Using artificial neural network (ANN) for modelling of load distribution along fully grouted cable bolts

Abstract The present study reports the application of a neural network for the prediction of the axial load distribution along fully grouted cable bolts based on a database of laboratory pullout tests on short samples. At first, the network was trained to predict the axial force of a series of short cables grouted with Portland cement at a specific w/c ratio that are then confined radially with a constant pressure. Another database which was used for the training is the pullout force of the cables confined within a confining medium (for example, a borehole in a rock mass or another material). This is classified as constant radial stiffness boundary condition. The back propagation training algorithm was used in the present study for the load calculation. Subsequently, a finite element model was used to model the axial load along long cable bolts based on the integration of the behaviour of short elements. This approach simplifies the determination of load along long cable bolts which are often installed in rock masses with different Young's modulus, rock mass displacement profiles and stress change histories.