Compression of the Multi-Physics Simulation Output Data Using Principle Component Analysis

Abstract

High-fidelity multi-physics reactor core simulation includes the pin-by-pin neutronics, thermal-hydraulics, thermal-mechanics and fuel performance physical models. Every run of a high-fidelity code consumes a lot of computation time, memory and usually requires an expansive multi-core computational cluster. All output data is to be saved for further analysis if one does not want to repeat the expansive calculations. The memory amount of highfidelity output data can be too large, e.g. in the case of uncertainty analysis. Therefore, the data compression algorithms are needed for reduction of the memory without accuracy loose. In this paper we suggest such algorithm based on Principle Component Analysis (PCA) adopted for multi-physics data.