This conference paper was submitted for presentation at the NAFEMS World Congress 2025, held in Salzburg, Austria from May 19–22, 2025.

Abstract

Surrogate models may be employed where conventional finite element analysis (FEA) will consume excessive analyst and/or computing time and capacity. Surrogate data driven models, applied with artificial intelligence (AI) and machine learning (ML) may use approximate mathematic models based on data (rather than being based on physics) which in theory require greatly reduced computing and human effort. In a finite element analysis context to employ a surrogate model, there is the need to firstly train the machine learning model with results from existing finite element analysis, then to predict results using the machine learning approach, and finally to map those results to the model to validate and further teach the tool. This approach has the potential to generate results much faster than traditional physics-based FEA models, but understanding of how and when it is an appropriate and viable approach, as an alternative to conventional finite element analysis, needs to be developed; this is an aim of the reported work. To advance the state-of-the-art of engineering surrogate modelling, the work reported here includes an assessment of both commercial and open source software solutions, effectively as a benchmarking, for addressing a structural integrity finite element analysis problem; based on predicting the attributes of a cracked pipe. The outputs are compared against best conventional FEA physics based modelling practice. The provision of sufficient training data is a concern in this type of work, and the approach taken to generate training and validation data and use of this will be discussed. This work begins to answer the fundamental questions about the future of finite element modelling and the role of finite element Analysts in engineering problem solving, the confidence which can be placed on such results, and allows comments to be made regarding future application of this surrogate approach in an engineering structural integrity assessment context.