Deflection Prediction of a Rail-Cum-Road Suspension Bridge Under Multiple Operational Loads With Improved GPR and FSF

Abstract

The deformation of the main girder is an important manifestation of the overall stiffness of suspension bridges, which is essential for assessing bridge performance. Nevertheless, it is difficult to achieve satisfied prediction without fully considering the overall operational loads. To this end, this paper proposes a method to predict the deflection considering multiple operational loads using the monitoring data of a high-speed rail-cum-road suspension bridge. Initially, an improved Gaussian process regression (GPR) model utilizing Bayesian optimization was employed to predict the deformation of the main girder under the condition of nontrain loads. Furthermore, the distinct contributions of temperature, wind, and vehicle load were analyzed. Subsequently, based on the strain and deflection induced by train loads, the sum of sinusoids method was proposed to construct fitting and shape function (FSF) for predicting the main girder deformation under the influence of train loads. Ultimately, the deformation considering overall loads was obtained by adding the deformation under the nontrain and train loads, and the predicted deformation result was verified using the measured data. When compared to other state-of-the-art machine learning algorithms, namely, artificial neural network (ANN), support vector machine (SVM), and decision tree (DT), the improved GPR demonstrates the highest accuracy in predicting the deformation of the main girder under nontrain loads with R² of 0.9478. In addition, the proposed sum of sinusoids FSF method accurately predicted the deformation of the main girder caused by train loads, with R² of 0.934. The deformation of the main girder under the influence of overall loads can lay a foundation for the early warning and evaluation of the suspension bridges.