Abstract: Three computation models of concrete frost resistance are established on basis of optimized BP algorithm, and each of them has an input layer, hidden layer and output layer. The model(1) is named 3-5-2 type in which 3 input neurons, 5 hidden neurons and 2 output neurons are covered. The models(2) and (3) are both named 3-5-1 type in which 3 input neurons, 5 hidden neurons and 1 output neuron are covered. In models(1) the water/cement ratio, unit cement consumption and sand percentage are taken as inputs, with both loss of weight and loss of strength after thawing as outputs. As to models(2) and (3), the inputs are the same to model(1) but the outputs of model(2) and of model(3) are loss of weight and loss of strength after thawing, respectively. The results of the computation conform well to that of tests. All models developed are realized with Visual Basic 6.0. The modeling shows that optimized BP algorithms are very efficient in quantitative analysis of concrete frost resistance.