A Robust Condition Monitoring Approach in Industrial Plants Based on the Pythagorean Membership Grades

In this paper, a novel approach for improving the performance and robustness of the condition monitoring system in industrial plants is presented. In the off-line stage of the proposal, the Pythagorean membership grade and its complement of a set of n classification algorithms are used to build the rule-based decisions for obtaining an enhanced partition matrix, which allows to improve the positioning of the center of the classes and data clustering. The use of Pythagorean fuzzy sets allow to obtain a larger classification space, and then the robustness of the condition monitoring system with respect to noise and external disturbances is improved. This represents a very powerful advantage in industrial plants, where process variables are affected by such features. The proposal was proven using the kernel fuzzy C-means and Gustafson-Kessel algorithms on experimental data sets and on the Tennessee Eastman process benchmark. The percentages of satisfactory classification obtained with the proposal were greater than 90% in most of the experiments. In all cases, the proposed methodology significantly outperformed the results obtained by other algorithms recently presented in the scientific literature.