基于均匀设计与BP神经网络优化制备SiO_2基相变调湿复合材料的预测模型

Prediction Model for Optimizing Preparation of SiO_2-based Phase Change and Humidity Storage Composites with Uniform Design and Back-propagation Neural Network

以SiO2为载体、脂肪酸为相变材料制备SiO2基相变调湿复合材料,运用均匀实验设计结合BP神经网络优化制备参数,对最优材料进行表征,建立了优化制备工艺与综合相变调湿性能的BP神经网络模型.结果表明,最优制备条件为溶液p H值为3.63、超声波功率100 W、去离子水与正硅酸乙酯物质的量比9.71、无水乙醇与正硅酸乙酯物质的量比5.18、脂肪酸与正硅酸乙酯物质的量比0.51;最优SiO2基相变调湿复合材料在相对湿度97.30%时的平衡含湿量为0.3057 g/g,从30℃到15℃的降温时间为1445 s,综合相变调湿性能为1.6014,实验结果与模型预测值吻合较好,相对误差为-1.70%~1.89%.脂肪酸包覆于SiO2的网络孔隙结构中形成最优SiO2基相变调湿复合材料,粒径主要分布在约100 nm.验证了利用二次回归方程对均匀设计实验的分析成果.

With SiO2 as carrier, fatty acid as phase change material, SiO2-based phase change and humidity storage composites were prepared. The scheme was optimized by uniform design in a combination with BP neural network to optimize preparation of SiO2-based phase change and humidity storage composites. The performance of optimal SiO2-based composites were characterized. The results show that the optimal parameters are solution pH value 3.63, ultrasonic wave power 100 W, molar ratio of deionized water to tetraethyl orthosilicate 9.71, molar ratio of absolute alcohol to tetraethyl orthosilicate 5.18 and molar ratio of fatty acid to tetraethyl orthosilicate 0.51. The optimal equilibrium moisture content under the relative humidity of 97.30% is 0.3057 g/g, cooling time from 30 ~C down to 15 ~C is 1445 s, and overall performance of phase change and humidity storage is 1.6014. The experimental results and the model prediction are in good agreement （relative error is -1.70%-1.89%）. The optimal SiO2-based phase change and humidity storage composites form with fatty acid coated on SiO2 network pore structure, and have the particle size distribution at about 100 nm. The above mentioned results verify the analysis with quadratic regression equation on the results obtained in uniform experimental design.