Decomposition-enhanced parameterized physics-informed neural network for uncertainty-aware interior acoustic analysis

Interior acoustic analysis considering uncertain-but-bounded parameters plays a pivotal role in aircraft and automotive noise reduction design, offering essential insights for robust acoustic optimization. Traditional iteration-based uncertainty propagation methods suffer from the inherent trade-off between accuracy and computational efficiency. To address the limitation, this paper proposes a versatile framework to perform uncertainty propagation analysis at multiple locations effectively with the aid of physics-informed neural network (PINN). For basic problems with regular geometric regions, a parameterized PINN model consisting solely of a single neural network is employed for rapid computation. For complex problem with irregular geometric region, a modified parameterized PINN modeling method integrating parametric interpolation and remainder term neural network is proposed, significantly enhancing the expressive capability and thereby improving prediction accuracy. Eventually, a two-dimensional rectangular mathematical example and a three-dimensional aircraft fuselage engineering application are investigated to validate the effectiveness of the proposed framework.