Prediction of stratified ground consolidation via a physics‐informed neural network utilizing short‐term excess pore water pressure monitoring data
Weibing Gong,Linlong Zuo,Lin Li,Hui Wang
DOI: https://doi.org/10.1111/mice.13326
IF: 10.066
2024-08-19
Computer-Aided Civil and Infrastructure Engineering
Abstract:Predicting stratified ground consolidation effectively remains a challenge in geotechnical engineering, especially when it comes to quickly and dependably determining the coefficient of consolidation (cv ) for each soil layer. This difficulty primarily stems from the time‐intensive nature of the consolidation process and the challenges in efficiently simulating this process in laboratory settings and using numerical methods. Nevertheless, the consolidation of stratified ground is crucial because it governs ground settlement, affecting the safety and serviceability of structures situated on or in such ground. In this study, an innovative method utilizing a physics‐informed neural network (PINN) is introduced to predict stratified ground consolidation, relying solely on short‐term excess pore water pressure (PWP) data collected by monitoring sensors. The proposed PINN framework identifies cv from the limited PWP data set and subsequently utilizes the identified cv to predict the long‐term consolidation process of stratified ground. The efficacy of the method is demonstrated through its application to a case study involving two‐layer ground consolidation, with comparisons made to an existing PINN method and a laboratory consolidation test. The results of the case study demonstrate the applicability of the proposed PINN method to both forward and inverse consolidation problems. Specifically, the method accurately predicts the long‐term dissipation of excess PWP when cv is known (i.e., the forward problem). It successfully identifies the unknown cv with only 0.05‐year monitoring data comprising 10 data points and predicts the dissipation of excess PWP at 1‐year, 10‐year, 15‐year, and even up to 30‐year intervals using the identified cv (i.e., the inverse problem). Moreover, the investigation into optimal PWP monitoring sensor layouts reveals that installing sensors in areas with significant variations in excess PWP enhances the prediction accuracy of the proposed PINN method. The results underscore the potential of leveraging PINNs in conjunction with PWP monitoring sensors to effectively predict stratified ground consolidation.
transportation science & technology,construction & building technology,computer science, interdisciplinary applications,engineering, civil