Using deep learning to explore local physical similarity for global-scale bridging in thermal-hydraulic simulation. (November 2020)

• Record Type: Journal Article
• Authors: Bao, Han; Dinh, Nam; Lin, Linyu; Youngblood, Robert; Lane, Jeffrey; Zhang, Hongbin
• Is Part Of: Annals of nuclear energy. Volume 147(2020)
• Publication Date: 2020-11
• Subjects: Deep learning -- Thermal–hydraulic simulation -- Global-scale bridging -- Local similarity; Nuclear energy -- Periodicals; Nuclear engineering -- Periodicals; 621.4805
• DOI: 10.1016/j.anucene.2020.107684
• Languages: English
• ISSNs: 0306-4549

Abstract

Highlights: Feature Similarity Measurement (FSM) is developed for bridging global scale gaps. FSM provides good predictions of simulation error for different extrapolative conditions. Prediction accuracy of FSM increases with a higher data similarity between training and target data. Abstract: Current system thermal–hydraulic codes have limited credibility in simulating real plant conditions, especially when the geometry and boundary conditions are extrapolated beyond the range of test facilities. Because mesh size is one of the model parameters for these coarse-mesh codes with simplified boundary-layer treatments, the mesh-induced error and model error are tightly connected, which makes it difficult to evaluate mesh effect or model scalability independently, as in classical scaling analysis. This paper proposes a data-driven approach, Feature-Similarity Measurement (FSM), to establish a technical basis to overcome these difficulties by exploring local patterns using machine learning. The underlying local patterns in multiscale data are represented by a set of physical features that embody the information from a physical system of interest, empirical correlations, and the effect of mesh size. After performing a limited number of high-fidelity numerical simulations and a sufficient amount of fast-running coarse-mesh simulations, an error database is built, and deep learning is applied to construct and explore the relationship between the local physical features and simulation errors. As a result, a data-driven model can be developed to provide an accurate estimate on the simulation error even when global-scale gaps exist. Case studies based on mixed convection have been designed for demonstrating the capability of data-driven models in bridging global-scale gaps.