Gas Turbine Fault Diagnosis Using Probabilistic Neural Networks

Abstract

Efficiency of gas turbine monitoring systems pri marily depends on the accuracy of employed algorithms, in particular, pattern classification techniques for diag nosing gas path faults. In recent investigations many tech niques have been applied to classify gas path faults, but recommendations for selecting the best technique for real monitoring systems are still insufficient and often con tradictory. In our previous work, three classification tech niques were compared under different conditions of gas turbine diagnosis. The comparative analysis has shown that all these techniques yield practically the same accu racy for each comparison case. The present contribution considers a new classification technique, Probabilistic Neural Network (PNN), and we compare it with the tech niques previously examined. The results for all compari son cases show that the PNN is not inferior to the other techniques. We recommend choosing the PNN for real monitoring systems because it has an important advan tage of providing confidence estimation for every diagnos tic decision made.

A Review on Gas Turbine Gas-Path Diagnostics: State-of-the-Art Methods, Challenges and Opportunities

“…The results revealed that the proposed scheme was capable of diagnosing all the considered fault scenarios with sufficiently high accuracy. Recently, the fault classification performance of PNN was compared with MLP and RBF by Loboda and Robles [153], and obtained similar accuracies. In general, as per this review, most of the previous PNN based GT diagnostic techniques were utilized for fault classification tasks.…”

Research on Gas-Path Fault-Diagnosis Method of Marine Gas Turbine Based on Exergy Loss and Probabilistic Neural Network

“…A PNN is a kind of neural network based on statistical principles, which is commonly employed to solve problems of pattern classification [30]. The PNN is an efficient and robust classifier, obtained when the Bayes strategy for decision making is combined with a non-parametric estimator for probability density functions [31].…”

“…The task of the pattern layer is to calculate the pattern, matching the relationship between the testing samples and the training samples, and to centralize the categories with high similarity. The number of neurons in the pattern layer is equal to the sum of the number of trained samples in all categories, and the neuron X ij computes the pattern layer output as follows [30]:…”

“…in the pattern layer is equal to the sum of the number of trained samples in all categories, and the neuron Xij computes the pattern layer output as follows[30]:…”

Enhanced turbine monitoring using emissions measurements and data reconciliation

“…The components of monitoring and detection include data extraction, fault detection, fault location and prognosis [1][2][3][4]. Much of the current literature focuses on fault location algorithms which can be broken down into pattern recognition methods such as fuzzy logic [5,6], genetic algorithms [7], Bayesian belief networks [8][9][10], and neural networks [11][12][13][14] and model identification methods such as Kalman filtering [15] and weighted least squares [16][17][18]. But despite continuous developments in all of these physical and computational techniques, serious gas turbine problems can still quickly develop before corrective actions can take place.…”