Reducing Test Cost of Infrared Detectors: A Machine Learning Approach to Failure Prediction of Infrared Detectors

Abstract

It is important to ensure the infrared detectors will work properly in practical application, so a series of tests must be carried out before putting them into application. In order to reduce test cost, this paper proposed a two-stage failure prediction model for infrared detectors during tests based on the logistic regression model. The solution has shown significant business values. With only the first three test stages’ data, under the condition that 99% of the normal detectors are classified as normal products, we can screen out 54.55% of the detectors which will fail before the last vacuum test. Furthermore, in the second stage prediction, we can screen out 80% of the final failure detectors.