A Fast, Statistical, Machine-learning Approach for Automotive Semiconductor Test Reduction

Modern electronic chips on advanced technologies include complex circuits performing a variety of functions. Extensive testing is required to identify and screen defective parts to prevent escapes that may cause field failures. As products mature, there is increased motivation to reduce test time without increasing quality risks, especially for automotive applications with stringent quality requirements. This work presents a novel, fast, scalable, and statistically rigorous approach for test-time reduction. The approach combines association analysis and a greedy algorithm to identify redundant tests for removal without compromising the quality of shipped parts.