適多工況之刀具剩餘壽命智慧監測方法研究

以人工經驗進行刀具壽命費時且缺乏客觀數據支援，無法準確的預測刀具壽命會造成非必要停機、加工精度降低及刀具成本提高。本研究透過實驗設計、數據特徵分析，建立磨耗與振動的關聯，比較GRNN與LSTM預測準確度，最終選用GRNN模型，並建立等效轉換法則，最終以C#建立智慧監測人機介面，完成即時預測刀具剩餘壽命的監測系統。實驗驗證結果顯示，基於AI模型評估刀具剩餘壽命，加工後預估剩餘壽命為1435秒，仍可完成853秒加工需求；在以等效轉換模組應用於多工況驗證下，顯示預測刀具剩餘壽命為430秒，實際壽命為392秒，預測誤差為42秒，由於剩餘壽命不足以完成下次加工製程，系統自動警示換刀。驗證結果顯示，所建立的方法與系統可有效用於刀具剩餘壽命預測。

Relying on human experience to assess tool life is time-consuming and lacks objective data support, leading to inaccurate predictions of tool life, which can cause unnecessary downtime, reduced machining accuracy, and increased tool costs. Through experimental design and data feature analysis, this study establishes the correlation between wear and vibration, compares the prediction accuracy of GRNN and LSTM, ultimately selecting the GRNN model, and develops an equivalent transformation rule. Finally, an intelligent monitoring human-machine interface is developed using C#, enabling a monitoring system for real-time prediction of tool remaining life. Experimental validation results show that, based on the AI model, the estimated remaining tool life after machining is 1435 seconds, sufficient to meet a 853-second machining requirement. In multi-condition validation using the equivalent transformation module, the predicted life is 430 seconds, while the actual life is 392 seconds, with an error of 42 seconds. Since it is insufficient for the next process cycle, the system automatically issues a tool change alert. Given the accuracy of remaining life prediction, this study holds significant reference value in the field of tool cutting.