Abstract Roller gap displacement sensor is an important part in the feedback channel of the hot rolling mill with AGC system. Structure of the sensor unreasonable designed must affect the stability of the rolling process. Establish the digital models of gap displacement sensor structure is the bases of rolling control system by computer simulation analysis, and it is the evidence of controller optimization, at the same time, it is an important means of research hot rolling mill vibration characteristics. At first, we have identified the roll gap dynamic model structure of the displacement sensor by theoretical analysis. Second, we have identified the dynamics model parameters of roll gap displacement sensor and have established the structure dynamic model of the roll gap displacement sensor by system identification technology and system identification toolbox of MATLAB. System input and output data were simultaneously recorded the location of mill vibration acceleration signal correspondingly. We using ARX model structure to establish system model, and using least squares estimation method (LSCE) to the estimate the parameter of the ARX model. The experimental results show that Identification model fitting to 79.48 percent, and it indicating a higher recognition accuracy of the model.