Fault isolation based on Bayesian fused lasso

Fault detection and isolation (FDI), which is a critical part of modern industrial systems, plays a key role in the maintainability, safety, and reliability of processes. Existing FDI approaches are dependent on varying degrees of knowledge of the process, limiting their implementation in practical industrial processes. Based on the least absolute shrinkage and selection operator (lasso), this paper proposes Bayesian fused lasso to overcome the above-mentioned problem. The fault isolation problem is converted into a quadratic programming problem with constraints, which can be solved satisfactorily by the Bayesian fused lasso. Ultimately, the probability distribution of every fault variable can be obtained. In the case of unknown fault directions, fault isolation is carried out. Therefore, the possibility of misdiagnosis was reduced. The reliability and effectiveness of the proposed method are illustrated with the case.