摘要：AbstractThis paper introduces an approach for real-time fault diagnosis in industrial processes. The approach aims to build a decision support tool (DST) that helps operators in large-scale processes diagnose faults and make the correct decisions that will keep production schedules on track. The idea is to combine diversified supervised and semi-supervised machine-learning methods to exploit the strength of each of them in fault diagnosis. Despite their accuracy in classifying faults, supervised methods have two limitations: they do not provide meaningful explanations about their results and cannot diagnose novel faults. Semi-supervised methods can detect and isolate novel faults but cannot disclose their root causes. The proposed approach uses the best of both, providing operators with a descriptive decision that improves the diagnosability of detected faults, whether novel or not. The effectiveness of the proposed approach is demonstrated using a benchmark industrial process.
关键词：KeywordsDecision Support SystemsFault DiagnosisIndustrial ProcessesSupervisedSemi-Supervised LearningInterpretable Predictive Models
