【作者】 白波；

【导师】 王逢瑚；

【作者基本信息】 东北林业大学 ， 木材科学与技术， 2008， 博士

【摘要】 本论文首先对秸秆进行表面碱液处理,然后采用不同比例的异氰酸酯和脲醛胶配比,加入不同量的FRW阻燃剂压制成板,并对板材进行力学性能测试和阻燃性能测试,优化出最佳配比和制造工艺;并在最佳工艺的基础上压制板材进行贴面,制造用于家具使用的低毒、阻燃型秸秆家具饰面人造板;最后应用人工神经网络建模进行秸秆人造板性能预测,可以得出以下几点结论:(1)通过碱处理的稻草原料的扩散—渗透系数K值都有所提高,说明经过处理过的稻草表面润湿性能得到改善,有利于胶粘剂的浸润,从而提高稻草人造板的胶接强度;通过正交实验得出稻草表面处理的最佳处理方案是:碱的浓度为0.4%～0.8%,处理时碱的用量为(料液比)1:2.5～1:4.5,原料的碱处理时间为12h～48h;麦秆表面处理的最优方案为:所用碱的浓度为0.6%,处理时碱的用量为(料液比)1:3.5,原料的碱处理时间为12h。(2)使用处理过的秸秆制得的人造板的内结合强度有大幅度提高,满足国家标准的要求;2h吸水厚度膨胀率降低,达到国家标准要求;静曲强度和弯曲弹性模量稍有下降,但下降幅度并不大,能满足国家标准的要求;(3)通过MDI、UF和FRW的添加量对稻草板物理力学性能指标影响的方差分析和极差分析结果的可以看出,MDI的添加量对稻草板MOE、MOR、IB、2hTS指标有显著影响;而UF和FRW对这四项指标的影响较小。随着MDI、UF施胶量的增加,稻草板MOE、MOR、IB的强度有上升的趋势,2hTS随施胶量和FRW的添加量的增加而下降;(4) MDI、UF和FRW的添加量对稻草板阻燃性能指标影响的方差分析结果表明,三个影响因子对稻草板的点燃时间、热释放峰值和峰值出现时间三个参数的影响均不显著;随着FRW的添加量的增加,热释放峰值有下降的趋势,阻燃剂中硼酸物质物质的存在降低了稻草的热释放速率,这是达到阻燃目的的最重要原因;(5)通过正交试验结果分析得到阻燃麦秸板的最优工艺条件为MDI施胶量4%,UF施胶量7%,FRW添加量10%,重复验证试验测得的各项力学性能指标均符合国家标准的要求,其力学性能指标分别是MOE为3461MPa,MOR为32MPa,IB为0.73MPa,2hTS为10.4%,;阻燃稻草板的最优工艺条件为MDI施胶量3%,UF施胶量6%,FRW添加量10%。重复验证试验测得的各项力学性能指标均符合国家标准的要求,其力学性能指标分别是MOE为2423MPa,MOR为29 MPa,IB为0.63 MPa,2hTS为7.9%;较素板相比,阻燃稻草板的阻燃性能有明显提高,达到阻燃的目的;(6)采用薄木、三聚氰胺两种材料贴面制得的稻草和麦秆人造板材的表面胶合强度均超过了国家标准,可以直接用于家具制造;(7)选用三层BP网络进行秸秆人造板性能预测,输入参数为异氰酸酯MDI、脲醛树脂(UF)、FRW新型木材阻燃剂的用量,对应的输出参数为秸秆人造板的静曲强度、内结合强度和热释放速率峰值。检验结果表明输出很好,它们与目标输出的相关系数都在0.99以上,综合考虑训练步长和训练时间,确定隐含层节点为8的网络;利用训练好的网络输出参数值与训练目标样本进行拟合,得出相关系数分别为0.999、0.994和1。但是验证性实验结果有一定的误差;用遗传算法对模型的权值和阈值进行优化后,网络的推广性能得到提高,通过进一步验证,优化后的网络能正确描述秸秆人造板工艺参数与性能之间的关系。更多还原

【Abstract】 This paper explored the effect of MDI,UF and FRW content on the mechanical and fire retardant property of straw based panels with surface alkali liquor processing.The material ratio and hot-processing parameters were optimized.In order to manufacture the straw based panel with low toxic and fire retardant,veneer overlaying was carried out under the optimized condition.Finally,the neural network techniques were used for the prediction of physical and mechanical properties of straw based panels.The results were showed as followed:(1) The conductivity and diffusion coefficient K of the straw material after alkaline liquid treatment increased obviously.This indicated that alkaline liquid treatment improved the surface wet ability of straw,which is helpful for the infiltration of resin.The results of orthogonal test showed that the optimized treating condition was alkaline liquid concentration as 0.4-0.8%,alkaline dosage as 1:2.5-1:4.5,alkaline-treated time as 12h-48h.(2) The IB of straw based panels after treated increased remarkably and it could satisfy the national standard.The 2hTS,MOE and MOR decreased slightly.The reason for this change is the ruin of fiber structure and the waxiness;the improvement of the straw surface wet ability is helpful to the forming of chemical bond.(3) The variance analysis of physical and mechanical property of straw based panel showed that MOE,MOR,IB and 2hTS were significantly affected by the content of MDI.With the increasing of resin content,the property of MOR,MOE and IB increased.The 2 hTS decreased with the increasing of FRW content.(4) The variance analysis of the fire retardant property of straw based panel showed that TTI,pkHRR and peak value appearance time were not affected by the MDI,UF and FRW content significantly.With the increasing of FRW content,the pkHRR decreased meanwhile. The existence of boric acid in FRW decreased the HRR that is the keynote of fire retardant mechanism.(5) The results of orthogonal test showed that the optimized processing condition was MDI content as 3%,UF resin content as 6%and the FRW content as 10%.The physical and mechanical value of straw based panel that pressed under the optimized condition were MOE as 2423MPa,MOR as 29 MPa,IB as 0.63 MPa,2hTS as 7.9%.Compared with untreated straw based panel,the fire retardant property of the straw based panel with FRW was obviously improved.(6) IB values of straw based panels with veneer and melamine overlaying exceeded the national standard,and they could be used in furniture manufacture directly.(7) Based on the three layers BP neural network techniques,the content of MDI,UF and FRW were chosen as the input parameters,corresponding to the MOR,IB and the fire retardant parameters HRR.The result showed that the correlation coefficient between the actual value and target output value is 0.99.Considering the step and the time of training,the hidden layer was set to 10 in the model.The correlation coefficients of fitting data between the trained output value and the target sample were 0.999,0.994 and 1 separately.This indicated that the result of the model was good based on the BP neural network,and more improvement was needed in the future.更多还原