Numerical Model Establishment and Neural Network Process Optimization for Simulated Selective Laser Melting Process

• Main Authors: CHEN,KUAN-YU, 陳冠宥
• Other Authors: WU,HSUAN-CHUNG
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/ud5988

碩士 === 明志科技大學 === 材料工程系碩士班 === 107 === The selective laser melting process is one of the layers of manufacturing. This study uses numerical simulations to simulate phenomena in processes in selective laser melting processes, such as spheroidization, voids, and cracks, which leads to the final The finished product reduces the mechanical properties of the finished product due to the presence of defects. This study has a mesoscopic scale simulation system that can handle thermal properties, fluid flow, surface tension, three-dimensional heat source, convection and conduction, by changing the laser power and scanning speed, and analyzing the bath temperature, bath size, holes and The surface topography has the same trend as the single-track melting experiment, and only 8 sets of simulation parameters are brought into the neural network. The learning ability of the neural network predicts the trend of the result, and finally, by the gene algorithm. Finding extreme value capabilities to predict process optimization parameters not only reduces the number of experiments or simulations, but also reduces the cost and time of the experiment.