运用模糊神经网络研究混凝土AAR问题

The approach of applying fuzzy neural network in research on the problem of AAR in concrete

提出运用模糊神经网络研究混凝土AAR问题。利用基于试验数据运算得到的模糊神经网络获得整体 (系统 )意义上的模糊规则 ,再利用训练好的网络扩充分析所需的其它模糊规则。通过对所有规则全面、深入的分析 ,从整体上找出引起混凝土AAR膨胀诸因素中每个因素作用的大小及其内在规律。研究表明 ,混凝土总碱量的大小对AAR膨胀起主要作用 ,水泥碱比外加碱对AAR膨胀的贡献更大。利用训练好的模糊神经网络可以对多因素作用下的混凝土AAR膨胀进行较准确的预测。

The approach of applying fuzzy neural network(FNN) in research on the problem of expansion of concrete due to alkali-aggregate reaction(AAR) is proposed for the first time.the fuzzy rules of obeying whole law are obtained by use of the FNN which is trained by inputting all of the test data in the FNN,and then other fuzzy rules needed for study are complemented by using the trained FNN.By studying all of the rules obtained comprehensively the capacity of each factor contributing to the expansion due to AAR and the inherent law among many factors causing the expansion due to AAR is discovered.The calculation shows that it is the total alkali content in concrete that plays a primary role on expansion due to AAR and the function of added alkali is less important than the alkali's of cement on expansion due to AAR.The expansions of concrete due to AAR which is affected by more than one factors can be predicted accurately by use of the trained FNN.