How our researchers achieve more accurate predictions in simulations
Industrial and engineering applications make a lot of use of simulation and modelling. They are increasingly complicated, because they require higher and higher fidelity to be accurate and useful.
Multi-scale problems are one of the greatest challenges in computational engineering when, for example, you are trying to understand how a large system like a turbine or pump is affected by small-scale phenomena such as molecular interactions.
Replacing physical experiments with non-destructive virtual testing is useful - but you don't want to wait forever for your simulation to complete. By using modern mathematical tools and high-performance computing, our researchers can now incorporate elements of uncertainty in simulations to account for lack of knowledge about input parameters, variability in operating conditions, or inappropriate modelling assumptions.
This is uncertainty quantification (UQ). It aims to provide more accurate predictions about the behaviour of systems. Industry is already very interested in this, because it answers key questions about the control of manufacturing processes, what constitutes normal operation, and the effect of uncertainties in trouble-shooting
Talk to our Physical Science & Engineering researchers about their work, and how we're developing rigorous systematic mathematical methods in data analytics, modelling, and simulation. Errors in computational engineering arise from things like numerical schemes, boundary conditions, and measurements. Our focus is on non-intrusive approaches that enable use of existing codes and straightforward parallel evaluation for quantities of interest.