應用田口方法於工具機軸承裝配製程最佳化之研究

• Main Authors: Ya-Chi Huang, 黃雅琪
• Other Authors: Dr. Wen-Tsann Lin
• Format: Others
• Language: zh-TW
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/21504676264144531624

碩士 === 國立勤益科技大學 === 工業工程與管理系 === 104 === Under intense global competition, Taiwan's machine tool industry is gradually being affected. And in the era of jungle, the only surviving chance is upgrading product quality. The keys to product upgrading include short lead time, low cost and high quality. Study on these key objectives, it will significantly enhance competitiveness for machine tool industry. This study is aimed at designs and assembly processes of motor shaft of headstock in machine tool. Through relevant personnel and literatures, understand the key technology and necessary resources, and integrate them into key technology roadmap. Then a technical layout for the product can be achieved. Figure out the quality characteristics of the key factors and by using Taguchi Methods OMEGA transformation to predict the outcome of the defective rate. Hence, the optimized reference data from motor shaft bearing packing in machine tool can be obtained. Such data can lead to lowering the defective rate effectively, lowering the cost and upgrading product quality. Then standardize the design changes; expecting the best quality can last and ensuring steady progress of production. Ultimate form for bearing can also be obtained by Taguchi Methods in deep groove ball bearings. And it is then assembled with help of fixture procedure. With dimensional tolerance of 0.005mm as optimal for bearing and bore in assembly, the defective rate has been improved from previous 35% down to 6%, which is close to 7.9% in earlier prediction. Thus, an effective improvement of quality in headstock is achieved. Finally, Back Propagation Neural Network was used as experimental network model in this study. Later, it was analyzed by Matlab software where input reference values were the three factors in Taguchi Methods and output reference values were the data obtained from headstock assembly experiments. Convergence status has been proven as MSE was equal to 0.0032932.