智能化刀具磨耗預測方法之研究

Building Machine Intelligence for Tool Wear Prediction

林聖哲(SHENG-CHE LIN)

指導教授 ： 鄭宗明

朝陽科技大學/理工學院/工業工程與管理系/碩士(2016年)

摘要

近年來，國內的老年化及少子化現象日趨嚴重，導致製造業將面臨到勞動人口缺乏之窘境，迫使台灣提出「生產力4.0」去應對，而台灣能夠立足全球工具機產業，就必須不斷強化競爭力。
CNC工具機切削加工精度之關鍵在於刀片鋒利程度，操作者判斷刀刃之鋒利程度皆以目視及手觸的經驗法則來認定。若能藉由電腦影像與人工智慧之導入取代專家經驗，在加工前準確預測某指派刀具於加工後之刃口狀態是否勝任切削工序，將可避免停機檢視及更換、重工、報廢之機率與其時間成本，提高刀具使用之經濟效益與加工品質，形成有效的機器智慧。
本研究將對刀片刃口之磨耗預測作深入探討，以期望能達到「工具機智能化」之程度。研究中將利用NC程式所定義之主動切削條件與從工具機上所擷取之主軸阻抗值分析出切削時間，與所拍攝到之刀片刃口磨耗量做關聯分析，應用類神經網路學習切削參數與刀片磨耗量間彼此之對應關係，產生一刀具磨耗模擬機制，以資訊系統大量、快速及精確之優點，依據演算法進行數據分析，即時有效的預測刀具磨耗量。

關鍵字

刀刃磨耗預測 ； 刀具壽命 ； 智能工具機 ； 類神經網路

並列摘要

The labor shortage problem in Taiwan is an emerged crisis and will become worse in the future. The government was forced to confront the problem with a “Production 4.0 policy”. Hoping that an intelligent and autonomous automation may relief the tension and provide advancement in technology, especially for the machine tool industry.
The quality of machining highly depend on the sharpness of the cutter tip. In the past, the examinations of the blades were carried out mainly by human expertise with manual or visual inspections. These operations can be replaced with numerical procedures using computer visions and artificial intelligence. If the inspection outcome were properly related to the NC program, then the tool wears may be analytically predicted according to the given NC tool path. The relation between the wear and the task will become the machine intelligence.
This research uses computer vision to capture tool wears task by task, and relate wear amounts bit by bit with the tool tip motions from an analyzed NC program and the spindle resistance using neural networks. The outcome shows that the prediction can be used to simulate tool wear before the cutting processes. Therefore, a proper tool selection or assignment, for economical and/or quality purpose, can be carried out by machine intelligence autonomously.

並列關鍵字

Tool Wear Prediction ； Cutter Tool Life ； Intelligent Machine ； Artificial Neural Networks

參考文獻

王怡惠，「從工業4.0看我國生產力4.0之挑戰」，臺灣經濟研究月刊，第38卷第8期，頁111-119，[民104]。

馬成傑，「監督式學習類神經網路於銑削斷刀即時監控之研究」，中原大學，[民99]。

Albertelli, P., Mussi, V., Ravasio, C. and Monno M. “An Experimental Investigation of the Effects of Spindle Speed Variation on Tool Wear in Turning”, Procedia CIRP 4, vol.22, pp.29-34, (2012).

Alonso, J., and Salgado, R., “Analysis of the structure of vibration signals for tool wear detection”, Mechanical Systems and Signal Processing, vol.22, pp.735-748, (2008).

Balan, G., and Epureanu, A., “The monitoring of the turning tool wear process using an artificial neural network”, Intelligent Production Machines and Systems, vol.47, pp.20-25, (2006).