RESUMO
Reconstructing transient states presents significant challenges, particularly within complex pipe networks. These challenges arise due to nonlinear behaviours, inherent uncertainties in the system, and limitations in data availability. This work proposed a novel approach employing Physics-Informed Neural Networks (PINN) to reconstruct transient states in pipe networks, even with limited sensor data. To integrate the complex topology of pipe network systems into neural networks, the method integrates the PINN framework with an efficient elastic water column (EWC) model which can be simply formulated across diverse pipe network configurations. The results showed the proposed PINN method can accurately reconstruct the pressure and flow variation at unmonitored locations, even provided with noisy data at a limited number of locations. One of its advantages lies in its ability to effectively capture extreme values that hold potential significance for pipe infrastructure, providing a promising avenue for pipe failure analysis and enhanced safety management. Laboratory experiments have also been conducted to validate the efficacy and reliability of this method, thus further underlining its potential for real-world applications.
Assuntos
Redes Neurais de Computação , Pressão , Modelos Teóricos , Abastecimento de ÁguaRESUMO
In water pipeline systems, monitoring and predicting hydraulic transient events are important to ensure the proper operation of pressure control devices (e.g., pressure reducing valves) and prevent potential damages to the network infrastructure. Simulating transient pressures using traditional numerical methods, however, require a complete model with known boundary and initial conditions, which is rarely able to obtain in a real system. This paper proposes a new physics-based and data-driven method for targeted transient pressure reconstruction without the need of having a complete pipe system model. The new method formulates a physics-informed neural network (PINN) by incorporating both measured data and physical laws of the transient flow in the training process. This enables the PINN to learn and explore hidden information of the hydraulic transient (e.g., boundary conditions and wave damping characteristics) that is embedded in the measured data. The trained PINN can then be used to predict transient pressures at any location of the pipeline. Results from two numerical and one experimental case studies showed a high accuracy of the pressure reconstruction using the proposed approach. In addition, a series of sensitivity analyses have been conducted to determine the optimal hyperparameters in the PINN and to understand the effects of the sensor configuration on the model performance.