Wavelet‐based data compression for flow simulation on block‐structured Cartesian mesh. (15th May 2013)

• Record Type: Journal Article
• Authors: Sakai, Ryotaro; Sasaki, Daisuke; Obayashi, Shigeru; Nakahashi, Kazuhiro
• Is Part Of: International journal for numerical methods in fluids. Volume 73:Number 5(2013:Oct.)
• Page Start: 462
• Page End: 476
• Publication Date: 2013-05-15
• Subjects: data compression -- wavelet transform -- embedded zerotree wavelet -- Cartesian mesh -- incompressible flow -- compressible flow; Fluid dynamics -- Mathematics -- Periodicals; 532
• DOI: 10.1002/fld.3808
• Languages: English
• ISSNs: 0271-2091

Abstract

SUMMARY: A data compression method based on image encoding techniques is presented for a flow simulation data set. An input flow field data set is converted into the octree structure by discrete wavelet transform, and then quantized finely or coarsely depending on its importance in the flow field. Embedded zerotree wavelet encoding as the image encoding technique and entropy encoding reduce the data size by making use of the octree structure created previously. The present compression method is incorporated in a block‐structured Cartesian mesh method called Building‐Cube method. The Building‐Cube method gives not only good performance in the flow simulation but also consistency with the embedded zerotree wavelet encoding in the data compression. Three compression cases for incompressible and compressible flow simulations, including a large‐scale simulation with O(10 8 ) mesh points, demonstrate that the present compression method gives both high compression ratios and good qualities of compressed data. Copyright © 2013 John Wiley & Sons, Ltd. Abstract : A data compression method based on image encoding techniques is presented and applied to the flow simulation data sets, including flow around a Formula 1 racing car with O( 10 8 ) mesh points. Visualization of flow field indicates compressed data reproduce velocity distribution well, compared with the original data. The data size after the compression is only 3.68% of the original data, showing good performance of the present method in terms of compressed data quality and data compression ratio.