A hybrid modeling framework using dimensional analysis for erosion predictions

The transport of solids in multiphase flows is common practice in energy industries. Safe and efficient operation requires reliable estimates of erosion rates in the pipelines. Numerous semi-mechanistic models exist for quantifying erosion rate. However, most models have large discrepancies in their predictions. This paper introduces a hybrid modeling framework to estimate erosion rates based on different flow regimes and pipeline orientations. The framework combines a semi-mechanistic model with Gaussian process modeling and dimensional analysis. Dimensional analysis and feature selection are utilized to identify significant dimensionless numbers for the erosion process. These numbers are used as inputs for the Gaussian process model and to suggest possible ways to improve the semi-mechanistic model. The hybrid modeling approach reduces the root mean square error by 36% compared to the semi-mechanistic model. The results suggest that a better multiphase flow model may improve the erosion rate predictions, especially in annular/mist flow boundary regions. (c) 2021 Elsevier Ltd. All rights reserved.