Abstract: At present, the accuracy is very low in the prediction of fabric diagonal bending rigidity with mathematic model. Several pieces of fabric are selected and their bending rigidity in every 10 degrees from the weft-wise to warp-wise is tested, and prediction of diagonal bending of the fabrics are carried out by BP neural network based on gradient algorithm. And verification and comparison are performed. Experimental results demonstrate that the proposed system can efficiently predict fabric diagonal bending rigidity with better accuracy and offers a new evaluation method for prediction of fabric diagonal bending rigidity.