Gas path diagnostics is a key aspect of the engine health monitoring process, which aims to identify and predict engine component faults using installed sensor data before they become catastrophic failures. To achieve high prediction accuracies however, current data-derived diagnostic models tend to be engine-type specific while the physics-model-based approaches are known to be time-consuming, especially for complex engine configurations. This paper proposes an integrated approach for accurate and faster isolation and prediction of multiple-degraded, gas turbine component faults which combines feature extraction using the Principal Component Analysis (PCA), machine learning capability of a multi-layer perceptron, artificial neural network (MLP-ANN) with the model-based, non-linear Gas Path Analysis (GPA) technique. In this hybrid approach, the PCA first transforms the measurement fault signature into a fault-feature domain, which becomes an input to the multi-label, multi-class ANN classifier model used to isolate the potential faulty components. The non-linear GPA finally quantifies the magnitude of degradation that produced the recorded fault signature. Once trained and validated, the PCA-ANN model is deployed as a data pre-processing mechanism before the actual GPA step. This method was validated and assessed based on the thermodynamic gas path model of a 2-shaft, high-bypass ratio, turbofan engine. For training and testing the ANN model, a total of 64,000 measurement samples for 16 sensors, with 1-, 2- and 3-component fault combinations of various magnitude, were generated by steady-state performance simulation of the engine model at the reference operating condition. The proposed methodology isolated faulty from clean components across all cases, thus significantly reducing the search space and calculation time for the GPA prediction step. A quantitative analysis of the integrated approach compared to their independent methods showed an overall increase in accuracy of 12% compared to a stand-alone regression ANN model, and 0.5% compared to the non-linear GPA only, having the same number of iterations. Finally, the method was shown to produce accurate fault prediction with the loss of up to 2 engine sensors and in the presence of sensor noise, by virtue of using the PCA for preliminary feature extraction.


An Integrated Principal Component Analysis, Artificial Neural Network and Gas Path Analysis Approach for Multi-Component Fault Diagnostics of Gas Turbine Engines

Division / Sponsor:
International Gas Turbine Institute (IGTI)

Paper Proceedings: https://asmedigitalcollection.asme.org/GT/GT2020/volume/84058

Turbo Expo - Virtual Conference

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.