Research On Predicied Vibration Of Rolling Mill Based On Data Driving

With the development of modern industry, All industries demand for plate and strip continues to increase. However, the quality of the strip also put forward higher requirements. The mill vibration problems not only cause strip thickness fluctuations, rolls and rolling oscillation marks, but also limit the rolling speed, greatly reduce production efficiency, severe vibration likely cause run accident of heap steel and thrust belt, a serious threat to safety mill equipment has become an urgent problem of the steel industry.Previous studies of the mill vibration focused on mechanism analysis, the vibration control and prevention, and the mill vibration signal analysis. But process equipment operating status monitoring data(PDA data) received less attention which contains a large amount of equipment operating status information. Industry 4.0 era, intelligent production process is the development direction of future production through analysis and mining of the industrial data, In this paper, data mining technology used in vibration mill studies, through analysis and mining of the mill process monitoring data(PDA data), mill vibration prediction based on data-driven are achieved.First, mill PDA data has the characteristics of non-linear and strong coupling. This paper takes two wider application of data mining algorithms which include BP neural network and support vector machine(SVM). For the weakness and shortcomings of BP neural network generalization ability, it introduced Ada Boost algorithm which integrate BP neural network that improve the prediction accuracy of BP neural network algorithm and establishment of a vibration mill prediction model based on BP-Ada Boost strong predictor. For the two key parameters difficult characteristics of support vector machine(SVM) algorithm which includes relaxation coefficient and the penalty factor, this paper introduces particle swarm optimization(PSO) to optimize the choice of the two parameters and establishes a rolling mill vibration prediction model based on PSO-SVM algorithm.Second, use the field test datas to validate two types of rolling mill vibration prediction model. The results show that: both algorithms are able to achieve mill vibration prediction, and PSO-SVM algorithm accuracy is better than BP-Ada Boost algorithm; it takes PSO-SVM mill vibration prediction model quantificationally analyzes the vibration intensity which is affected by the changes in process parameters, optimizes those parameters which have a great impact in the vibration intensity and finally reduce the vibration phenomenon significantly.