Prediction Of Discharge Moisture Content Of Loosening And Conditioning Based On Bayesian Network

The tobacco industry is an important pillar industry that promotes the sustainable and healthy economic development of the province and even the whole country.Tobacco production has sophisticated technological processes.Among them,the most critical factor that determines the quality of cigarettes is the moisture content of cut tobacco,and the loosening and conditioning process is one of the key links to control moisture.At present,the quality control of loose and condition products faces challenges such as multivariable,weak controllability,and complicated processes.Therefore,how to effectively control the moisture of the cut tobacco outlet has become an important issue in the current tobacco production process.This paper constructs a network model based on the Bayesian network analysis method.In this model,the water addition amount,drum speed,process hot air temperature,process flow rate,cumulative amount of material per unit time and discharge moisture content are nodes.The tobacco-making process is long and the types of equipment are various.Therefore,effective data interception rules are established first,and data preprocessing is completed with the help of R software.At the same time,the effective data set should be divided into two groups: training set and testing set.Based on the correlation between variables,a Bayesian network topology of continuous data is constructed.The influence weight of five key parameters on the discharging index is calculated by the structure diagram,which provides a reference for the sorting of node variables in the follow-up work.To further construct the discharge moisture content forecasting model of loosening and conditioning process,the following steps are performed.Firstly,all parameters are discretized according to the tobacco making process standard.The discharge moisture content of loosening and conditioning process is divided into five intervals: less than16,[16,17),[17,18),[18.19)and greater than or equal to 19.Then using the training set data and the BNT toolbox in MATLAB software to learn the structure and parameters of Bayesian network,so as to obtain the discharge moisture content forecasting model of loosening and conditioning process based on Bayesian network.The network inference is completed by cluster tree propagation algorithm.In addition,the testing set data is imported to verify the network model from two levels of learning precision and prediction accuracy.The results show that the model has high accuracy.Finally,in the application stage of the model,the direct and indirect factors affecting the discharge moisture content can be intuitively obtained according to the structure of the constructed Bayesian network model.Based on this,the parameter priority adjustment scheme in tobacco making process can also be put forward.Meanwhile,according to the prediction results,the optimal value ranges of other process parameters are found when the discharge moisture content is controlled at [17,18)and [18.19).The distribution diagram of the discharge moisture content at this time is given and compared with the distribution of the overall data.The results show that the optimal scheme can effectively increase the probability that the discharge moisture content value falls into the standard range in the loosening and conditioning process,which can provide technical support for further optimizing the tobacco production process and realizing fine management.