Aerodynamic model identification of supersonic aircraft using Bayesian approach-based Box–Jenkins structure

Abstract

Nonlinear aerodynamics complexities of trending supersonic fighter aircraft entail formulation of a robust and reliable System Identification (Sys ID) technique that is capable of giving deep insight into its nonlinear characteristics and being self-capable of fitting into future advancements. This study discovers a decoupled longitudinal aerodynamic model of an open-loop supersonic aircraft using a novel algorithm that blends grey-box modeling architecture i.e. Box–Jenkins (BJ) structure with Bayesian approach, named as Box–Jenkins–Bayesian–Estimation (BJBE). BJ model utilizes a nonlinear least square estimator for parameter identification, which has been improved by the Levenberg–Marquardt algorithm for parameter error minimization, and further refinement is accomplished through Bayes’ theorem using its maximum-a-posteriori characteristics. Bayesian estimation, due to its a-priori feature, fully explores grey-box modeling BJ structure, which no other estimation technique does. The proposed solution involves the construction of a discrete-time BJ model using a simulated input–output dataset generated from the Flight Dynamic Model of F-16 aircraft, followed by the reduced-order model using Bayesian information criteria and parameter optimization using Bayesian theorem. A closer analysis of results has been conducted through statistical techniques like residual analysis, best-fit percentage, fit percentage error, mean squared error, and model order. Results show good agreement between model predictions and simulated flight data with an accuracy of 82.42%. Based upon this research, control laws of supersonic jets have been investigated through a novel technique, further leading to the development of its flight simulator module.