A hybrid forcasting model of temperature gradient in longitudinal slab tracks on bridge-subgrade transition zones

Abstract

A hybrid particle swarm optimization and support vector machine (PSO-SVM) model was proposed to accurately forecast the temperature gradients in longitudinal slab ballastless tracks on a bridge and subgrade transition zone in this paper. Based on the monitoring experiment in climate extremes, the characteristics of meteorological factors and internal temperature were analyzed. Subsequently, the nonlinear correlations of meteorological factors and temperatures gradient on three different foundations were analyzed by five copula functions. Moreover, two theoretical temperature gradients’ thresholds were deduced, the SVM and PSO-SVM model were then established to forecast the temperature gradients, respectively. Results show that the track slab on transition zone has the largest temperature gradient among three foundations. The Frank copula function has the best effect in reflecting the joint distribution between meteorological factors and temperatures gradient. Moreover, the forecasting effect based on the displacement threshold requirement is better than that the axial force threshold requirement. The PSO-SVM model with the Radial Basis Function kernel has the best forecasting effect with the accuracy greater than 93 %.