Detecting Defects with Data Science
A frequent application of data science in the industry, and more precisely in semiconductor manufacturing, is to detect failed components.
In this article, Anirudh Kondaveeti, a principal data scientist at pivotal, uses classic data science techniques to detect failed wafers on a production assembly line. A process that includes feature extraction, dimensionality reduction through Non-negative Matrix Factorization, outlier detection, and clustering. The final visualization helps uncover common defect patterns occurring in the manufacturing process.