WorksMathematics
Time-independent equivalent matrix representation for time-domain boundary element analysis of large-scale 3D acoustic problems
Abstract
This paper proposes a time-independent equivalent matrix representation for large-scale three-dimensional (3D) acoustic analysis with the time-domain boundary element method (TDBEM). In conventional TDBEM, pair-dependent retarded times couple spatial interactions to historical boundary data, causing substantial storage and history-convolution costs. The proposed representation decouples compressible spatial interactions from propagation history, allowing reuse of the same compressed spatial representation across propagation delays and throughout time marching without additional temporal interpolation of the kernel. For conventional space-time discretizations, blockwise low-rank compression yields the compressed equivalent matrix method in the -matrix format (CEM-), with dynamic cluster-pair tracking to skip inactive interactions during wavefront propagation. For low-frequency, long-wavelength problems with a bounded number of nonzero time-lag matrices, the framework is extended to the -matrix format (CEM-) using delay-consistent partitioning, shared coupling matrices, and nested cluster bases. Under sparsity and bounded-parameter assumptions, storage and per-step cost of CEM- both scale as , excluding preprocessing. Numerical results agree well with conventional TDBEM and semi-analytical reference solutions. For CEM-, the storage requirement for a sphere with is reduced from approximately 349.0 GB to 1.6 GB, with a speedup of approximately . For CEM-, simulations with up to are performed in the low-frequency, long-wavelength regime. A million-degree-of-freedom underwater-vehicle model is also analyzed. These results demonstrate the capability of the proposed framework for large-scale 3D transient acoustic analysis using TDBEM.