A Machine Learning Enabled Multi-Fidelity Platform for the Integrated Design of Aircraft Systems

Abstract

The push toward reducing the aircraft development cycle time motivates the development of collaborative frameworks that enable the more integrated design of aircraft and their systems. The ModellIng and Simulation tools for Systems IntegratiON on Aircraft (MISSION) project aims to develop an integrated modelling and simulation framework. This paper focuses on some recent advancements in the MISSION project and presents a design framework that combines a filtering process to down-select feasible architectures, a modeling platform that simulates the power system of the aircraft, and a machine learning-based clustering and optimization module. This framework enables the designer to prioritize different designs and offers traceability on the optimal choices. In addition, it enables the integration of models at multiple levels of fidelity depending on the size of the design space and the accuracy required. It is demonstrated for the electrification of the Primary Flight Control System (PFCS) and the landing gear braking system using different electric actuation technologies. The performance of different architectures is analyzed with respect to key performance indicators (fuel burn, weight, power). The optimization process benefits from a data-driven localization step to identify sets of similar architectures. The framework demonstrates the capability of optimizing across multiple, different system architectures in an efficient way that is scalable for larger design spaces and larger dimensionality problems.

Engineering Applications of Artificial Intelligence in Mechanical Design and Optimization

“…The optimization procedure employs a data-driven localization stage to recognize files with comparable structures. The framework displays the capacity to optimize across numerous system architectures in a manner that is efficient and scalable for bigger design spaces and problems of greater sizes [82].…”

Machine Learning in Aerodynamic Shape Optimization

“…For a low-dimensional interpretable space (for example the flight conditions), the physical region could be defined by flight mechanics specialists [316], but it is intractable to manually determine the desired regions for high-dimensional geometric design space. Design space filtering has been proposed in ASO as an efficient way to exclude the abnormal regions, and similar attempts have been applied in other design problems [317,318]. This approach does not reduce the number of design variables; instead, it shrinks the design space by defining a constraint function to evaluate the abnormality of samples.…”

A Safety-Focused System Architecting Framework for the Conceptual Design of Aircraft Systems

“…Chakraborty and Mavris introduce heuristics-based checks to ensure that system architectures are defined correctly to provide meaningful results when evaluated by their integrated systems sizing and performance estimation framework [46]. Recent work by Garriga et al [47] has demonstrated the use of configurational filters in determining the feasibility of a large design space of landing gear braking and flight control system architectures. Using configurational rules and heuristics to test if system architectures are feasible (at least from a configurational point of view) can also be extended to incorporate some safety aspects.…”