Tzu-Ching Yang
Facility Engineer
Winbond Electronics Corporation
Education
- Master of Science (M.S.) in Energy and Refrigerating, Air-Conditioning Engineering, National Taipei University of Technology (TAIPEI TECH)
Experience
- Semiconductor Facility Engineer: Specializing in energy efficiency optimization and system stability for Fan Filter Units (FFUs) and Make-up Air Units (MAUs) in high-tech wafer fabs.
AI & Data-Driven Energy Efficiency: Driving practical energy-saving initiatives in cleanroom HVAC systems utilizing artificial intelligence and data analytics.
Hands-on Cleanroom Optimization: Actively involved in optimizing cleanroom air-conditioning systems and evaluating filtration media to validate tangible energy reduction and carbon mitigation benefits.
Smart Factory & Net-Zero Integration: Focusing on the practical deployment of AI, net-zero initiatives, and smart factory applications using tools such as JMP, Power BI, and GPT.
Speech Title and Abstract
Optimizing FFU Operational Strategies in Semiconductor Fabs: A Data Science and AI-Powered Energy-Saving Practice
As semiconductor fabrication plants face the urgent challenge of transitioning toward net-zero emissions, Cleanroom Fan Filter Units (FFUs)—due to their high energy consumption and absolute critical impact on production environment stability—have become a core focus of green transformation. Traditional operational strategies rely heavily on conservative rules of thumb, keeping fan speeds excessively high, which results in substantial energy waste and carbon footprint pressures.
This study introduces a systematic data science methodology to break through the limitations of traditional trial-and-error approaches. First, a Full Factorial Design of Experiments (DOE) was implemented to capture multi-dimensional dynamic data, followed by Principal Component Analysis (PCA) to effectively eliminate multicollinearity among environmental factors and streamline key metrics. Furthermore, Random Forest and LASSO regression algorithms were combined to identify core influential variables, building a "Z-score Comprehensive Decision Model." Under the strict prerequisite of ensuring 100% compliance with cleanroom cleanliness, temperature, and humidity standards, this model successfully identifies the system's "golden energy-efficiency balance point."
Empirical results demonstrate that the optimized strategy achieved an annual energy savings of 51,862 kWh in a single target zone, yielding significant carbon reduction benefits. Beyond these immediate results, this project establishes a standardized AI analytics SOP. When scaled across the facility, the projected annual energy savings potential exceeds 1 million kWh. This case study successfully defines a sustainable management benchmark for high-tech fabs transitioning from intuition-based decision-making to AI-and-data-driven operations.