Integrating supervised learning and applied computational multi-fluid dynamics. (December 2022)

• Record Type: Journal Article
• Authors: Catsoulis, Sotiris; Singh, Joel-Steven; Narayanan, Chidambaram; Lakehal, Djamel
• Is Part Of: International journal of multiphase flow. Volume 157(2022)
• Publication Date: 2022-12
• Subjects: Computational Fluid Dynamics -- Multiphase flow -- Machine learning -- Data-driven model -- Simulation-based Digital Twin; Multiphase flow -- Periodicals; Écoulement polyphasique -- Périodiques; Multiphase flow; Periodicals; 620.1064
• DOI: 10.1016/j.ijmultiphaseflow.2022.104221
• Languages: English
• ISSNs: 0301-9322

Abstract

Abstract: The transition from iterative methods of engineering design towards physics-based modeling has been assisted by the advent of Computer-Aided-Engineering. However, post-processing of simulation results, based on a standard workflow providing base-case simulations complemented by selected operating conditions, has changed little. In this work, we propose a new paradigm for handling simulation data by deploying machine learning to encompass a wide spectrum of operating conditions, bypassing the need for additional simulations. This hybrid physics-based and data-driven modeling procedure yields to what we refer to as a Simulation-based Digital Twin. In this paper, we make the case for Computational Fluid Dynamics in multiphase flow systems, although the workflow can be generalized to any other computational engineering method. We quantify the computational speed-up to conclude that the combination of these two fields generates potential for improvement on the conventional methods used in the broad area of computational engineering. Highlights: We propose a workflow to automate post-processing of CFD data by using machine learning. We quantify the superiority of machine learning techniques in providing faster forecasts. The strategy allows the simulation digital twin to be used for real-time process optimization.