Predicting the mechanical behavior of steel frames exposed to elevated temperatures using artificial neural networks

The structural behavior fire-resistant steel is essential; fire-resistant steel is proven to have higher strength at elevated temperature than that of conventional steel. Also, the requirement of fire protection in the fire-resisting steel can be relaxed as compared with conventional steel structures. However, the design criteria for the application of the fire-resisting steel in steel columns are still limited. Experimental approach into the analysis of fire-resistant steel frames is costly and expensive. Such analyses aim at evaluating the variations of the ultimate strength of steel frames due to the reduction effects on strength resulting in the increasing temperature. An alternative approach to model the mechanical behaviour of steel frames when exposed to fire at high temperatures is presented in this work. The concept is based on a series of stress-strain curves obtained experimentally at various temperature levels. An artificial neural network (ANN) is employed to predict the stress-strain curve under such condition. The numerical results obtained from ANNs of stress levels for the material were compared with the experimental data. A New model for reduction factor is introduced and compared with other models. Using ANN was found to be an efficient tool for modelling the material properties of steel frames for high temperature applications. Key-Words: Material models, artificial neural networks, fire-resistant steel, fire load, stress strain curves, strength of materials, elevated temperature.