Study of Parameter Optimization and Tool Life Prediction based on Hybrid-Index on Machine Learning method

• Main Authors: Po-Jung Yang, 楊柏融
• Other Authors: Shih-Ming Wang
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/5r67f7

碩士 === 中原大學 === 機械工程研究所 === 106 === This study is mainly to establish a Hybrid-index parameter optimization and tool life prediction method, which can make the CNC machine tool obtain better energy consumption while satisfying the surface precision, and predict the life of the used tool to improve the processing efficiency. And can more accurately grasp the time point when the tool needs to be replaced, to avoid the product surface accuracy exceeding the required range. The research method is to use the Google’s library Tensorflow in the Python programming language to build an artificial neural network architecture to learning. In the study, the first analysis of the spindle speed, feed rate, processing conditions, unit energy consumption and surface accuracy recorded in the past experiments, in addition to discussing the relationship between the parameters, and the data into the established neural network for training And analysis, functional design is to calculate the best processing parameters in accordance with the needs set by the user, and predict the life of the tool. The machine learning model completed through training can also be used for parameter optimization design for processing applications that are not included in the experimental modeling data or for a small amount of data, which can improve the shortcomings of the past experimental modeling that require a large amount of experimental data and limited application range. Experimental verification results show that the optimized machining parameters can be within 3 % of the required surface accuracy deviation, while improving efficiency and optimal energy consumption, and predicting tool life within 6% of the deviation time.