Multilayer, back-propagation artificial neural networks were used in the analysis of two problems on concrete fatigue durability. The first problem deals with prediction of fatigue number of cycles on the basis of four known characteristics of concrete and parameters of load cycles. The second problem is associated with identification of concrete strength f_c. The neural networks were trained and tested by means of more than 400 patterns from laboratory tests whose evidence was completed by Furtak [1]. Neural prediction is much more precise than prediction by empirical Formula (2.1).
展开▼