A Real-Time Analysis Tool based on Surrogate Models for Accurate Prediction of Combined Flight External Loads

ABSTRACT: The proposed effort aims to develop and demonstrate an innovative real-time tool for accurate, rapid analysis and prediction of combined external loads using non-proprietary physics-based models and existing flight-test data as the starting point. The salient aspects of the proposed solution are: (1) mathematically formal multivariate regression and surrogate modeling techniques to enable accurate, real-time load prediction; (2) a novel data fusion approach to accommodate and combine data from various sources to update and enhance prediction, (3) input selection/screening to identify primary load-driving parameters; and (3) a modular software framework to facilitate seamless integration with the USAF-designated tool (IADS) into AF real-time control room environment. In Phase I, key components, including a surrogate model module, a data fusion module, an input selection & sensitivity analysis module, and a real-time load prediction engine will be developed. Feasibility will be demonstrated via multiple case studies of USAF interest, in which physics-based models, wind tunnel data, and flight test data will be analyzed using the developed software and its performance (e.g., accuracy, speed, reliability, and integrability to existing tools) will be assessed. The Phase II effort will focus on capability extension, algorithm optimization, IADS integration, and extensive software validation and technology demonstration. BENEFIT: The proposed simulation technology will be of value in military (Air Force, Army, MDA, etc.), NASA, and civilian sectors. The tool will serve as a technology backbone specifically meeting the needs of Air Force Test Center (AFTC) for prediction of combined static and dynamic external loads using non-proprietary physics-based models and existing flight-test data in a real-time environment. It delivers flight test engineers and aircraft design engineers a valuable tool to: (1) determine, on-the-fly, the condition of external loads and their effects on flight state; (2) quickly screen and identify the primary parameters contributing to load generation and measurement-prediction difference for flight test diagnostics and planning; (3) combine data from various sources to update and improve the model prediction; and (4) design advanced, reliable control strategies for developing high-performance aircrafts and vehicles. The success in the proposed research will markedly enhance the flight test analysis, assessment, and decision-making capabilities of USAF. Among non-DoD applications, the proposed software will find use in various engineering sectors, including NASA, automobile, combustion, power and aerospace industry, chemical plants, biomedical, among others. The product would directly contribute to these vital areas by enabling accurate, real-time, parametric analysis and prediction capabilities, which can be used in (1) efficient simulation for concept evaluation and optimized design with dramatically shorter turnaround time; (2) on-field system diagnostics and troubleshooting, (3) sensitivity analysis and correlation identification of system parameters; and (4) proper design and arrangement of test procedures for rational use of instruments and facilities.