INTELLIGENT DIAGNOSIS FOR TURBINE BLADE FAULTS USING ARTIFICIAL NEURAL NETWORKS AND FUZZY-LOGIC

An on-line diagnostic system, which consists of three components: data acquisition, feature extraction, and pattern recognition, is described for detecting two main. faults of turbine blades. Due to some significant noise in the signals, multiple vibration sensors are applied in order to obtain more-accurate results. The system identifies the existence of an unbalanced blade, or a loose blade, or the combination of these two faults. A feature-extraction algorithm, a frequency analyzer, was developed, and the features are formulated as the inputs of an artificial neural network using backpropagation. Fuzzy models are employed to dynamically update the training parameters, training rate, momentum, and steepness of the activation function, in order to speed up the training speed. The validity of the system was demonstrated with 90% discrimination accuracy in an experiment when multiple sensors were used. Moreover, the training time is reduced by 97.3% while all three training parameters are dynamically updated.