Abstract
ENGINE COMPONENT FAULT DIAGNOSTICS USING ARTIFICIAL NEURAL NETWORK
*Diwa James Enyia
ABSTRACT
Power generation in Nigeria has been on the increase over the last decades, since most of the hydraulic resources that are used to drive the hydropower plants have already been deployed. In order to keep electrical energy at affordable prices, several modalities have been considered, such as reduction in operating and maintenance costs of thermoelectric facilities. On this ground, engine component fault diagnostic technique are intended to enhance maintenance quality, reduce engine downtime, and thereby increasing plant availability, while maximising operational profit by keeping engine efficiency at standard level. In this technical research, an artificial neural network ANN based fault diagnostics technique have been developed to assess the health condition of a single shaft heavy duty engine ALSTOM GT11N2. The engine employed was modelled using software known as PYTHIA 2.8, where actual engine data was used in the matching process, afterwards, component parameters degradations were applied to the model to generate neural network training and validation samples. Finally, a three level nested structure was established comprising fault detection, isolation, and quantification functions, each represented by specific neural network architecture. Which are trained to tackle both measurement data without and with noise. In the earlier case, the individual networks presented excellent performance, while in the latter case, good performance have been achieved, despite a few problems with the fault isolation network. Nonetheless, the whole nest neural networks have presented good and acceptable performance respectively when analysing these two cases.
[Full Text Article] [Download Certificate]