Fast Differentiable Modal Simulation of Non-linear Strings, Membranes, and Plates

TL;DR

This paper presents a fast, GPU-accelerated, differentiable modal simulation framework for non-linear strings, membranes, and plates, enabling efficient inverse modelling and real-time applications in acoustics and sound synthesis.

Contribution

It introduces a novel, differentiable, GPU-based modal simulation framework built with JAX, significantly improving computational efficiency and enabling gradient-based inverse modelling for complex physical systems.

Findings

Outperforms CPU and GPU implementations in speed

Enables accurate parameter recovery from data

Supports real-time sound synthesis applications

Abstract

Modal methods for simulating vibrations of strings, membranes, and plates are widely used in acoustics and physically informed audio synthesis. However, traditional implementations, particularly for non-linear models like the von K\'arm\'an plate, are computationally demanding and lack differentiability, limiting inverse modelling and real-time applications. We introduce a fast, differentiable, GPU-accelerated modal framework built with the JAX library, providing efficient simulations and enabling gradient-based inverse modelling. Benchmarks show that our approach significantly outperforms CPU and GPU-based implementations, particularly for simulations with many modes. Inverse modelling experiments demonstrate that our approach can recover physical parameters, including tension, stiffness, and geometry, from both synthetic and experimental data. Although fitting physical parameters is more sensitive to initialisation compared to other methods, it provides greater interpretability and more compact parameterisation. The code is released as open source to support future research and applications in differentiable physical modelling and sound synthesis.