MULTIDISCIPLINARY ROBUST OPTIMIZATION OF EARLY-STAGE AIRCRAFT WING DESIGN UNDER UNCERTAINTY

Early aircraft design must cope with uncertainties that exist due to lack of design maturity, lack of knowledge about aerodynamic loads and constraints as well as model form uncertainties due to lack of confidence in the underlying physics simulation models. The ability to properly account for these uncertainties in early stage aircraft design optimisation becomes a key enabler for industry to properly trade optimum performance against the risks of achieving the performance and meeting the constraints (including key constraints on time to market). The project aims to build a data-driven computational framework (RDO-Air) for robust design and optimization of aircraft (A/C) structures in a multidisciplinary framework. RDO-Air would be applicable for early stage A/C design under a set of stringent design constraints imposed by considerations of weight penalty, flexibility in flight conditions, flight envelopes and risk minimization using a Bayesian inference-based optimization technique. It is important to highlight that the proposed design optimization under uncertainty would lead to ultimately lead to a multi-objective optimizing which explicitly considers the cost penalty for balancing the A/C performance with the risk of failing to meet the design loads.

The specific application area for this work is related to the optimisation of the "jig twist" in the manufacture tooling and the impacts of this in terms of the achieved flight shape, aerodynamic performance, and on the overall design loads envelope. Within this specific use case the design loads have already been "fixed" including a "design margin" on the loads envelopes to cover uncertainties, and the challenge is to define the "optimal" jig twist in presence of design uncertainties (such as wing structural stiffness) to trade the confidence of achieving the performance benefits against the risk of exceeding the current design loads envelope.

The project will involve the following:
- Simulation of aerodynamic loads and performance indicators in a high dimensional parameter space defining early stage A/C wing configuration using Airbus' multidisciplinary aircraft simulation platform
- Reduced order modelling using stochastic surrogates which to enable rapid stochastic sampling required for sensitivity analysis, Markov Chain Monte Carlo algorithms for Bayesian inference.
- Multi-objective optimization of parametrized A/C wing using Bayesian inference-based robust optimization framework which gives the probabilistic estimates of optimal wing parameters along with the confidence of meeting the target design loads.
- Assimilating the actual cost function of balancing the aerodynamic performance with the risk of re-design for failing to meet the load specs within the robust optimization framework, which would incorporate informed strategic decision making for a streamlined A/C design workflow.