Abstract: Based on measured data of steel plate rolled by single stand 2500 four high reversing mill, the prediction model for front end bending behavior has been established by artificial neural network method. Results showed that artificial neural network calculated value of plate front end bending during rolling conformed to measured value ； as temperature difference between top and bottom surface of rolling work piece was higher, the temperature difference, deformation zone shape characteristic and deformation extent were main factors to effect rolling work piece front end bending. For heavy plate products, decreasing pass reduction could decrease bending, and for thin plate, increasing pass reduction could decrease bending.