Prediction model for load effective distribution width of slab in composite box girders using gene expression programming

•Four loading tests on slabs in composite box girders were carried out and elaborate finite element models were developed.•The elastic load effective distribution width is recommended for the serviceability limit state and an elastoplastic load effective distribution width is proposed for the ultimate limit state.•A comprehensive parametric study was conducted to obtain the main influencing parameters and the database for the GEP model.•Formulas for the elastic and elastoplastic effective distribution widths are proposed by using GEP approach and show superior performance compared to the models in design codes. To determine the loading condition of the deck slab in girder bridges, the concept of the load effective distribution width is widely accepted in practical design. However, effective models for calculating the load effective distribution width of deck slab in steel–concrete composite box girders (SCCBG) are still lacking, which may bring about significant errors in slab design. In this study, a rational model to calculate the load effective distribution width of deck slabs in SCCBG in both the elastic and elastoplastic stages is proposed using the gene expression programming (GEP). First, four loading tests were conducted and elaborate finite element (FE) models were developed to investigate the load distribution of the slab in SCCBG. The FE models were validated by the test results and then applied to conduct a comprehensive parametric study to obtain the influences of the design parameters and the database for the GEP model. On this basis, formulas forthe load effective distributionwidth in the elastic and elastoplastic stages were proposed by using the GEP algorithm. Furthermore, extra testing data were applied to verify the validity of the proposed models. By comparing the results given by several bridge codes, it is demonstrated that the proposed models can provide more accurate and reliable predictions and are quite simple and useful for pre-design applications.