The diagnostic ability of probabilistic neural networks (PNN) for detecting sensor faults on gas turbines is examined. The structure and the features of a PNN, for sensor fault detection, are presented. It is shown that with the proposed formulation, a powerful tool for sensor fault identification is produced. A particular feature of the PNN produced is the ability to detect sensor faults even in the presence of engine component malfunction, as well as on deteriorated engines. In such situations, the size of bias that can be identified increases. The way to establish the limits of sensor bias that can be detected is presented along with results from application to test cases with realistic noise magnitudes. The diagnostic procedure proposed here is also supported by an engine performance model. The data used for setting up and testing the PNN are generated by such a model.
Propulsion diagnostic method evaluation strategy (ProDiMES) offers an aircraft engine diagnostic benchmark problem where the performance of candidate diagnostic methods is evaluated while a fair comparison can be established. In the present paper, the performance evaluation of a number of gas turbine diagnostic methods using the ProDiMES software is presented. All diagnostic methods presented here were developed at the Laboratory of Thermal Turbomachinery of the National Technical University of Athens (LTT/NTUA). Component, sensor, and actuator fault scenarios that occur in a fleet of deteriorated twin-spool turbofan engines are considered. The performance of each diagnostic method is presented through the evaluation metrics introduced in the ProDiMES software. Remarks about each methods performance as well as the detectability and classification rates of each fault scenario are made.
The paper presents the use of different approaches to engine health assessment using on-wing data obtained over a year from an engine of a commercial short-range aircraft. The on-wing measurements are analyzed with three different approaches, two of which employ two models of different quality. Initially, the measurements are used as the sole source of information and are post-processed utilizing a simple “model” (a table of corrected parameter values at different engine power levels) to obtain diagnostic information. Next, suitable engine models are built utilizing a semi-automated method which allows for quick and efficient creation of engine models adapted to specific data. Two engine models are created, one based on publicly available data and one adapted to engine specific on-wing “healthy” data. These models of different detail are used in a specific diagnostic process employing model-based diagnostic methods, namely the Probabilistic Neural Network (PNN) method and the Deterioration Tracking method. The results demonstrate the level of diagnostic information that can be obtained for this set of data from each approach (raw data, generic engine model or adapted to measurements engine model). A sub-system fault is correctly identified utilizing the diagnostic process combined with the engine specific model while the Deterioration Tracking method provides additional information about engine deterioration.
Propulsion Diagnostic Method Evaluation Strategy (ProDiMES) offers an aircraft engine diagnostic benchmark problem where the performance of candidate diagnostic methods is evaluated while a fair comparison can be established. In the present paper, the performance evaluation of a number of gas turbine diagnostic methods using the ProDiMES software is presented. All diagnostic methods presented here were developed at the Laboratory of Thermal Turbomachinery of the National Technical University of Athens (LTT/NTUA). Component, sensor and actuator fault scenarios, that occur in a fleet of deteriorated twin-spool turbofan engines are considered. The performance of each diagnostic method is presented through the evaluation metrics introduced in the ProDiMES software. Remarks about each methods performance as well as the detectability and classification rates of each fault scenario are made.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.