Gas Turbine Off-Design Performance Adaptation Using a Genetic Algorithm

Abstract

Gas turbine gas path diagnostics is heavily dependent on performance simulation models accurate enough around a chosen diagnostic operating point, such as design operating point. With current technology, gas turbine engine performance can be predicted easily with thermodynamic models and computer codes together with basic engine design data and empirical component information. However the accuracy of the prediction is highly dependent on the quality of those engine design data and empirical component information such as component characteristic maps but such expensive information is normally exclusive property of engine manufacturers and only partially disclosed to engine users. Alternatively, estimated design data and assumed component information are used in the performance prediction. Yet, such assumed component information may not be the same as those of real engines and therefore poor off-design performance prediction may be produced. This paper presents an adaptive method to improve the accuracy of off-design performance prediction of engine models near engine design point or other points where detailed knowledge is available. A novel definition of off-design scaling factors for the modification of compressor maps is developed. A Genetic Algorithm is used to search the best set of scaling factors in order to adapt the predicted off-design engine performance to observed engine off-design performance. As the outcome of the procedure, new compressor maps are produced and more accurate prediction of off-design performance is provided. The proposed off-design performance adaptation procedure is applied to a model civil aero engine to test the effectiveness of the adaptive approach. The results show that the developed adaptive approach, if properly applied, has great potential to improve the accuracy of engine off-design performance prediction in the vicinity of engine design point although it does not guarantee the prediction accuracy in the whole range of off-design conditions. Therefore, such adaptive approach provides an alternative method in producing good engine performance models for gas turbine gas path diagnostic analysis.

Gas Turbine Gas Path Fault Diagnosis based on Adaptive Nonlinear Steady-State Thermodynamic Model

“…For previous gas path analysis methods (e.g., approaches developed by authors [2,6,7]), health parameters for gas path components were defined using component absolute performance parameters (e.g., mass flow rate G and rotational speed n for compressor and turbine). When the gas path measurements are obtained at slightly different ambient conditions from baseline, they should be corrected to the baseline condition to eliminate the deviation of component performance parameters caused by varying ambient conditions (e.g., ambient temperature, pressure and humidity).…”

“…For the previous gas path analysis method, (e.g., approaches developed by authors [2,6,7]), because the gas path component performance parameters are adaptive parameters, two steps should be included where the first step is to estimate engine degraded component performance parameters p using a performance adaption technique with gas path measurements, and the second step is to estimate component health parameters SF  by comparing healthy and degraded performances at the component level to obtain the magnitude of shift of the characteristic curves on component' maps due to degradation. Here, a new GPA approach has been developed to detect engine performance degradation by using gas path measurements z to directly output the deviations of component health parameters SF  .…”

Review on Advanced Health Monitoring Methods for Aero Gas Turbines using Model Based Methods and Artificial Intelligent Methods

“…Through each interval, the change in the independent parameter becomes increasingly smaller and the process can be stopped when the change in the independent parameter has reached a convergence criterion that suits the needs [36]: (6) wwhere M is the number of measurements, is the actual deteriorated measured parameter vector, is the calculated deteriorated measured parameter vector that is based on the detected component parameter vector , and is the convergence criteria.…”

“…Through each interval, the change in the independent parameter becomes increasingly smaller and the process can be stopped when the change in the independent parameter has reached a convergence criterion that suits the needs [36]:…”

Compressor performance modelling method based on support vector machine nonlinear regression algorithm

“…A lot of efficacious methods have been proposed to improve the calculation precision of the thermodynamic performance model for gas turbine engines, mainly through correcting the known component characteristic maps [12][13][14] or producing new ones [15] based on gas path measurable parameters. Simani et al [16] introduced a gas turbine thermodynamic modelling method which used an optimization algorithm to seek an optimal set of scaling factors for the component maps, and subsequently the proposed method was expanded by Lambiris et al [17].…”