Fault diagnosis on CVD equipment via Neural Network Approach

• Main Authors: Chien-Hsun Lai, 賴建勳
• Other Authors: Tai-Yue Wang
• Format: Others
• Language: zh-TW
• Published: 2007
• Online Access: http://ndltd.ncl.edu.tw/handle/78959151770731764176

碩士 === 國立成功大學 === 工業與資訊管理學系專班 === 95 === The integrated circuits (ICs) industry is the industry with technology and capital intensively. Thus how to make the equipment utilization higher and to compensate the equipment depreciation cost, to reduce product cost, and to create a company’s competitive advantages are the targets to be pursued. In semiconductor manufacturing, the chemical vapor deposition (CVD) equipment is a key system in producing integrated circuits. To maintain equipment in good condition and stable throughput rate, CVD faults should be diagnosed accurately and timely. At present, the equipment maintenance still depends deeply on the engineers’ experience. Due to the fact that high-tech employee has higher leaving job rate, the technical experience is not easily transferred and enterprise knowledge can not be aggregated effectively. For maintaining the IC manufacturing equipments effectively, some methods were developed by scholars such as case-based reasoning (CBR) and wavelet theory et al. These fault diagnosis approaches, however, still can not meet the needs in practice. Thus a model for CVD fault diagnosis is needed. In this research, a system consisted of artificial neural network (ANN) and expert’s experience is presented. A back-propagation neural network (BPN) was used to capture the causal relationships between fault symptoms and root causes. The results have shown that proposed model has an excellent prediction capability of CVD machine fault root causes diagnosis. On the other hand, the result for time prediction of recovering the CVD machine is not as good as the results of CVD machine fault root causes diagnosis due to the respondents have different cognition for recover time. However, if we can define the recover time clearly, the fault diagnosis performance could be raised and can be applied in process equipment for related manufacturing fields.