In the realm of semiconductor manufacturing, Automated Optical Inspection (AOI) stands as a critical process, ensuring the quality and reliability of products. Traditionally, the optimisation of inspection tolerances in AOI has been a subjective task, heavily reliant on the expertise and intuition of engineers. This approach, while practical, often leads to inconsistencies and inefficiencies. To address this challenge, a team of researchers has embarked on developing an intelligent, data-driven framework aimed at optimising inspection tolerances in a more objective and consistent manner.
The team, comprising Shruthi Kogileru, Mark McBride, Yaxin Bi, and Kok Yew Ng, has identified a significant gap in the current research landscape. Most existing studies focus primarily on minimising false calls, often at the risk of overlooking genuine defects. This oversight can have serious implications, particularly in critical sectors such as medical, defence, and automotive industries, where product quality is paramount.
The researchers’ innovative approach introduces the use of percentile rank, among other logical strategies, to ensure that actual defects are not disregarded. By leveraging data-driven techniques, the framework aims to achieve a point where every flagged item is a true defect, thereby eliminating the need for manual inspection. This not only enhances efficiency but also offers significant time and cost savings.
The proof of concept for this framework has yielded promising results. At the 80th percentile rank, the system achieved an 18% reduction in false calls while maintaining a 100% recall rate. This means that the system successfully identified all genuine defects, demonstrating both its reliability and effectiveness.
The implications of this research extend beyond the semiconductor industry. In the defence and security sector, where precision and reliability are non-negotiable, such a framework could revolutionise quality control processes. By ensuring that every defect is detected and addressed, the framework could significantly enhance the safety and efficacy of defence technologies.
Moreover, the framework’s ability to reduce false calls offers substantial operational benefits. In high-stakes environments, the reduction of false positives can lead to more efficient use of resources, allowing defence personnel to focus on genuine threats rather than investigating false alarms.
As the researchers continue to refine their framework, its potential applications in the defence and security sector are likely to expand. The move towards data-driven, objective decision-making processes represents a significant step forward in the ongoing quest for technological excellence and operational efficiency. Read the original research paper here.