Physics Informed Neural Networks for Dynamic Load Reconstruction for Plant Piping

This study aims to offer a method for calculating the dynamic loads on a vibrating pipework in a process plant based on the Theory of Inverse Problems (IP) and physics-informed neural networks (PINN). Vibrating pipes present a significant risk of fatigue failure, which can potentially cause catastrophic damage. However, the lack of quantitative information on applied loads precludes the conventional design of the system. Mathematically, the governing partial differential equation (PDE) is ill-posed. In this scenario, a data-driven strategy based on neural networks was studied in an inverse theoretical framework was studied. Deep Neural Networks (DNNs) were used to model the forcing function, and the PDE was solved in the time domain for the displacement response. The target data were the displacement readings of the in-situ vibration. This problem is reduced to an optimization problem that minimizes the errors. Two cases, one with single loading and the other with dual loading, were presented to validate the theory. These results clearly demonstrate the effectiveness of the proposed method. The widespread use of artificial intelligence (AI) and machine learning (ML) in multidisciplinary engineering domains is the main motivating factor. This study is significant from the perspective of monitoring industrial equipment conditions.