Аннотация:The report examines the effects that occur when acoustic waves (the Berlage pulse) are applied
to linearly elastic solids formed by a periodic structure. For a series of numerical experiments
based on the finite element method, various types of lattice structures were modeled in the
Fidesys strength analysis package. In these lattice structures, the parameters of the undulation of the
bars forming the lattice structure, as well as the frequency of the applied pulse and the distance at
which the sound insulation level was measured, were varied. The number of points forming cells of
the lattice structure was also varied. For the calculation using the finite element method, grid
convergence is established and a formula is found that allows predicting the level of accuracy for a
given grid refinement. In the course of direct calculations, the presence of frequency filtering was
established for some of the specified variable parameters. The analysis of dependences of variable
parameters and sound insulation level is made. Also, based on a series of virtual experiments, a model
predicting the level of sound insulation is built based on combining the following machine learning
methods: Gradient Boosting, Random Forest, Gaussian Process.
This algorithm was configured using the Python programming language and the scikit-learn library.
The use of this algorithm allowed to reduce the time for calculating the sound insulation level by
several hundred times, as well as to solve the inverse problem: specifying the necessary parameters of
the lattice structure at a fixed level of sound insulation. As a development of the idea, further research
of three-dimensional lattices, layered lattices, lattices with viscoelastic filler, lattices with artificially
introduced deformations in the path of acoustic wave propagation, and variation of a larger number of
lattice parameters is proposed. It is also possible to use high-order beam spectral elements to improve
the accuracy of calculations instead of additional refinement of the finite element method grid.