Data-driven prediction method for flexural performance of ECC composite sandwich panels

Abstract

To address the inadequate load-carrying capacity and ductility of precast concrete sandwich panels (PCSPs), as well as their inability to meet the current trend of lightweight and high-strength structures, this study proposes a novel engineered cementitious composite (ECC) composite sandwich panel (ESP) by applying ECC to the outer wythe. Existing design methods are unable to accurately quantify either the composite action between the inner and outer wythes or fully consider the complex design parameters when predicting the flexural performance of this new type of sandwich panel. In response, two normal concrete sandwich panels (NSPs) and two ESPs were first tested to reveal their flexural performance using a four-point bending test. Test results indicate that the application of ECC to the outer wythe can significantly improve the flexural performance of the sandwich panel, especially in terms of crack resistance and ductility. Subsequently, finite element parametric analysis demonstrated the impact of connector arrangement on the flexural performance, and design formulas for the degree of composite action (DCA) based on connector arrangement were proposed. Finally, a data-driven prediction model integrating experiments, finite element analysis, and neural networks was developed to accurately predict the flexural carrying capacity and ductility of the sandwich panel. By utilizing gene expression programming, the explicit function can be derived for the prediction model, which provides practical guidance for engineering applications.