Karlsen, J. and Rigby, S.E. (2024) The role of AI in engineering: Towards rapid inverse blast analysis. In: Proceedings of The 4th International Conference on Structural Safety Under Fire & Blast Loading (CONFAB 2024). The 4th International Conference on Structural Safety Under Fire & Blast Loading (CONFAB 2024), 09-10 Sep 2024, London, UK. ASRANet Ltd
Abstract
As artificial intelligence grows increasingly commonplace in areas of analysis that were once the exclusive domain of the engineer, it is critical to establish the scope for cooperation and competition between the two. This paper therefore evaluates their individual performance in addressing a problem of extreme practical significance: increasing the computational efficiency of inverse explosion analysis. At present, methods for determining the equivalent yield and position of an explosive lack the speed necessary to meaningfully inform life-saving, post-blast response; a consequence of their use of inefficient exhaustive ‘brute force’ algorithms. Therefore, this paper details and evaluates two alternative searching routines, one representative of AI and the other physics-informed. Aiming to retain the equivalent charge yield and position estimation accuracy of traditional methods, but with reduced computation time, the first scheme is a genetic algorithm that uses the principles of machine learning optimisation to more efficiently explore the domain. The second, trilateration, continues to employ an exhaustive search, but on a smaller domain that is strategically constrained using prior understanding of the governing physics. The schemes’ predictive accuracy and computation time are then assessed through the inverse analysis of blast wave arrival time data from six small-scale, free-field experiments. The alternative algorithms exhibit an insignificant reduction in accuracy compared to exhaustive search; the maximum increase in estimation error is 6.6 mm for position and equates to 0.21% of the true charge mass for the yield. There is, however, considerable increases in computational efficiency, with the genetic algorithm and trilateration requiring 1,600- and 109,000-times fewer computational iterations to complete, respectively. Overall, the implementation of a more intelligent searching routine within iterative inversion is proven to generate considerable reductions in computation without a significant loss in accuracy, thereby supporting the rapid analyses necessary to meaningfully inform the coordination of life-saving, post-blast response. Additionally, while artificial intelligence is seeing increasing integration within engineering, this work demonstrates the immense value that continues to be brought by a deep, physical understanding.
Metadata
Item Type: | Proceedings Paper |
---|---|
Authors/Creators: |
|
Copyright, Publisher and Additional Information: | © 2024 ASRANet Ltd. This is an author-produced version of a paper subsequently published in Proceedings of the The 4th International Conference on Structural Safety Under Fire & Blast Loading (CONFAB 2024). Uploaded with permission from the copyright holder. |
Dates: |
|
Institution: | The University of Sheffield |
Academic Units: | The University of Sheffield > Faculty of Engineering (Sheffield) > Department of Civil and Structural Engineering (Sheffield) |
Funding Information: | Funder Grant number Engineering and Physical Sciences Research Council 2786653 |
Depositing User: | Symplectic Sheffield |
Date Deposited: | 13 Sep 2024 10:40 |
Last Modified: | 14 Nov 2024 15:23 |
Status: | Published |
Publisher: | ASRANet Ltd |
Refereed: | Yes |
Related URLs: | |
Open Archives Initiative ID (OAI ID): | oai:eprints.whiterose.ac.uk:217126 |