基于遗传神经网络的农作物秸秆板材优化设计

Optimization design of panels made by crop straw based on genetic neural network

利用弱碱对农作物稻秸进行表面化学处理后,通过热压方式压制阻燃秸秆人造板,研究异氰酸酯、脲醛树脂胶和FRW(fire retardant of wood)阻燃剂的加入量对农作物稻秆板材力学性能和阻燃性能的影响。研究中采用正交试验和神经网络构建工艺和性能之间的非线性映射模型,再利用遗传算法对神经网络模型的权值和阈值进行优化,然后利用训练好的模型和设定的板材性能对工艺参数进行优化设计。通过验证性试验发现网络模型优化的工艺参数配比为异氰酸酯、脲醛树脂、FRW阻燃剂的加入量分别为1.926%、2.40%和15.381%,采用优化的工艺生产的板材性能和实际值之间的静曲强度、内结合强度和热释放速率峰值与目标值的误差分别为11.7%、20%和8%,相比于没有优化的工艺,其误差分别减小了35.3%、17.5%和39%。

With the treatment of alkalescency,the crops rice straw particleboards were produced by the way of hot pressure.Effects of different addition amounts of isocyanate（MDI）,urea formaldehyde resin（UF）and the fire retardant of FRW（fire retardant of wood）on the mechanical properties and flame-retardant performance were explored.In this study,the orthogonal experimental design and neural network were employed to construct the mapping model between the performances and the techniques.The genetic algorithm was adopted to optimize the weight and threshold of the model.The optimum design of the technique parameters were determined by the trained model and the given performances of the materials.With the proof generalization test,the optimized technological parameters of MDI,UF and FRW were 1.926%,2.40%and 15.381%.The properties of rice straw particleboards hot-pressed by the optimized techniques were analyzed.Compared with errors of the non-optimization model,the error of modulus and rupture（MOR）,Internal bond strength（IB）and heat release rate（HRR）was 11.7%,20%and 8%,respectively.And after optimization by GA,the errors of performances were decreased 35.3%,17.5%and 39%,respectively.