Ergonomics Management of High-Tech Fab Based on Visual Recognition-Case Study for Etching Process PM

The etching process is an important item in the advanced semiconductor process. After each layer of film and photolithography, it must be etched. Under the current large-scale use of single-wafer processing machines, the PM frequency of the etching machine is greatly increased; so many equipment engineers have serious problems. For cumulative ergonomics injury, under the provisions of the Occupational Safety and Health Act of Taiwan, each semiconductor factory must implement an effective ergonomics injury prevention plan. Therefore, based on the aforementioned problems, this study hopes to use AI methods to assist company management to reduce ergonomics injury, the situation allows workers to work with more peace of mind. However, this study is mainly based on visual recognition technology to perform the operation recognition of PM etching personnel, and compares the recognition results with the cloud system calculation, and the calculation method mainly uses CNNs. The results found that the use of visual recognition technology in this study can effectively identify 95% of unsafe actions. When it is compared with the default alert and alarm values, the APP signal can be transmitted to the site manager with the WiFi connection from cloud system. For the operation the non-conformity rate has been reduced from 40% at the beginning to 15%, for the reduction from 38% to about 15%, the improvement trend is in line with wafer operations, which is an important contribution of this study.