ARTIFICIAL INTELLIGENCE FOR ENHANCED SEMICONDUCTOR MANUFACTURING: FEATURE SELECTION FOR YIELD IMPROVEMENT

Abstract

In semiconductor manufacturing, ensuring high yield rates is critical for optimizing production
efficiency and minimizing costs. However, the vast number of signals collected from sensors and
process measurement points often contain a mix of relevant information, noise, and irrelevant data,
making it challenging for engineers to identify the key factors affecting yield. Feature selection
techniques are instrumental in addressing this challenge, as they help identify the most relevant
signals that significantly impact yield. In conventional semiconductor manufacturing, engineers are
inundated with an extensive array of signals, making it cumbersome and time-consuming to pinpoint
the critical factors influencing yield excursions. This data overload often results in suboptimal process
efficiency and increased production costs. Traditional approaches may not effectively distinguish
between useful information and noise, leading to inefficient troubleshooting and reduced yield rates.
Additionally, manual feature selection processes are labour-intensive and may not uncover complex
causal relationships between variables, limiting their effectiveness in enhancing semiconductor
manufacturing operations. To overcome the limitations of the conventional approach, this study
proposes the use of artificial intelligence-based feature selection techniques. By leveraging
sophisticated algorithms, the proposed system will rank features according to their impact on
semiconductor manufacturing yield. These techniques will not only streamline the identification of
crucial variables but also unveil causal relationships within the data, providing a deeper understanding
of the production process. The application of cross-validation ensures robustness and reliable
evaluation of feature relevance for predictability using error rates. The goal is to empower engineers
with a more efficient and data-driven approach to semiconductor manufacturing, resulting in
increased yield rates, reduced production costs, and shorter learning cycles. The preliminary results
presented here demonstrate the potential of this approach, highlighting the promise of artificial
intelligence in revolutionizing semiconductor manufacturing optimization.