The establishment of early warning model of steel industry based on BP neural network is discussed in this paper. The topology of the network chooses three layers of BP network structure. Hidden layer nodes selects Sigmoid function as the activation function, and the output layer select purelin function as the activation function. Error function and capacity utilization composite index are combined. Capacity utilization composite index of the next period is taken as expected value of the current issue. Initial parameters selection of the model is given. The experiment results show that the network's early warning ability is better.