【作者】 王飞；

【导师】 周修理；

【作者基本信息】 东北农业大学 ， 物联网工程， 2018， 硕士

【摘要】 21世纪,我国的人造板产量已然位居世界前列。混合材料板作为人造板工业的重要组成部分,其需求和产量也随之增加。我国作为世界人口大国,森林覆盖率仅仅占21.66%,还达不到世界平均水平;据资料显示,我国每年废弃和焚烧的农作物秸秆已达到2~3亿吨,不仅导致自然资源严重浪费,也带来严重的环境污染。因此,如何高效节约综合利用木材和植物秸秆,缓解木材资源供需不平衡,提高自然资源的利用率、增加其附加值、改善环境、实现自然资源的可持续发展显得尤为迫切。热压是混合材料板生产的主要工序之一,对其力学特性起决定性作用。研究从工程应用的角度出发,以混合材料板热压过程为研究对象,将机器学习和智能算法应用到混合材料板力学特性预测中,利用MATLAB2015b和开源软件LIBSVM-3.22进行分析和研究,主要研究内容和研究结果:(1)通过分析混合材料板热压过程发现:混合材料板力学特性(静曲强度、弹性模量、内结合强度)主要受原料参数(含水率)及热压控制参数(热压温度、压力、时间)的影响;(2)确定正交试验因素水平进行混合材料板热压试验,对其力学特性进行测试并整理实验数据,利用支持向量机回归原理构建混合材料板力学特性SVR预测模型;(3)混合材料板力学特性SVR预测模型性能主要受惩罚因子C与RBF核函数参数g影响,针对网格搜索法优选参数易陷入局部最优的问题,本文采用全局优化算法(遗传算法和粒子群算法)优化和选择C和g,构建全局最优的PSO-SVR/GA-SVR预测模型;(4)对比SVR、PSO-SVR和GA-SVR的实验结果并进行分析讨论,结果表明PSO-SVR预测模型可更好地描述热压参数与混合材料板力学特性之间的非线性关系,能够根据自变量快速准确地预测混合材料板力学特性。相比较SVR和GA-SVR,PSO-SVR预测模型预测精度更高、稳定性更强、泛化性能更好。(5)验证预测模型的稳定性,基于Matlab GUI结合PSO-SVR预测模型以混合材料板热压正交试验结果作为数据集分别搭建混合材料板力学特性预测界面。对混合材料板热压过程的研究,为农作物秸秆的高附加值应用提供新思路。PSO-SVR预测模型可为混合材料板的性能预测和热压控制参数选择提供理论参考。更多还原

【Abstract】 21st century,the output of man-made boards has been ranked in the forefront of the world.As an important component of the man-made boards industry,the demand and output of mixed material boards are also increased.China,as a world’s most populous country,has a forest coverage rate of only 21.66%,which still falls short of the world’s average.According to statistics,China’s annual crop straw that has been discarded and burned has reached 200-300 million tons,which can causes serious waste of natural resources and environmental pollution.So,how to efficiently economize the comprehensive utilization of wood and plant straw,alleviate the imbalance between supply and demand of timber resources,improve the utilization rate of natural resources,increase the added value,improve the environment and realize the sustainable development of natural resources is particularly urgent.Hot-pressing is one of the main processes in the production of mixed material boards,which plays a decisive role in the mechanical properties of mixed material boards.From the perspective of engineering application,taking the hot-pressing process of mixed material boards as the research object,applies the machine learning and intelligent algorithm to the prediction of the mechanical properties of mixed material boards,and uses MATLAB2015 b and open-source software LIBSVM-3.22 for analysis and research.The main research contents and research results are as follows:(1)By analyzing the hot-pressing process of mixed material boards,it was found that the mechanical properties(compressive strength,elastic modulus,internal bond strength)of mixed material boards are mainly affected by the raw material parameters(moisture content)and hot pressing parameters(hot pressing temperature,pressure and time);(2)Determine the level of orthogonal experiment to conduct the hot pressing test of mixed material boards,test its mechanical properties and sort out the experimental data,and construct the SVR prediction model of the mechanical properties of mixed material boards by using the support vector regression principle;(3)Mechanical properties of mixed material boards SVR prediction model performance is mainly affected by the penalty factor C and RBF kernel function parameters g.For the grid search method,the optimal parameters are easy to fall into local optimum.This paper adopts the globaloptimization algorithm(genetic algorithm and particle swarm optimization algorithm).Optimize and select C,g build a globally optimal PSO-SVR/GA-SVR prediction model;(4)Comparing the experimental results of SVR,PSO-SVR and GA-SVR and analyzing and discussing.The results show that the PSO-SVR prediction model can better describe the nonlinear relationship between the hot-pressing parameters and the mechanical properties of mixed material boards,Rapidly and accurately predict the mechanical properties of mixed material boards based on independent variables.Compared with SVR and GA-SVR,PSO-SVR prediction model has higher prediction accuracy,stronger stability and better generalization performance;(5)Verify the stability of the PSO-SVR prediction model.Based on Matlab GUI and PSO-SVR prediction model,the prediction UI of mechanical properties of mixed material boards is established by using the results of orthogonal test of mixed material plate hot-pressing as data set.The research on the-hot pressing process of mixed material boards provide new idea for the application of high added value of crop straw.PSO-SVR prediction model can provide theoretical reference for performance prediction and hot-pressing control parameter selection of mixed material boards.更多还原