Neural networks speed up vibration analysis of nonlinear mechanical parts
AFT Neural Function Approximators for 1D Nonlinear Force Laws
Computational Engineering, Finance, and ScienceMachine Learning
Summary
When machines have parts that don't move smoothly, like springs that behave oddly or surfaces that rub together, it can be hard to predict how they will vibrate. The authors found a way to use neural networks to learn how these tricky parts behave without needing slow, repetitive calculations. This method focuses on individual parts and can handle many types of nonlinear behaviors more efficiently. It can help engineers analyze complex machines faster by replacing a time-consuming step with a learned shortcut.
What this means in practice
- •For mechanical engineers: Analyze vibration behavior of assembled structures with nonlinear parts faster using neural network surrogates.
- •For automotive design teams: Improve simulation speed of vehicle suspension systems featuring nonlinear springs and friction elements with learned force models.
Authors
Miriam Goldack, Johann Groß, Malte Krack, Merten Stender
Abstract
Nonlinear contacts and friction strongly influence the vibration response of assembled structures, but their accurate numerical treatment is computationally demanding. The harmonic balance method is widely used to compute periodic steady-state responses, yet the required alternating frequency-time scheme becomes costly for nonsmooth and hysteretic nonlinearities and must be repeated throughout the nonlinear solution process. Here we show that this procedure can be replaced by neural networks that directly map displacement Fourier coefficients to nonlinear force coefficients and provide the corresponding Jacobian through automatic differentiation. The surrounding solver and continuation algorithms remain unchanged for the computation of frequency response curves. The neural networks exclusively learn individual nonlinear elements rather than complete system responses. Physics-based nondimensionalization and phase normalization facilitate the learning process and enable a single trained network to cover a wide range of parameter combinations. Building on the cubic spring, unilateral spring, and Jenkins elements considered here, the approach points toward a reusable library of nonlinear-element surrogates that can be combined in arbitrary number and location within a mechanical system. By bypassing the iterative force evaluation in time domain, the method offers favorable computational scaling for high-resolution analyses and systems with many nonlinear elements.