Abstract: The roll gap calculation model of L₂ system is the foundation for controlling the thickness of hot rolled medium and heavy plates. Especially when employing the feedback pressure AGC control, the accuracy of roll gap calculation model of L₂ system directly affects the thickness of the head and the whole plate, consequently affecting the yield of medium and heavy plates. A 2 680 mm semi-continuous roughing rolling mill utilizes the pressure feedback AGC control system, but faces the problem of low thickness accuracy in rolling stainless steel medium and heavy plates. To address this problem, a multi-factor regression analysis was conducted using process data from the intermediate billet rolling process and actual thickness data measured by the thickness gauge. Sixteen indicators were selected for analysis to determine the correlation and path analysis of the thickness of the intermediate billet in cold state. The analysis results indicate a collinear effect of various factors on the thickness of the billet in cold state. The main influencing factors identified in the process of rolling medium and heavy plates are the roll gap, number of rolled pieces per unit of the roll, and slab discharge temperature. A multivariate fitting regression model was developed based on these key influencing factors to correct the roll gap of the R₇ stand during the rolling process. Extensive production practices have demonstrated that the application of the fitting regression model can significantly enhance the setting accuracy of the R₇ stand roll gap. This improvement has led to an increase in the overall thickness accuracy rate of medium and heavy plates from 84.47% to 98.92%.