Substructured optimization for a speakerphone design task

Abstract

This work explores the use of substructures and parallelized optimization techniques in order to design speakerphone casings, the same problem considered by Berggren et al. [in 10th World Congress on Structural and Multidisciplinary Optimization (2013)]. For this work, a simplified model consisting of a 1D beam model with enforced vibration at one end is used to represent the supporting structure of the speaker and microphone of the device. The thickness of the supporting structure is varied in order to reduce coupling between the speaker and microphone. Minimization of the structural vibration in the microphone region over frequencies between 300 and 3400 Hz is examined for optimization. A substructuring approach using spectral elements and Legendre polynomials for the thickness profile reduces the computation cost such that many evaluations of the model may be obtained in a reasonable time. The Python optimization library PyGMO is used to distribute optimization tasks over multiple CPUs in order to further accelerate the design process. The approach is shown to successfully optimize a thickness distribution targeting the frequencies of interest while reducing computation costs and is compared to previously published results for this example problem.