Research on multi-source data assimilation for core wall rockfill dams based on MPINN

Key Points

• The aim is to develop a method for assimilating multi-source data for comprehensive monitoring of rockfill dams using MPINN.
• Developed a multi-source data assimilation method using MPINN incorporating coupled stress-seepage theory.
• Applied the method to Twohekou core wall rockfill dam using various monitored data.
• Conducted comparative experiments to evaluate prediction accuracy and robustness of the MPINN model.
• MPINN effectively reconstructed the pore water pressure and stress fields from sparse data.
• Demonstrated superior prediction accuracy compared to other methods, enhancing robustness in forecasts.
• Validated the data assimilation approach, providing a novel pathway for comprehensive dam monitoring.