基於數控銑削過程的能耗快速建模

Abstract

機械加工過程中會消耗大量能源，進而對環境造成影響。為了降低加工能耗，工具機的能源模型被用來確定最低能耗的切削參數。此外，由於工廠中有許多型號的工具機，需要一個快速建模方法以減少停機時間來對每一台工具機進行建模。本研究提出了一種基於數控程式碼快速建立數控銑削能耗模型的方法。這種方法透過兩種關鍵技術顯著減少了收集訓練和驗證數據所需的時間：（1）磁性電壓感測器，可直接安裝在工具機電箱的總電源開關上，無需對內部電力系統進行修改；（2）基於田口穩健設計，大幅減少了收集訓練數據所需的切割實驗數量。功率消耗測量分為兩部分：（1）在空切情況下，進行不同主軸速度和進給速度的實驗，通過主軸馬達和進給軸馬達的功率特性，大幅減少空切實驗的次數；（2）在不同主軸速度、進給速度、切削深度和切削寬度下進行切削實驗。基於主軸和進給軸能耗的特性以及田口穩健設計，大幅減少了切削參數的組合。整個獲取能耗模型所需的時間大約為30分鐘，其中估計安裝電氣感測器需要15分鐘。該方法應用於10種不同型號的工具機，能耗模型的誤差率約為3.98%，凸顯了此方法的有效性，從這10台機床獲得的能耗數據可供開源使用，促進了數控銑削加工能耗的協作研究。此外，除了機台需要快速建模之外，使用不同刀具也會造成不一樣的切削功率，本研究建立可以預測不同材料的切削功率模型，以硬度為特徵去預測同一把刀具切削不同材料的切削功耗。

This study presents a methodology for rapidly establishing an energy consumption model based on NC code for CNC milling operations. The approach significantly reduces the time needed to gather training and validation data through two key techniques: (1) magnetic voltage sensors, facilitating direct installation onto electrical joints within the electrical box without the need for modifications to the machine tool electricity, and (2) a greatly reduced amount of cutting experiments required for collecting the training data based on a Taguchi method. The power consumption measurements are divided into two parts: (1) experiments conducted at varying spindle speeds and feed rates without cutting, and (2) cutting experiments conducted at varying spindle speeds, feed rates, depth of cut, and width of cut. The combinations of cutting parameters are significantly reduced based on the nature of the dependency of the parameters to the energy consumption, and using the Taguchi method. The overall time required for obtaining the power consumption model is around 30 minutes, including 15 minutes estimated for electrical sensors installation. The methodology is applied to 10 distinct machine tools, demonstrating the versatility and applicability of the modeling process across diverse machining environments. The achieved error rate of up to 3.98% underscores the efficacy of this approach. In addition to the need for rapid modeling of the machines, the use of different tools also results in varying cutting power. This study establishes a model to predict cutting power for different materials, using hardness as a characteristic to predict the cutting power consumption of the same tool when cutting different materials.