Fault Detection via Domain-Knowledge-Based Training Data Refinement
A domain knowledge–based data refinement methodology for detecting defective products that cannot be filtered out in the LQC process.
2023.07 – 2024.10
·Motivation
Defective products that are not detected during the LQC process are shipped to customers and later returned due to quality issues.
·Methodology
Based on domain knowledge, the conditions of abnormal patterns are specified and used to refine the model's training data. A self-supervised learning method is employed, accounting for data contamination and the extremely limited information on minor defects.
·Contribution
Before refinement, defective products were rarely detected; after refinement, more than 80% were identified. The model also learned to distinguish abnormal from normal data more effectively.
