Numerical Assessment of Multi-Splitter Mufflers Using Neural Networks, the Boundary Element Method, and the Genetic Algorithm

Abstract

Recently, research on new mufflers lined with sound-absorbing material has been addressed in the industrial field. On the basis of the transfer matrix method and the stiffness matrix method, most researchers have explored noise reduction effects. Yet, the maximum noise reduction of a compact silencer equipped with sound-absorbing splitters within a constrained space, which often occurs in modern industries, has been ignored. Therefore, the optimum design of mufflers becomes essential. In this paper, a one-chamber muffler equipped with multiple sound-absorbing panels within a fixed length is assessed. In order to facilitate the assessment of optimal mufflers having multiple sound-absorbing splitters, an approximated simplified objective function (OBJ) is established in advance by linking the boundary element model (BEM) with a polynomial neural network fitted with a series of real data, input design data (muffler dimensions) and output data obtained by BEM simulation. To assess the optimal mufflers, a genetic algorithm (GA) is applied. Before the GA operation can be carried out, the accuracy of the mathematical models must be checked using the experimental data. On the basis of the fixed total thickness of the splitters, the open area of the flowing channel can be assured. Therefore, not only the influence of the backpressure can be minimized, but also the cost of the sound absorbing splitters can be economically saved. Optimal results reveal that the maximum value of the sound transmission loss (STL) can be improved at the targeted frequencies. Consequently, the optimum algorithm proposed in this study provides an efficient way to find a better silencer for industry.