This paper is focussed on the research area of fault detection and diagnosis in a complex thermodynamical system, as an axial flow compressor is. The main contribution reached includes a new approach which hybrids a physical model and a Multi-Layer Perceptron (MLP) using the best advantages of both types of modelling. The physical model was used to generate different fault simulations by shifting physical parameters related to faults. After these simulations, the different fault profiles obtained were collected within a fault dictionary. Fault detection was carried out by a MLP whose residuals against the real outputs of the system determined which samples could be considered abnormal. In order to identify and diagnose the fault, the anomalous residuals observed by the MLP were compared with the fault profiles in the dictionary, obtaining a correlation that provided fault causes hypothesis. This methodology has been successfully applied using axial compressor operational data extracted from a real power plant.
Directores: Miguel Ángel Sanz, Antonio Muñoz
Autor: Jesús A. García-Matos
Lugar: Aula de Seminarios

Añadir a agenda electrónica:

When facing an unknown forecasting problem, accuracy on the predictions as well as useful information about the underlying physics of the process are mostly appreciated. In this paper the Thor model; a fully interpretable model with automatic identification is presented. Based on additivity assumptions and piecewise linear regression, it allows the analyst to gain insight about the problem by examining the automatic model selected. Monte-Carlo simulations have been run to assure that the model selection procedure behaves correctly under weakly dependent data. Moreover, comparison over other well-known methodologies have been done to evaluate its accuracy performance, both in simulated data and in the context of short-term natural gas demand forecasting.
Director: Eugenio Sánchez Úbeda
Autor: Alberto Gascón
Lugar: Aula de Seminarios

Añadir a agenda electrónica: