Fast Vibroacoustic Optimization of Mechanical Structures Using Artificial Neural Networks

Abstract

An artificial neural network (ANN) is adjusted to make analytical approximation of objective function for a specific structural acoustic application. It is used as the replacement of the main real objective function during the optimization process. The goal of optimization is to find the best geometry modification of the considered model which is supposed to produce lower values of the radiated sound power levels. The result of this study shows that the function approximation by neural networks can reduce the duration of optimization procedure. Furthermore, the tuning of ANN internal parameter settings is a real challenge to be considered.

Enhancement of the vibro-acoustic performance of anti-tetra-chiral auxetic sandwich panels using topologically optimized local resonators

“…In another words, it is assumed here that whole radiated energy from the surface is being transformed into acoustic sound power, with no interaction between the panel and air. Due to these reasons the amount of sound power calculated from the ERP method is higher than practical values, however the trends of the vibroacoustic response are still valid [45]. The ERP method can provide accurate estimations for the vibro-acoustic evaluation of the system and be considered as an upper limit for sound power.…”

Trade-off among mechanical properties and energy consumption in multi-pass friction stir processing of Al7075 alloy employing neural network–based genetic optimization

“…ANN method is more suitable for the applications where there is no way to describe the problem with an analytical function. For example, optimization of manufacturing processes, 21–25 vibro-acoustic optimization of mechanical structures 29 and optimization of trusses design. 30…”

Generating Component Designs for an Improved NVH Performance by Using an Artificial Neural Network as an Optimization Metamodel

“…Often, the geometry optimization works directly on the FE model, which requires high computation times. In a holistic and fast development process the optimization should be based on fast calculating metamodels that represent the behavior of the FE model [13,18,[21][22][23][24][25][26]. In a previous publication, we showed polynomial metamodels to be capable of a targeted reduction in specific frequency bands of a component FRF [27].…”

“…[13]). In the automotive industry, artificial neural networks (ANN) currently establish themselves as a solution to overcome these obstacles [13,21,24,26,28,29]. In a previous publication we used an ANN to predict the FRF of an abstract component based on one geometry parameter [30].…”