BP神经网络模型在烟草烘烤过程中叶温变化预测中的应用

Establishment of a BP neural network model for predicting leaf temperature in tobacco baking process

【目的】建立合适的BP神经网络模型,了解散叶烘烤过程中一系列烘烤因素对叶温变化的影响,为烤烟烘烤调制过程中叶温变化研究提供参考。【方法】运用叶温测定仪和温湿度自控仪记录烘烤过程中干球温度、湿球温度、相对湿度及干球温度与叶温的差值,并将此4项指标作为输入变量,叶温作为输出变量,建立一个拓扑结构为4-4-1的BP神经网络模型。【结果】所建立的BP神经网络模型模拟结果很快收敛,预测结果的绝对误差与相对误差小,预测所用的20组数据中相对误差>1%的有8组数据,相对误差>2%的有2组数据,相对误差<1%的有12组数据。【结论】所建立的BP神经网络模型在对烟叶烘烤过程中叶温变化的预测效果较好。

[Objective]In order to research the influence of a series of curing factors on leaf temperature in baking process, a suitable BP neural network model was built to provide references for leaf temperature variation. [Method]The dry bulb temperature, wet bulb temperature, relative humidity, and the difference betwean dry bulb temperature and leaf temperature were recorded using a leaf temperature tester and temperature-humidity control detector. Then the four indicators in the baking process were used as input variables, and leaf temperature as an output variable to set up a BP neural network model of 4-4-1. [Result]The results showed that the simulant result converged soon, and the absolute error and relative error of prediction were minimum. There were 8 sets of data whose relative error was 〉1%. There were 2 sets of data whose relative error was 〉2%. There were 12 sets of data whose relative error was〈1%. [Conclusion]The established BP neural network model had content prediction effects for tobacco leaf temperature in the baking process.