Abstract: A solid granule flowrate model was proposed by using simple dynamic recurrent neural networks. Considering dynamic recurrent neural network's shortcomings of complex structure and low convergence speed of training algorithm, a kind of recurrent neural network was adpted. whose structure is very simple. This RPE algorithm was adapted to the simple recurrent network by making improvement and complementarity, and the weight and the threshold of the network can be adjusted at the same time. The results of modeling show the speediness and the high-precision of this method.