依据响应曲面法和BP神经网络对木纤维脉冲-旋流气流干燥工艺优化

为探讨脉冲-旋流气流干燥工艺对木质纤维终含水率的影响,选用Box-Behnken试验设计进行响应曲面法试验,考查杨木纤维脉冲-旋流气流干燥过程中主要工艺参数对终含水率的影响。结果表明:影响杨木纤维终含水率的工艺参数递次是初含水率、进风温度、进料速度和风速。选用TensorFlow框架借助python编程语言建构BP神经网络,搭建终含水率预测模型,结果显示:足够的样本数据反映出规律特征后,预测模型的优化效果得以改善。将响应曲面法试验和BP神经网络模型优化效果进行对照,响应曲面法和BP神经网络的ERMS值、R2值分别为0.2264、0.9834和0.4419、0.9794,反映出响应曲面法的优化水平更好。研究结果旨在为丰富木质纤维气流干燥理论体系及其工业化应用提供技术支持和理论依据。

Optimization of Wood Fiber Impulse-Cyclone Airflow Drying Process with RSM and BP Neural Network

We used the Box-Behnken test design to conduct the RSM experiment for investigating the effect of impulse-cyclone airflow drying process on the final moisture content of wood fiber.The effects of four factors,including inlet air temperature,air inlet speed,feed rate and initial moisture content,on the final moisture content of poplar fiber impulse-cyclone airflow drying equipment were investigated.The process parameters affecting the final moisture content of poplar fiber are initial moisture content,inlet air temperature,feed rate and wind speed.Using the Tensor Flow framework to construct BP neural network with Python programming language,the final moisture content prediction model was configured,and the optimization effect of the prediction model was improved after sufficient sample data reflects the regular features.Compared with the optimization effects of RSM and BP neural network models,the ERMS and R^2 values of RSM and BP neural networks were 0.2264,0.9834 and 0.4419,0.9794,respectively,reflecting that the optimization level of RSM were better.The research results could provide technical support and theoretical basis for enriching the theoretical system of wood fiber airflow drying and its industrial application.