Introduction:

To achieve high levels of quality production with efficient use of raw materials, industries must have fault diagnosis systems for processing and analyzing the information obtained through data acquisition and control systems. The performance of fault diagnosis systems is affected by noise, information loss in the data acquisition process, the presence of unknown faults, and in the case of multi-mode processes, the occurrence of faults during transitions between stationary modes. The latter problem derives from the fact that diagnostic methods developed for stationary modes cannot be applied satisfactorily during transitions.

Methods:

In the present paper, a group of new paradigms is presented to provide solutions to the above-mentioned problems through the effective use of data-driven methods, clustering, imputation, hybrid algorithms, and computational intelligence tools. The proposals are validated in benchmark problems established as study cases in the scientific literature representing chemical processes, electromechanical systems, and urban water distribution networks.

Results:

Besides demonstrating the effectiveness of the proposals, the set of benchmark processes considered is very important for our country in its prospects for development, saving and caring of the environment.

        · abstract in Spanish     · text in Spanish     · Spanish ( pdf )