Modelling of double-K fracture approach in concrete fracture by artificial neural networks

A cracked building can only be analyzed the best way utilizing the principals of fracture mechanics. To analyze a concrete a structure according to fracture mechanics, fracture parameters of cementitious material must be determined. Experimental studies have shown that fracture parameters of concrete are particularly influenced by the four material parameters compressive strenght, maximum aggregate size, water-cement ratio and aggregate type. Many non-linear fracture models have been proposed by design codes and investigators to determine fracture parameters of concrete. To characterize failure of concrete structures, the Double-K Model needs two fracture parameters: the unstable stress intensity factor K_IC^unand the initiation stress intensity factor K_Ic^ini. The most important difference from the other models of fracture method, on the other fracture models, taking into account not only the development of the sudden crack of concrete structures modeled with parameters, in addition to double-K method is that it takes into account the criteria of the crack initiation to spread. The main benefit of using an Artificial Neural Network (ANN) approach is that the network is built directly on experimental data by using the self-organizing capabilities of the ANN. The presented fracture model was developed by utilising 193 noisy test data taken from the literature, which were obtained via different test methods in different laboratories. In the study of the concrete material parameters; aggregate type, maximum aggregate size (dmax), compressive strength of concrete (f_c), water-cement ratio (w/c) geometric parameters of material; the initial crack length (a_0) and the effective crack length (a_e) and the initiation stress intensity factor K_Ic^ini in double-K model is to establish a relationship based on artificial neural networks. The results of an ANN-based ECM look viable and very promising. Key words: Concrete, Fracture Mechanics, Double-K Model, Artificial Neural Networks