In order to study the complex process mechanism that affects the rolling mill vibration under dynamic rolling state and the process optimization method for vibration suppression， the data relationship is established by using the dynamic process data and equipment test data. The application method of industrial data in rolling mill stability control is explored. Taking the F3 stand vertical vibration of a 1 580 mm hot strip mill as the research object， the dynamic process parameters and vibration data of the strip production process were tracked and collected. Before building the data model， the industrial production data ragarding screening and cleaning were completed. The data model of Bayesian neural network driven by mechanism is established， which is trained and verified by using dynamic process parameters and vertical vibration acceleration of the work roll. Results show that the proposed model can achieve the purpose of predicting the running state through process parameters. According to the industrial data relationship revealed by the model， the reasonable control range of rolling force and rolling speed was given， and the effect of vibration suppression was verified. This research has a significance for realizing the industrial data application in intelligent control of rolling process.