The computational cost of industrial-scale models can cause problems when performing sampling-based reliability analysis. This is due to the fact that the failure modes of engineering systems typically occupy a small region of the performance space and thus require relatively large sample sizes to accurately estimate their characteristics.

This talk explores two methods for reducing the cost of reliability analysis whilst preserving the accuracy of estimated quantities. The first approach, based on Markov chain Monte Carlo sampling, can be used when several thousands of code evaluations are available. The second method, built on the ideas of Gaussian process-based optimisation, lowers this requirement from tens to hundreds of evaluations.

About the Presenter

Dr Peter Hristov is the PDRA on DATA-CENTRIC, an EPSRC fellowship project which aims to develop transparent and accountable computational engineering models. He holds a PhD in computational engineering and uncertainty quantification, and a bachelor’s degree in aerospace engineering from the University of Liverpool. Dr. Hristov is currently affiliated with the Institute for Risk and Uncertainty at the University of Liverpool. His research interests lie in developing uncertainty-aware models for aerospace and structural applications. He’s actively engaging in bridging the gap between academia and industry.

Optimisation Community

This event was hosted by the NAFEMS Optimisation Working Group (OWG). The OWG has formed an online Community to help disseminate best practice and encourage the adoption of optimisation methods and technology. You can discuss this and other presentations on the Optimisation Community Forum. For more information and to get involved go to the Optimisation Community webpage.