Failure Detection and Primary Cause Identification of Multivariate Time Series Data in Semiconductor Equipment

Downtime caused by equipment failure is the biggest productivity problem in 24-hour a day operations of semiconductor industry. Although some failures are inevitable, increases can be gained if causes detected quickly and repaired, thus reducing downtime. Univariate control charts commonly used to detect failures. However, because complexity process structural characteristics equipment, detection identification may difficult. The purpose this study use correlations variables predict parts replaced identify primary proposed method consists four steps: (1) conversion multivariate time series data into signature matrixes, rid="deqn2" xmlns:xlink="http://www.w3.org/1999/xlink">(2) anomalies through convolutional autoencoder, rid="deqn3" xmlns:xlink="http://www.w3.org/1999/xlink">(3) learning classification models with supervised methods that residual matrixes fault sections, rid="deqn4-deqn7" xmlns:xlink="http://www.w3.org/1999/xlink">(4) application an explainable algorithm interpret model. effectiveness applicability demonstrated actual obtained from 8-inch ashing produces semiconductors on silicon wafers.