Application Of Fuzzy Neural Network To Classification About The Grade Of Swelling-Shrinkage For Expansive Soils

Because there are many and complex factors that affect classification about the grade of swelling-shrinkage for expansive soils. And the connection among the factors is fuzzy. So classification about the grade of swelling-shrinkage for expansive soils is typical question of indeterminacy. To reasonably classify expansive soils, this paper introduces a kind of method that is applicable to classify expansive soils. It is fuzzy neural network method (compensation fuzzy neural network and adaptive fuzzy neural network), which can deal with fuzzy information and automatically design and adjust rules and membership functions of fuzzy system. Thus it combines advantages of neural network and fuzzy logic. So this provides a new way of classifying the grade of swelling-shrinkage for expansive soils.In this paper different judgment indexes of swelling-shrinking grades have been summed up on analyzing many factors affecting expansive soil grades and taking account of the methods of classification about the grade of swelling-shrinkage for expansive soils both at home and abroad. Basing on MATLAB language, I have compiled training process of compensation fuzzyneural network and adaptive fuzzy neural network, and classified three groups of expansive soils measured data that have different judgment indexes. Average value of the three groups of accuracy of classification for expansive soils is used to estimate validity of the two methods. The results indicate that classification effect of the two methods is good and error is small. Fuzzy neural network method will be a potent method of classification for expansive soils.Using single method of classification for expansive soils may mistakenly judge. For best effect of classification for expansive soils, so in this paper BP neural network, fuzzy comprehensive evaluation, gray cluster and table lookup are used for reference methods. Finally, the paper does some deeply comparison on the method of fuzzy neural network with the reference methods . Then the study and analysis are put forward in the paper. Some examples proved that the method of fuzzy neural network is more practicable and efficient in the field of classification about the grade of swelling-shrinkage for expansive soils, and it is worthy of making further research and promoting in future.