Combining Back-propagation Neural Network with Bayesian Network to Adjust the Best-fitted Quadrant Protrusion Control Parameters in CNC Machine Tools

• Main Authors: Li-Wei Lee, 李立維
• Other Authors: 葉賜旭
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/t5d6f8

碩士 === 國立臺北科技大學 === 機械工程系機電整合碩士班 === 106 === The efficient adjustment of correction parameters in CNC machine tools on machining paths under different environmental conditions and accurate prediction of results in machining paths have always attracted considerable research interest. Compared with the trial and error method, a Taguchi experiment can regularly find appropriate parameters. However, this experimental process is cumbersome and time-consuming. In the uncertain parameters model, a back-propagation neural network is generally used to solve the nonlinear model. Here, this study confirms the number of hidden layers and the number of neurons in each back-propagation neural network algorithm. Finally, this study selects the Bayesian algorithm that best matches with the back-propagation neural network to estimate the appropriate correction parameters under different environmental factors, and thus confirm the robustness of the network model. Payload, feed-rate, and circular radius are used here as environmental factors. Furthermore, a circular machining path is used in the process of adjusting the correction parameters. The optimization condition is the quadrant protrusion, which occurs during the process of circular motions. The relevant correction parameters are backlash correction parameters (backlash acceleration, backlash acceleration effective time, backlash acceleration stop distance). This study investigates whether the backlash correction parameters estimated using the Bayesian algorithm in combination with the back-propagation neural network can meet the requirements. Finally, our experimental results show that this method can indeed be used to estimate a range of correction parameters in CNC machine tools.