For large-scale numerical simulations on supercomputers, data transfer and storage present significant efficiency and productivity issues. Therefore, the jointed hierarchical precision compression number-data format (JHPCN-DF) technique was proposed for efficient visualization and analysis of plasma particle-in-cell simulation data. It is also available for lossless and lossy compression with user-defined errors. We implement a lossy compression method of JHPCN-DF in finite element code and evaluate the compression effectiveness and compression data accuracy in linear static and dynamic structural analyses. Our technique achieves the required accuracy, even for dynamic problems, and provides a significant increase in compression performance for variable datasets.