BP神经网络预测矿物聚合材料强度的影响因素  

Factors Affecting Strength of Geopolymer Predicted by Back Propagation Neural Network

作　　者：张祖华[1] 姚晓[1] 诸华军[1] 华苏东[1] 陈悦[1] 

机构地区：[1]南京工业大学材料科学与工程学院,南京210009

出　　处：《硅酸盐通报》2008年第3期640-644,共5页Bulletin of the Chinese Ceramic Society

基　　金：国家高新技术计划“863”课题(2006AA06Z225)资助

摘　　要：基于矿物聚合材料制备过程中激发剂和配料组成的多样性与复杂性,采用BP神经网络方法考察了影响矿物聚合材料强度的3个因素:激发剂NaOH溶液的浓度、配料中碱硅物质的量比和铝硅物质的量比。结果表明:BP神经网络可以准确地预测矿物聚合材料抗折强度和抗压强度(误差在10-2数量级);高碱激发下(COH-=12mol/L),M2O/SiO2对抗压及抗折强度影响显著,预测M2O/SiO2=0.332,Al2O3/SiO2=0.441时制品的抗压强度达30.96MPa,抗折强度高达9.33MPa;SEM分析和MIP实验证明,提高激发剂NaOH溶液浓度和碱硅物质的量比更有利于形成内部结构完整、强度更高的矿物聚合材料。Based on the flexibility and complexity of source materials activating system during geopolymer synthesized process, back propagation neural network was used to study three main factors affecting strength of geopolymer, i.e. concentration of sodium hydroxide solution, alkali silicon ratio and aluminum silicon ratio in mole. The results showed that the compressive and flexural strength could be predicted precisely at the order of 10^-2. At high concentration of alkali activator （ COH^- = 12 mol/L）, M2O/SiO2 became a notable factor, which valuing 0.332 of prediction, while Al2O3/SiO2 equaling 0.441, was corresponding to a maximum compressive strength of 30. 96 MPa and flexural strength of 9. 33 MPa. SEM and mercury intrusion method confirmed that increasing the concentration of sodium hydroxide solution and alkali silicon ratio was beneficial to form completely interior structured geopolymer with high strength.

关 键 词：矿物聚合材料 煅烧高岭土 BP神经网络 强度 

分 类 号：TU502[建筑科学—建筑技术科学]