The prediction of fire performance of concrete-filled steel tubes (CFST) using artificial neural network

Search for enhancing the efficiency has led to composite structures such as concrete-filled steel tubes (CFST) with increasing applications across the world. The fire performance of CFST columns needs to be understood comprehensively. In this paper, the effectiveness of the most important parameters on fire resistance rating (FRR) and residual strength index (RSI) was evaluated using artificial neural network (ANN). Relationships were derived from the developed ANN model for predicting the FRR and RSI of CFST columns. Near 300 experimental data points were extracted from the literature and were used for training, validating, and testing the ANN models. The correlation coefficient (R) of the network for FRR and RSI is 0.967 and 0.97, respectively. The derived relationships from the ANN model have the R coefficient of 0.61 and 0.74 for FRR and RSI, respectively. Also, the limitations of existing empirical relationships were compared with the derived relationships. The comparative results indicated that the ANN model was more stable and accurate than the existing relationships. Moreover, a graphical user interface was developed to predict the FRR and RSI of CFST columns. The derived relationships can be used for the design of new CFST columns, retrofit the existing ones, and risk assessment in fires.

Automated summary

This paper evaluates the impact of key parameters on the fire performance of CFST columns using artificial neural networks and derives predictive relationships for FRR and RSI. The developed ANN models show higher stability and accuracy compared to existing empirical relationships, with correlation coefficients of 0.967 and 0.97 for FRR and RSI, respectively. The derived relationships can be used for designing new CFST columns, retrofitting existing ones, and conducting risk assessments in fires.