Experimental and statistical investigation on temperature gradient of CFST truss chords

Abstract

Due to mutual shading occlusion among staggered components, the solar temperature fields of CFST truss are complex, and the temperature actions of the truss chords differ from that of single-tube members, posing challenges in the calculation and assessment of thermal effects. This study proposed a temperature action mode tailored to independent thermal effects of CFST truss. Experimental testing over 1 year was conducted on a CFST truss specimen, with temperature gradients of chords analyzed using an extreme value statistical method and various machine learning algorithms. Structural parameters affecting temperature gradients were also examined through numerical simulation. The findings were applied to assess thermal responses in a real bridge scenario. Significant differences in the temperature gradients were observed between upper and lower chords, with the upper chords showing larger vertical temperature difference and the lower chords exhibiting slightly larger lateral temperature difference. A linear relationship in the temperature actions across different chords was notable. Bending effects induced by temperature gradients in chords may equal or surpass those caused by vehicle loads, necessitating careful consideration in CFST bridge design. The proposed extreme value statistical method for bridge temperatures, based on fourth-order linear moments, effectively determined optimal thresholds and probability distributions, demonstrating excellent performance when applied to tested data. Advanced machine leaning-based predictive models for bridge temperatures offer enhanced flexibility and accuracy compared to traditional predictive formulas.