Abstract
The ability to reasonably predict the response of steel structures under fire effects is of great importance in structural fire safety design. This paper presents neural networks prediction of axial load capacity for steel columns in fire. An algorithm of back propagation neural network with the log-sigmoid activation function is adopted because of its precision and results enhancement of foretelling. The legitimacy of the technique is tried by contrasting and distributed test information on steel columns at surrounding and elevated heat. The examinations demonstrate such technique gives great correlation with test result. Parametric studies have been done to evaluate the impacts of cross sectional shape, slenderness ratios and eccentricity of loading on the carrying capacity of steel columns under fire. The slim sections of steel columns with slenderness ratio domain (100-140) react distinctively by showing an abundantly decreased rate of loss in strength within the temperature domain (20°C - 300°C). This domain diminishes further with expanding slenderness ratios, and for middle columns with slenderness ratio domain (40-80), is like that of stumpy columns however at decreased buckling stress. Be that as it may, in this scope of (L/R) ratios the lessening in stress with expanding temperature is regular and demonstrates no sudden drop, because of the collaboration amongst buckling and yielding. On other hand, the eccentricity of loading on the carrying capacity of steel columns under fire shows that the slender column, (slenderness ratio) greater than 120, the column demonstrates a diminishing impact of used eccentricity of loadings with expanding slenderness ratios. This might be as a consequence of more impelled thermal bowing that is straightforwardly relative to the column length. And the load-eccentricity characteristics of the intermediate column, (slenderness ratio) domain (20 – 60), are schemed at increasing temperature gradient. It is fascinating to observe that the eccentricity of the limit of maximum column load capacity slightly effected with temperature gradient. It is trusted that the important data gave in this work will be helpful in giving a superior comprehension on the genuine behavior of steel sections in fire and a great step in improving the method of design.
