Abstract: A 3-layer BP neural network is used to identify fabric defects.An unsuitability rule is proposed innovatively to optimize the neural network structure,improve the training speed and identification correctness,which is executed during the selection of neuron number in hidden layer and the choice of training methods.The designed neural network is trained with test samples,including warp-lacking,weft-lacking,double weft,loom bar,oil stain and hole,according to the unsuitability rule.Experimental results show that the neural network designed with the way proposed in this paper can satisfy the identification requirement and which possesses advantages of high identification veracity and speed.