Robust Design under Uncertainty

The goal of engineering design is to create technological systems that satisfy specific performance objectives and constraints over a period of time. Usually, there exist many feasible designs that satisfy the required objectives. For this reason, it is desirable to choose an optimal design according to some criterion. Modern engineering systems are inherently complex. This complexity means that endogenous (geometry, material properties) and exogenous (loads) information is never complete. This lack of information can be captured by modelling uncertainties as random variables, whose distributions can in principle be obtained from expert opinion, literature or test data. The objective of performance-based design is therefore to determine the optimal design that minimises an expected loss function which depends on both the characteristics of the design space and the model parameters that encode the characteristics of the system under study. This PhD places a strong emphasis on techniques and methods such as surrogate modelling, computational Bayesian inference and stochastic optimisation, applied to problems arising within the field of computational fluid dynamics (CFD).
The first six months will be used as an initial phase of substantial foundational work. This will include a combination of working through relevant literature, both probability related and CFD related, as well as attending lectures of probability theory and computational inference lectures. Furthermore, there was a placement at the industrial partner's site (completed before Christmas), at which I underwent training on CFD software, and began attempting to utilise this new knowledge and skillset to replicate the results of a relevant academic paper.
From months six to twelve (there may be some overlap depending on progress and other factors), I will continue working through relevant literature, at this point beginning to look into more specialised areas. Furthermore, to supplement my reading, I will also attend a mixture of seminars, workshops and conferences in the fields which hold an interest for me. I will also complete my work on replicating the results of the academic paper, and begin coding up my first stochastic optimisation (single-objective) solver, with the intention of finishing this by the end of the year.