Rotating machinery fault identification using model-based and data-based techniques integration.

Abstract

The use of modern maintenance techniques in rotating machinery has changed the scenario in energy companies, both concerning generation, such as in hydroelectric, thermoelectric, and wind power plants, and the exploitation of resources such as oil and natural gas. The quest for excellence makes new sophisticated and refined tools more necessary and, in this way, academic research is increasingly approaching industrial reality. Wishing to contribute to the answer to such demand, so that in the future new integration and asset management tools can be developed for industrial applications, this work proposes a hybrid methodology in which already recognized techniques are applied and integrated to highlight their qualities and overcome their weaknesses. Applying model-based techniques, inverse methods, Bayesian inference, and data-driven statistical methods, the proposed methodology enables machines and faults modeling and simulation, providing information that would allow the identification, detection, and diagnosis of malfunctions in real machines. To illustrate the methodology, three examples were developed, in which different machines with mechanical unbalance were considered.