Research On The Soft Sensor Model Of Rotary Kiln Calcining Zone Temperature Based On GBMA-SVR

Rotary kiln is the main production equipment for iron ore pellets.The temperature of the calcination zone is one of the key technical indicators that determine the quality of the finished pellets.Improving the temperature control level of the calcination zone is of great significance to improving the quality of the finished pellets.However,when the rotary kiln is running,the thermocouple signal transmission slip ring is prone to poor contact and the temperature measurement signal is unstable.In addition,under strong oxidation and high temperature environments,the service life of thermocouples is extremely short.To replace faulty thermocouples,the rotary kiln must be shut down,which greatly affects the field production efficiency.The accuracy of non-contact temperature measurement methods(such as infrared temperature measurement)is low in harsh environment.In view of the low accuracy of domestic rotary kiln temperature control and the difficulty of monitoring,it is of great significance to establish a soft-sensing model to predict the temperature in the calcining zone of the rotary kiln.Therefore,this paper proposes to use soft sensing technology to realize on-line temperature prediction of rotary kiln calcination zone.Through the relationship between the auxiliary variables easy to be measured and the temperature in rotary kiln,a mathematical model is established to calculate the temperature prediction value of calcination zone.In this paper,three research methods: support vector machine regression,data preprocessing,and intelligent optimization algorithm are used to establish the temperature prediction model of the rotary kiln,and provide timely and effective data guidance for the field staff.Soft measurement technology is an effective tool for online evaluation of unmeasured or difficult to directly measure variables.The performance of soft sensing depends to a large extent on its convergence speed and generalization ability.Based on soft measurement technology,we propose a new soft measurement model IsomapSVR.Firstly,SOM neural network is used to divide the sample data into two categories: training set and test set,so as to ensure the fairness and symmetry of data segmentation and make the prediction results more convincing.Isomap method is used to reduce the dimension of input data.It can not only reduce the structural complexity of the proposed model,but also accelerate the learning speed.Then support vector machine regression(SVR)model is used to predict the variables.In order to solve the problem of determining the hyperparameters in the Isomap dimensionality reduction algorithm and support vector machine regression,this paper proposes a new biomigration algorithm(GBMA)based on the full-enclosed topology to optimize the hyperparameters of Isomap and SVR.To improve the stability and generalization ability of the model.First,the adaptive weight factor is introduced to improve the overall convergence performance of the biological migration algorithm(BMA),and to ensure the convergence accuracy of the algorithm.Secondly,in order to strengthen the information exchange between species,a central point symmetry topology is constructed.Eight species are selected to share the location in the adjacent position of each species,so as to balance the exploration and development ability of the algorithm.Thirdly,chaos theory is introduced to improve the initialization state of species to avoid premature convergence and falling into local optimal state.The proposed GBMA algorithm is evaluated by CEC2017 test set,and some common algorithms are selected to test with GBMA at the same time,Through the above analysis results,it is found that the performance of the proposed GBMA algorithm is better than other optimization algorithms.The improved biological migration algorithm is applied to the joint optimization of Isomap and SVR parameters to obtain the optimal parameters,and the parameters are input into the model to predict the temperature in the calcining zone of the rotary kiln,which is difficult to directly measure.Simulation results show that the proposed soft sensor modeling method has higher learning speed and better generalization ability.Compared with other models,this model is better than other models,which proves that this modeling method is feasible.Finally,in order to apply the designed temperature prediction model for the calcining zone of the rotary kiln to the industrial site,the database is integrated on the basis of C#,and a soft sensor is established by MATLAB and PLC,which can control the kiln of the rotary kiln without affecting the original industrial control system.The temperature is predicted,and the predicted result is displayed in the host computer,laying the foundation for the industrial application of soft measurement technology.