基于KPCA和PSO-SVM的木材干燥过程在线优化建模研究

Optimized Online Modeling of Timber Drying Process Using Kernal Principal Component Analysis and PSO-SVM

针对木材干燥过程样本数据存在较多噪声的问题,采用核主成分分析方法对木材干燥数据进行预处理,然后利用粒子群优化的支持向量机建立木材干燥系统的在线预测模型,并进行在线预测。仿真研究表明,对数据预处理后,降维训练样本建立的木材干燥模型能够获得很好的预测精度,计算量小,速度快。在线模型能够实时反映系统当前状态,在线优化模型结构并预测系统下一步输出,实现了木材含水率特性变化的动态预测。模型输出误差小、泛化能力强,能够满足实际干燥过程在线预测控制的需要,具有良好的实际应用价值和工业前景。

This paper presents two parts of our research. In the first part of the work, the experimental data of interference and redundancy in timber drying process was pre-processed using Kernel Principal Component Analysis （ KPCA）, and in the second part, a timber drying online model based on PSO-SVM was established to predict online. The results of simulation results show that the model established has strong capability to be applied in practical kiln drying of timber with better predictive accuracy and high computing speed, and realize dynamical prediction of tim- ber moisture content change characteristics, which has a good practical application value and industrial prospect.