data and determines the degree of tool wear. Therefore, This research will use the tool visual inspection instrument provided by the case company to measure the tool wear and usage, and record the tool length and diameter after the tool is processed on the processing machine, in order to be able to predict the tool life, tool Health, therefore, this research is specifically focused on the same processing material material-the analysis of various measurement data when using three processing tools on the turret can be digitalized to escape the previous visual inspection by the processing machine operator, Touch the way to judge the basis of tool change. This study uses LSTM as a regression analysis model to predict future tool length and diameter changes. After inputting historical data into the neural network model, we learn the change curve of tool length and tool diameter wear over time., Predict the next tool length and tool diameter. When the predicted length is lower than the established threshold, you can know that this tool will reach its service life in the next use. |
