Towards a diffusion model of acoustic energy flow in large undamped structures

It is shown that energy transport parameters for vibrations in a complex structure may be extracted from the kind of data that might be obtained from limited and cost effective direct numerical simulations (DNS). A diffusion model is proposed which promises to avoid the requirement of statistical energy analysis (SEA)-like substructuring. The model is successfully fit to simulated data with an underlying nature which is unambiguously diffusive, as opposed to DNS data where the underlying nature is in question. The data is constructed with noise that mimics the sorts of fluctuations expected in practice. The algorithm's robustness depends heavily on the number of adjustable parameters.