Prompt design of direction air supply using CFD-based proper orthogonal decomposition method considering locally dense sample distribution

Abstract

Achieving long-lasting and effective respiratory protection in highly polluted industrial sites with a cost of a small amount of air, particularly during mobile operations by workers, poses a challenge. Direction air supply (DAS) has been developed to address this challenge. However, the trial-and-error design of DAS parameters is computationally expensive and often leads to inaccurate optimizations. This work proposes a prompt inverse design method for DAS based on proper orthogonal decomposition (POD). The data structure of this work is compatible with Hermite interpolation. Furthermore, an innovative approach is introduced, utilizing a locally dense strategy for sample distribution to enhance optimization reliability. The accuracy of the CFD-based POD method for reconstructing concentration fields is verified by comparing the deviation and Pearson correlation coefficient of the predicted concentration with simulated data. The optimal DAS parameters for three target protection areas are determined through a traversal-screening process, under dual constraint indexes of protection factor and DAS air volume. In addition, the typical concentration structures extracted using POD and the sample saturation are discussed. The findings of this work contribute to a broader application of DAS in industrial sites.