Construction of Processing Prediction Model and Parameter Optimization Model for Ultrasonic Assisted Turning

建構超音波輔助車削加工預測模式與參數最佳化模式

楊慶彬(Ching-Been Yang) ； 江新祿(Chyn-Shu Deng) ； 鄧琴書(Hsiu-Lu Chiang)

《中國機械工程學刊》 33卷6期 (2012/12) Pp. 497-505

https://doi.org/10.29979/JCSME.201212.0004

摘要

超音波輔助車削在光學元件或難切削材料可以有效降低切削力、提升加工品質與減少刀具磨耗。由田口分析結果，超音波輔助車削加工之最佳參數水準組合為A1B1C3D2。而影響表面粗糙度品質的因子中，轉速是最顯著的，其次為進給率，超音波型式與切削深度影響則不明顯。本文提出「階段式田口類神經網路模式」，結合田口法及類神經網路建構超音波輔助車削加工之預測模式。最後結合超音波輔助車削加工預測模式及Levenberg - Marquardt method搜尋法則，建立了超音波輔助車削加工參數最佳化模式。由超音波輔助車削加工參數最佳化模式搜尋出最佳參數組合為進給率=0.1625（0.0015 mm/rev）、切削深度=0.1625（0.0015 mm）、轉速=0.843（475 rpm）及超音波型式=0.5（超音波），預測之表面粗糙度Rapre為1.71 μm。而驗證實驗值為1.69 μm，預測值與實驗值的誤差為1.2%。

並列摘要

Ultrasonic assisted turning is a technique capable of reducing cutting force and tool wear while increasing manufacturing quality in optical components or materials that are difficult to cutting. Rotational speed was the most significant factor influencing the quality of surface roughness, feed rate the second, whether with ultrasonic type or not and depth of cutting revealed marginal effects. This study combines Taguchi method with artificial neural networks (Taguchi-neural network) to construct a predictive model for ultrasonic assisted turning. Then, the Levenberg-Marquardt method was incorporated into the predictive model to establish parameters to provide an optimal model for ultrasonic assisted turning. The optimal parameters derived using the resulting model include feed rate 0.1625 (0.0015 mm/rev), depth of cutting 0.1625 (0.0015 mm), rotational speed 0.843 (475 rpm), and with ultrasonic assisted. Surface roughness Rapre predicted was 1.71 μm, while the tested one was 1.69 μm. Thus, the error of which between prediction and experiment was 1.2%.

並列關鍵字

Ultrasonic assisted turning ； Taguchi method ； Surface roughness ； artificial neural network ； Levenberg-Marquardt method