Application of Machine Learning for Design-by-Analysis of Pressure Equipment

Abstract

Finite Element Analysis (FEA) is used extensively for design by analysis of Pressure Equipment (PE). While FEA provides the stress distribution within the PE geometry, the analyst is required to manually identify certain parameters for performing the analysis. The objective of this work is to supplement or replace the manual procedures of this analysis using Machine Learning (ML). Two distinct ML models to replace two such manual procedures in Design-by Analysis of PE have been developed. The models are trained on 605 distinct datasets obtained from stress-analysis of commonly found discontinuity region of PE. The ML models are trained to identify regions of discontinuity and predict linearized stresses which accelerates the analysis process. The results show that the ML models are sufficiently accurate to significantly supplement the analysis procedure.