Abstract: In the steel strip cold rolling process, the radial deformation of roll system will affect the thickness precision of the steel strip, and it is difficult to be detected on line. In allusion to this problem, a method of data mining is put forward Based on the measuring system of auto gauge control (AGC), and by using the collected data such as the exit thickness, downstroke, rolling force and rolling speed, data mining is carried out on the basis of the thickness difference formula, and the roll deformation is computed. Before the data mining, the data are rematched and arranged to solve the problem of thickness difference signal hysteresis caused by the mounting distance of thickness meter. After the date mining, the exponential smoothing is used to smoothen the computed result, and the trends extrapolation is introduced to predict the current radial deformation value of the roll system. Adaptive smooth coefficient algorithm is adopted to further improve the prediction precision. The data mining and prediction methods for radial deformation of roll system are applied to a 4-high mill, a clear and complete deformation process is obtained, which lays the foundation for further research on the axial deformation rule of roll system and the compensation method, thereby improving the thickness control precision

Key words: Cold rolling, Data mining, Deformation, Exponential smoothing