Studies on Process Optimization Based on the Cutting Force Prediction Model of an Artificial Neural Network for Five-axis Milling

• Main Authors: PengXu, 徐鵬
• Other Authors: Rong-Shean Lee
• Format: Others
• Language: en_US
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/55647548718398851201

碩士 === 國立成功大學 === 機械工程學系 === 103 === The evolution of the five-axis machining systems and Computer Numerical Control (CNC) machine tools has provided considerable advantages for high-precision manufacturing. However, due to a conservative machining strategy,parameter value shave usually been preset as constants to avoid tool damage or breakage. Unfortunately, this practice leads to a great expense of machining time. So improving production efficiency is an important issue for machining applications such as five-axis milling, especially when machining complex surface parts. With the development of virtual simulation technology, optimizing machining parameters before machining is now recognized as a feasible method to improve efficiency. Based on this consideration, this thesis proposes a novel milling process optimization method to regulate milling constraints and adjust parameters so as to maximize the five-axis milling efficiency. As cutting force is the primary constraint, the cutting-force model is first analyzed to identify the necessary force components. The employed artificial neural network (ANN) is trained with collected milling data to predict milling force. Then,a model with all constraints, including drive conditions and force,is established to compute the optimal spindle speed and feed rate in each cutting engagement interval. With the optimized results of each milling interval, a series of process optimization algorithms are proposed to evaluate and integrate the optimal parameters in the process. All these processes are carried out in a virtual machining environment. Finally, the new milling data could be used to directly modify the cutter location (CL)file. In Addition, several case examples have been provided to verify the optimization performance of this method, which was found to be effective and reliable.