Prediction of stability and deformation for seawalls using RSM-SVR method

Seawalls serve as crucial barriers against wave activity in coastal and riverine environments, but their stability and deformation are influenced by multiple factors, including surcharge loads, wave forces, seepage, localized scour, and construction effects. Accurately predicting their performance under real-world operating conditions remains a significant challenge. In this paper, the ancient masonry seawall, located on the north bank of the Qiantang River in China, is taken as a case study. First, the centrifuge experimental test was carried out to determine the ultimate bearing capacity of the ancient masonry based on the dimension of the prototype scale. Then, we developed numerous finite element models incorporating various boundary conditions, including surcharge loads, wave action, and localized scour to analyze the response of the seawall structure. Furthermore, we established a dataset covering the ultimate bearing capacity of the seawall by integrating the results of finite element models with the response surface method (RSM). Finally, an RSM-SVR (Support Vector Regressor) method was proposed to predict the stability and deformation of the seawall based on the results of a series of machine learning methods. The results show that the predicted results of the SVR algorithm presented a good agreement with the labeled values, which provides some references on the real-time evaluation and early warning of the seawall during its operating condition.