Combining Active Learning and Fisher Discriminant Analysis for the Semi-supervised Process Monitoring

Abstract Fault detection and fault classification are two extremely important parts in process monitoring. However, obtaining the true labels of the industrial data is often time-consuming and expensive in practice, which brings great difficulties to the fault classification. To solve this problem, a new semi-supervised fault classification method is proposed, which combines the active learning and fisher discriminant analysis (FDA). The limited capacity of FDA can be effectively improved by integrating the active learning for semi-supervised fault classification. The experiments on the Benchmark of Tennessee Eastman process (TEP) prove the effectiveness of the proposed method.