摘要
Semiconductor manufacturing is water-intensive that generates tremendous wastewater during the wafer cleaning for lengthy wafer fabrication processes. In particular, the effluent discharged from wafer cleaning processes contains a huge amount of hydrogen peroxide. The high concentration of hydrogen peroxide will affect the yields of the efficiency of the wastewater treatment plant and cause the environmental risk of pollution. Catalase has been employed in the pre-treatment process of wastewater treatment plants for hydrogen peroxide removal. The pre-treatment process includes multiple pumps for catalase emission adjustment. The oversupply of catalase may lead to the extra cost of the company, while the insufficient quantity of the catalase will cause the risk of environmental contamination, and the adjustment of hydrogen peroxide concentration is not real time. To address realistic needs, this study aims to develop a UNISON framework for hybrid optimization to minimize the potential risk of the hydrogen peroxide wastewater treatment while considering the uncertainty for forecasting the wastewater treatment demands for Industry 3.5. In particular, the proposed framework includes the hydrogen peroxide forecasting model based on long short-term memory and the decision model for catalase dosage pump to derive robust optimal treatments. An empirical study was conducted in a leading semiconductor company to estimate the validity of the proposed framework. The results have shown its practical viability with robust performance under different scenarios for cleaner semiconductor production and sustainability.