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

Abstract

In modern vehicle development, suspension components have to meet many boundary conditions. In noise, vibration, and harshness (NVH) development these are for example eigenfrequencies and frequency response function (FRF) amplitudes. Component geometry parameters, for example kinematic hard points, often affect multiple of these targets in a non intuitive way. In this article, we present a practical approach to find optimized parameters for a component design, which fulfills an FRF target curve. By morphing an initial component finite element model we create training data for an artificial neural network (ANN) which predicts FRFs from geometry parameter input. Then the ANN serves as a metamodel for an evolutionary algorithm optimizer which identifies fitting geometry parameter sets, meeting an FRF target curve. The methodology enables a component design which considers an FRF as a component target. In multiple simulation examples we demonstrate the capability of identifying component designs modifying specific eigenfrequency or amplitude features of the FRFs.

A review of machine learning methods applied to structural dynamics and vibroacoustic

“…An NN-based surrogate was used instead of costly FEM simulations in [285] to perform parametric optimization of vehicle suspension hardpoints to minimize structure-borne road noise. The optimization approach used a criterion combining an up-limit and a matching target to the FRF curve to control amplitudes at specific frequencies and the frequency shift.…”

A Review of Machine Learning Methods Applied to Structural Dynamics and Vibroacoustic

“…Parametric optimization of the kinematic hardpoints of a vehicle suspension aiming to decrease road noise was performed through an NN surrogate model replacing costly FE analysis [173]. The optimization approach combines criteria that set an FRF curve as an up-limit and as a matching target, aiming to control both amplitudes in a specific frequency and frequency shift.…”

Performance Evaluation of Machine Learning Algorithms for Regression of Finite Element Analysis (Fea) Simulation Design Data

“…Typically, the development process involves successive iterations of design creation, simulation, and analysis, requiring significant efforts from the engineer and computational resources until optimal metrics are achieved with parameters within the constraints [1]. As a result, some methods have emerged to expedite and facilitate this process, among which Finite Element Analysis (FEA) combined with an optimization algorithm stands out for its higher accuracy [2,3].…”

“…As for DOE, it is a statistical application used to design experiments and analyze the results to identify relationships between design variables and responses. Through multi-objective optimization by creating metamodels from DOE, it is possible to obtain optimal results while analyzing the sensitivity between parameters and responses of the Finite Element (FE) model [3,5].…”

“…However, they have shown limitations in handling more complex systems with large amounts of data, parameters, and nonlinear relationships. In this context, Machine Learning (ML) comes into play to overcome these limitations, as it demonstrates better performance due to its adaptability features [3,6]. Therefore, this study aims to evaluate the performance of conventional ML techniques as metamodels for design variables with more than one objective response.…”