Prediction Of Sandy Soil Liquefaction During Earthquakes By The Neural Network

Sandy soil liquefaction make foundation lose bearing capacity and produce non-uniform settlements. It bring huge calamities. So, it is important to forecast and evaluate the possibility and the damage degree of liquefaction quickly and effectively. Based on macroscopical earthquake calamities and laboratory tests,traditional methods for estimation and grade evaluation of liquefaction are inducted by means of generalization, analyses and statistics, which have some practicability and some limits. Every kind of methods have some uncertainty and limits,because influence factors of sandy soil liquefaction are complicated. Therefore, it is necessary to establish the method of evaluation of liquefaction , which embrace general and many indexes.The fundamental principle and network model of BP network is expounded in this thesis. BP network model is improved and established according to the characteristic of the problem studied. The procedure based on the model is redacted by using the Fortran language. The thesis also discusses the influence by change the parameter of activation function on hidden layer .1)The data of sandy soil seismic liquefaction are collected and inputed to date file according the sequence and certain format of the data.2)BP neural network model which forecastes and evaluates sandy soil liquefaction is established. The data in file which have been managed are inputed to the neural network procedure and are used to train the network. The procedure is tested by inputing the testing specimen. The results producing by the procedure is compared with the factual results on the scene of sandy soil liquefaction. N13)Some neural network models are established by changing the the parameter of activation function on hidden layer, then are trained and tested by the same data. By compare, the influence to the network is discussed.The results show that the BP neural network model of estimation of liquefaction is practical and effective method. It provide a new method and way for solving these problems whose academic answer is not perfect and function relation is not clear. So it is valuable and has bright future.