Jürgen Branke is Professor of Operational Research and Systems at the University of Warwick, Coventry, U.K. He has been an active researcher in the area of evolutionary optimization since 1994 and has published over 180 papers in international peer-reviewed journals and conferences. His current research interests include Bayesian optimization, multiobjective optimization, handling of uncertainty in optimization, dynamically changing optimization problems, simulation.based optimization, and the design of complex systems. Prof. Branke is Editor-in-Chief of the ACM Transactions on Evolutionary Learning and Optimization, Area Editor of the Journal of Heuristics and the Journal on Multi-Criteria Decision Analysis, as well as an Associate Editor of the IEEE Transactions in Evolutionary Computation and the Evolutionary Computation Journal.
Bayesian optimisation is a relatively new but powerful technique for computationally expensive black-box optimisation. It builds a surrogate model of the fitness landscape, and then uses this model to iteratively decide which additional designs are most promising to evaluate. This talk will give a brief introduction to Bayesian optimisation and then focuses on recent developments that allow to use Bayesian optimisation for a) multi-concept design, where one has to select among different design concepts, each with different design variables, and b) multi-task problems, where solutions for several related problems have to be identified.
Pranay Seshadri is the Aeronautics Group Leader in the Data-Centric Engineering Programme at the Alan Turing Institute. He is also a Postdoctoral Fellow at the Department of Engineering in the University of Cambridge. He obtained his PhD in Computational Engineering in 2016 from the University of Cambridge.
Techniques for subspace-based dimension reduction are starting to make their way into engineering workflows. They facilitate powerful inference tailored for both design and manufacturing tasks by identifying linear combinations of parameters that are important with respect to key output quantities of interest. The central idea is to use these methods for empowering more efficient (and faster) optimisation, uncertainty quantification and sensitivity analysis studies, by drastically reducing the number of model evaluations. In this talk, I will provide an overview of such techniques with an exposition of the formulas and code. I will also provide a few industrial examples of how subspace-based dimension reduction can be used to solve salient problems in aeronautics.
Nuno Lourenço has 20 years of engineering experience spanning several industries, particularly automotive and aerospace, where he has held several leadership positions in Product Development. Nuno leads simulation for Body Engineering at Jaguar Land Rover since 2016 covering multi-disciplinary mechanical systems simulation.
Finite Element Analysis is well developed and mature in the Automotive Industry; it is widely recognised as an enabler to the product development process. The constant advances in high performance computing clusters in most Automotive OEMs mean that today’s models for crash, dynamics, vibration, noise, fatigue are very detailed and can avoid costly late engineering changes, provided there is discipline in model set-up, geometric and non-geometric data. However, as vehicle systems get more complex, there is an increased risk of simulation failing to predict the performance in future tests. This poses significant challenges for traditional use of FEA as a prediction & development toolset as the system being simulated may be prone to having multiple behaviour modes.
The author proposes that the role of FEA-based simulation needs to evolve to enable a better understanding of a system’s behaviours and be focused more on the requirements definition stage of systems engineering. To manage the systems engineering process effectively, we must combine large FEA data sets with experimental observations to derive multi-physics forecast models that can be used alongside software and electronics model-based engineering in the early product definition stages.