The Application of Regression Models to Enhance Gas Turbine Engine Fault Tolerance

Introduction. Improving the fault tolerance of automatic control systems (ACS) for gas turbine engines (GTE) relies on structural redundancy, achieved by duplicating measurement channels for key engine parameters. However, determining which channel provides dependable information poses a challenge. A solution proposed involves employing an integrated mathematical model as an "arbitrator". This article focuses on presenting regression models for the main parameters of a gas turbine engine. The study aims at developing a GTE parameter model based on regression models, assess the models' adequacy on both training and predictive datasets, and identify the optimal mathematical model. The article addresses the mathematical model structure of GTE main parameters, presents an experiment setup methodology, examines linear and polynomial regression models, calculates model adequacy, and selects the best models. Findings and conclusion. Regression models using machine learning were built to evaluate the GTE main parameters. During model analysis, various mathematical combinations of the main parameters were considered alongside the main parameters themselves. The research identified significant model regressors and optimal models based on the learning algorithm. A comprehensive analysis of model adequacy revealed satisfactory results for the parameters  (P2; hnc; αвна; n2; n1), while the search for alternative model types for the parameters (T4; αди; αруд)  is proposed.