Diagnóstico de fontes de perturbação em plantas industriais

Abstract

In industrial processes, oscillatory and non oscillatory disturbances usually travel along propagation paths, and are indicative of degradation in control loops, disturbances, limitations in equipments or control system. In this work, a methodology based in two algorithms from literature to detect these disturbances and the root cause analysis of theses disturbances is proposed. The detection of oscillatory signals is done using the filtered autocorrelation from the power spectral density. The non-periodic disturbances are detected through analysis of principal components of the power spectral density filter. The diagnosis of both disturbances is made based on the cause and effect relations of the loops and energy analysis. The application of the methodology to a power plant illustrates the usefulness of the proposed algorithms