Jiménez Sánchez, D., Quiñones Grueiro, M., Silva Neto, A. J. e Llanes Santiago, O., A Regularized Inverse Problem Approach for Robust Condition Monitoring in Industrial Systems, Capítulo, pp. 177-198, em Shi, P., Stefanovski, J. and Kacprzyk, J., (Eds) Complex Systems: Spanning Control and Computational Cybernetics: Foundations. Studies in Systems, Decision and Control, Vol. 415, ISBN: 978-3-031-00977-8 (Hardcover), 978-3-031-00978-5 (eBook), Editora Springer, Cham, 2022.
Condition monitoring is very important in modern industry in order to increase the safety of industrial plants and the economic benefits. Schemes based on model inversion or system inversion represent an important branch of the available solutions in model based condition monitoring. These techniques allow the development of detection, isolation and successful estimation of the fault magnitude. However, most of the proposed methods do not consider the noise present in industrial control systems which significantly affects the performance of the condition monitoring systems. They do not consider either the occurrence of multiple faults. In this paper, a proposal for robust condition monitoring, formulated as the solution of a regularized inverse problem in discrete linear time invariant systems is presented. Single and multiple faults are reconstructed by using the vector of residuals in the presence of noise. Tikhonov regularization is used to obtain a stable solution when noise in the measurements is considered. The proposed approach is applied to a case study with satisfactory results.