Abstract: The nonlinear behavior and stochastic features for concrete in an anchorage zone or called an end block subject to local compression loading and under complex stress state are briefly reviewed．Results of experimental study for4groups of specimens are introduced．Parameters‚such as the stiffness of the end plates‚the ratio of the compres-sion areas‚and the reinforcement effects are considered．Theses results are used as learning samples for training an artificial neural network（ANN）．By selecting network architecture with sufficient training and testing‚an ANN model with necessary accuracy for predicting the bearing strength of concrete is obtained．An application example shows that ANN can be used as one of approaches for predicting bearing strength of concrete subject to local compression．

Key words: confined concrete, bearing local strength, artificial neural network, prediction