Remote Acquisition And Application For Measured Shape Data Of Cold Rolled Strip

To break the technology monopoly of foreign shape meter,our college has independently designed and developed a new generation of the shape meter with entire seamless measuring roller and wireless signal transmission processor.However,the application of shap detection and control system needs long-term practical production experience to constantly learn and optimize,but the production environment of the steel plant is dreadful and hazardous,so a remote data acquisition system is needed instead of the researchers to collect the actual production data for a long time.The installation error of the coiling machine of a steel mill is large,and the coiling machine is fixed on the cement base,the adjustment is very inconvenient,so only use the target curve for error compensation.Besides,in order to calculate the influence coefficient of each control method of the shape control system,it need to establish the shape prediction model.The interference of installation error makes it difficult to establish the shape prediction model based on traditional mathematical models.Based on the above problems,following aspects are researched:First of all,this paper design detailedly the data sending program,4G data transmission terminal,data transmission rule,and the cloud server program.After the completion of the system,test the system in a steel plant.The result shows that the data display of cloud server interface is consistent with the shape computer interface.At the same time,we check the pressure display of two interfaces,verify that the delay is low,and the system can guarantee the real-time display of shape data.The second,by using the data obtained from remote data acquisition system,this paper find the suitable shape curve as the shape target curve through the learning of the improved BP neural network.Input bandwidth,exit and entry thickness and the rolling pass can get the desired shape target curv.This paper find the suitable shape curve as the shape target curve through the learning of the improved BP neural network.In order to solve the problem of can't learn all the samples and easy to fall into the local optimal solution of the classical BP neutral network,a small batch learning algorithm is introduced.The results of test show that the target curve obtained by the improved BP neural network is consistent with the shape curve as learning goals.The last,by using the data obtained from the system,the strip shape prediction model is established based on the deep belief network(DBN)and the deep BP neural network(DNN).The combination of the two kinds of neural networks overcomes the problem of slow training due to the increasing number of network layers and gradient diffusion.It is proved by the test that the shape prediction model constructed by DBN-DNN network has high prediction accuracy,the predicted shape can be used to calculate the influence coefficient of each control method for the strip shape control system.