Work roll wear in hot rolling process has a substantial effect on strip profile and shape defect. To predict the wear profile, it is necessary to be familiar with significant parameters and line actual conditions.In this project, wear has been completely introduced and influential parameters have been investigated. Definition of strip profile and its elements as well as wear influence on the strip profile and ridge-buckle defect are illustrated. Essential theories due to work roll wear prediction after each pass and after a rolling program as an important goal of this project have been explained.Two distinct work roll models wear prediction have been proposed. The first model is regarding wear prediction after each pass and the second one is to be used after a rolling program. In the first model, hot rolling process has been modeled by use of finite element method, and wear is obtained by means of the resultant pressure distribution during the length of the roll barrel. To obtain the wear experimental coefficient in the equation, summation wear of each pass was obtained and it was calibrated by the work roll actual wear of a sample tested in the Mobarakeh Steel Company. Results were validated by comparing the predicted and actual rolling forces.In the second model, which was proposed for wear prediction after a rolling program, tonnage and wear graphs versus the roll barrel length have been depicted. Then, the roll barrel was divided into forty equal segments and the passed tonnage of each segment was obtained according to the passed tonnages and widths of a rolling program. This is the input term for the neural network training while its output is the wear quantity of each segment. This training is done via the graphs of work roll wear obtained from the Mobarakeh Steel Company.The fist model is more powerful than the second one because all the influential parameters are taken into account. However, the second model is better in terms of the solution speed. Differences between theoretical and actual rolling forces in the first model are justified by strip yield stress overestimation and roll flattening ignorance. Also, differences between the predicted wear profile and the measured contour in the second model can be a consequence of unawareness of some influential wear parameters, few numbers of samples for neural network training and unpredictable local wears during the length of roll barrel. In general, the obtained results of two models are reasonable in comparison with the actual results. Furthermore, predicted wear graphs are found to be in qualitative agreement with the industrial measured ones. Key Words: wear profile, hot rolling, work roll, FEM, neural network