Abstract: Using the neural network technology, a prior knowledge based BP network model of rock blasting earthquake effects in the construction of the Three Gorges project has been put forward. Due to high accuracy of forecasting this model can meet the imperative necessities in the construction. The field experiments of rock blasting at the Three Gorges temporagry ship lock foundation have been performed to obtain the training samples. The prior knowledge based neural network model and its algorithm have for the first time been formulated. The results obtained from the blasting experiments show that the new network model has a decided advantage over the standard BP model.