Advanced artificial intelligence network to forecast the fire characteristics of concrete-filled steel columns

Composite constructions, such as concrete-filled steel columns (CFSC), have gained popularity worldwide as a result of efforts to increase effectiveness. A thorough understanding of CFSC fire performance is required. Artificial intelligence (AI) was used in this study to assess the efficiency of the key factors affecting the fire characteristics of CFSC. The created AI models, including deep neural networks (DNNs), convolutional neural networks (CNNs), and Backpropagation-Particle Swarm Optimization (BP-PSO), provided relationships for predicting the fire resistance and residual strength of CFSC after fire exposure. The literature was mined for over 200 experimental data points, which were then utilized to train, validate, and test the AI models. To ascertain the fire resistance rate and residual strength index of CFSC, the experimental findings were also contrasted with predictions from already-in-use models, and brand-new, extremely accurate models were created. Deep neural networks, in particular, used AIs that demonstrated a highly accurate prediction. Additionally, the suggested models might be employed as extremely effective and precise instruments to ascertain the fire properties of CFSC due to their great agreement with actual data (R²>0.94).