模糊神经网络在高强混凝土强度预测与配合比设计中的应用

针对模糊系统中规则结论为数值和线性函数的两种表示方式 ,找到了它们的共同点 ,将它们置于同一网络结构中 ,形成规则结论为数值和线性函数 (T -S模型 )的两种模糊神经网络 (FuzzyNeuralNetworks,简称FNN) ,导出了它们的网络模型及其学习算法。并首次将其应用于高强混凝土强度预测和配合比设计中。文章还介绍了一种简单有效地从样本数据中提取模糊规则及确定FNN参数初值的方法。运算结果表明 ,FNN不仅具有很高的预测精度 ,而且网络的结点和权值均具有明确的物理意义 ,可以借此深入分析高强混凝土综合性能与影响它们的因素之间的非线性关系

Application of Fuzzy Neural Network in Strength Prediction and Optimal Design of High Strength Concrete

The common ground of the value-type conclusion and the linear functional type conclusion in fuzzy system is found and the models and learning algorithms of the two kinds of fuzzy neural networks (FNN) are formed by putting them in a network structure, which are at the first time applied to predict strength and design the mix proportion of high strength concrete. Furthermore, a simple method to extract fuzzy rules and to determine the primary parameters of FNN from the sample data is recommended. According to the operational results it is shown that FNN has high predicting accuracy, furthermore its nodes and weights have definite physical meaning, by which the non linear relationships between the comprehensive performances of high strength concrete and the factors affecting its performances can be analyzed.