Using Machine Learngin for Early-Stage Aircraft Wing Design

早期飞机结构设计必须探索较大的设计空间，以保证最佳的气动性能和承载性能。当我们进入一个具有增强功能的新型航空结构（如变形翼）的时代时，这一点尤为重要。在这个阶段（与随后的详细设计阶段相反），设计师还必须应对由于缺乏设计成熟度、空气动力载荷知识和模型预测而存在的不确定性。该博士项目旨在与空客合作，利用机器学习开发数据驱动的强大飞机机翼设计工具箱。由于机器学习能够在无数可能性和协变量的情况下预测最佳设计/操作条件，因此机器学习在制造和设计中迅速普及。机器学习将成为项目在实现性能和满足约束的风险之间适当交易最佳性能的关键推动者。贝叶斯机器学习方法将在数据驱动的框架内用于飞机结构的稳健设计和优化，在考虑重量惩罚、飞行条件灵活性、飞行包线和风险最小化的严格设计约束下

Early-stage aircraft structural design must explore a large design space to ensure optimal aerodynamic and load bearing performance. This is especially important as we enter an age of novel aerostructures with enhanced capabilities such as morphing wings. At this stage (as opposed to the subsequent detailed design stage), designers must also cope with uncertainties that exist due to lack of design maturity, knowledge about aerodynamic loads and model predictions. The PhD project aims to use machine learning to develop a data-driven robust aircraft wing design toolbox in collaboration with Airbus. Machine learning is seeing a rapid uptake in manufacture and design due to its ability to predict optimal design/operational conditions given countless possibilities and covariates. Machine learning will be a key enabler for the project to properly trade optimum performance against the risks of achieving the performance and meeting the constraints. A Bayesian machine learning approach will be used within a data-driven framework for robust design and optimization of aircraft structures under a set of stringent design constraints imposed by considerations of weight penalty, flexibility in flight conditions, flight envelopes and risk minimization