Electronics Industry Use Cases
Circuit-board FDC (Fault Classification & Diagnostics)
Machine learning is unquestionably the best technique for Fault Detection & Classification in the testing of assembled printed circuit board-level electrical modules, or other fully assembled electrical modules. Time series & sequence test data from electrical testers can be as input to deep learning & machine learning models that will classify faults by known type to very high accuracy of precision and recall (low false positive and low false negative rates). In many cases the algorithms can also determine the root cause of failure (for example a specific discrete component that is out of spec). These methods can also predict assemblies that are in spec, but will likely fail after burn-in at the customer, or have a sub-standard time-to-failure, hence boosting quality.
Circuit-board inspection / image-based defect detection
Convolutional Neural Networks (CNNs) allow for the highest accuracy final inspection & detection and classification of defects in assembled circuits boards, based on imaging of these pieces. Wrong components, missing or misplaced components, and board damage can all be detected with high accuracy.
Advanced packaging techniques: POP, Si Interposer/RDL, FOWLP/FOPLP, etc.
For OSATs or OEMs employing any of these advanced, high density packaging techniques, both of the electrical-fault FDC and image-based defect detection techniques above are highly relevant and applicable. Therefore, these use cases are of high interest to the OEM providers of inspection and test equipment, in order to differentiate their solutions, and also to OSATs or Fab's doing their own advanced packaging who want to perform their own customized advanced testing.