Recently, research on new techniques for an economical muffler, which is hybridized with a single-connected curved tube, has been applied in the industrial field. Most researchers have explored noise reduction effects based on the transfer matrix method and the stiffness matrix method. However, the maximum noise reduction of a silencer within a constrained space, which frequently occurs in engineering problems, has been neglected. Therefore, the optimum design of mufflers becomes an essential issue. In this paper the shape optimization of a one-connected tube muffler with a fixed length is examined. In order to speed up the optimization assessment, a simplified objective function (OBJ) is established by linking the boundary element model (BEM― developed by the commercialized software, SYSNOISE) with neural network model (NNM). Instead of a complicated mathematical model (BEM), a polynomial neural network, which is constructed using a neural network fitted with a series of real data ― input design data (muffle dimensions) and output data approximated by BEM data in advance. To assess the optimal mufflers, a genetic algorithm (GA) is applied. Moreover, the numerical cases of sound elimination with respect to various parameter sets and pure tones (350, 500, and 650) have been exemplified and discussed. Before the GA operation can be carried out, the approximation between BEM and experimental data is checked. In addition, both the BEM and NNM are compared. It is found that for a one-connected curved tube muffler the BEM and the experimental data are in agreement. Moreover, the BEM and NNM conform. Optimal results reveal that the maximum value of the sound transmission loss (STL) can be improved at the desired frequencies. Consequently, the optimum algorithm proposed in this study can provide an efficient way to develop an optimal muffler hybridized with a one-connected curved tube for industry.

Included in

Engineering Commons