- 學位論文
運用類神經網路建立CNC銑床加工精度最佳化預測模式
The Application of Neural Network Forcasting to Establish the Optimal Mode for CNC Milling Process
摘要
CNC銑床加工，在一般傳統加工產業與高科技產業都占有著舉足輕重的地位，交通、鋼鐵、建築、電子、醫療…等，說其帶動整個工業進步繁榮也不為過，現今的加工廠幾乎都有此設備，除了特殊零件，製程已不再是問題，如何用更簡便的方式、更快的速度、做出更精密的零件，值得繼續研究。 一般對銑削加工最佳化的研究著重於主軸轉速、切削深度、進給率、刀刃數這4項的配合，現今技術已經相當成熟，本研究的動機要以業界用於傳動系統元件常用之藍十字合金鋼(SNCM439)為工件材質，做出最短工時與符合精度之物件。選擇適當的主軸轉速、切削深度、進給率和刀刃數對藍十字合金鋼材進行銑削加工後，經由類神經網路預測，突破傳統技術分析與時間序列分析法的侷限性，提升預測的準確率，所得最短工時與符合之精度，以幫助加工廠預測工件最佳化較客觀且準確的參考依據。 運用類神經網路中的誤差倒傳遞演算法建構的倒傳遞類神經網路預測模型，分析出本研究類神經網路預測模型最佳參數。研究結果顯示：(1) 利用全因子參數設計可以找出輸入及輸出變數的數據資料；(2)類神經網路被選為訓練加工精度參數數據的工具； (3) 根據實驗最佳精度與最短工時排名，找出最佳精度與最短加工時間両者取最適當參數條件，進而提供金屬加工時的參數調整依據；(4) 經由類神經網路訓練，可以加速找出最佳參數組合；(5) 從研究結果顯示，本研究所建立的類神經網路預測模型預測能力優良。
並列摘要
CNC milling process is very important in both the traditional industry and high-tech industry. It is also very popular in different kinds of metal prcessings. The processing procedure is no longer an issue. However, how to use more precise and faster process that can produce sophisticated parts with better quality is worth of continuing study. The technology of applying CNC milling processing is quite mature. There are lots of studies related to CNC milling process. Most of them focused on the optimizations for spindle speed, depth of cut, feed rate, and the number of blades parameter designs. In this study, we investigated Blue Cross alloy steel (SNCM439) for the workpiece material. It is commonly used in the driveline components in the industry. The motivation of this study is to investigate the optimal parameter designs that can shorten processing time with better precision. In this study, we use parameter designs and neural network trainings to predict the optimal parameters for spindle speed, depth of cut, feed rate and the number of Blue Cross alloy steel blade for CNC milling process. From this study, we can conclude that (1) the parameter design can help the collections of input and output data, (2) the backpropagation neural network training have been investigated and implemented, (3) the performance of the prediction can help better predictions of the parameters for CNC milling process, (4) the optimal parameters were established for faster processing time and better precision, (5) the performance measure showed that neural nnework training can help making good predictions.