Enhanced Long-Time prediction of Gas-Solid flows via Integrating Reduced-Order model with Cluster-Based network model

Abstract

Understanding gas–solid flows is crucial due to their extensive industrial processes. Integrating data-driven models into the analysis has been increasingly recognized for its ability to reduce computational costs while decoding the intricate flow behaviors. In this work, a coupling approach using Singular Value Decomposition (SVD) and Cluster-based Network Model (CNM) was developed, where SVD is employed to extract and decompose the essential information into key modes then CNM is performed for exploring the spatiotemporal correlations between modes, ultimately achieving robust long-time predictions of the system. Applied to various fluidized beds, including bubbling and spout beds, this method demonstrates stability and accuracy in predicting gas–solid flows by capturing crucial flow patterns and minimizing data redundancy. With its industrial scale running time stability, the coupling approach shows promise as a cost-effective tool for the design and optimization of real industrial processes.