振动力作用下饱和砂土液化的模糊神经网络预测

Fuzzy Neural Network Forecast for the Liquefaction of the Saturated Sand under the Action of the Vibration Force

根据饱和砂土在振动力作用下会发生液化,而液化的程度又具有模糊性的特点,将模糊数学与神经网络算法结合,建立饱和砂土液化预测的模糊神经网络系统。该系统是1种前向多层网络,它将传统的模糊逻辑控制器基本元件和功能与具有分布学习能力的神经网络相联系,通过实际工程样本数据训练获得系统模糊推理中饱和砂土液化的基本参数,从而优化推理系统。利用模糊神经网络系统建立饱和砂土液化的模糊规则,剔除对饱和砂土液化影响不大的因素,突出对饱和砂土液化影响较大的因素。系统以饱和砂土的相对密度、标准贯入击数、上覆有效应力、振动力幅值等参数作为预测饱和砂土液化的判别指标,预测砂土在振动力作用下发生液化的可能性。应用该系统对实际工程中35个饱和砂土样本进行液化预测,其结果与工程实际有很高的符合度。

Based on the fact that the saturated sand may become liquefaction under the vibration force and the liquefaction potential has the fuzzy characteristics,the fuzzy mathematics and neural network algorithm were combined to construct a fuzzy neural network system for forecasting the liquefaction of the saturated sand.The system is a forwards-multilayer network,integrating the basic element and function of the traditional fuzzy logistic controller with the neural network that has the ability of distributed-study.It can obtain some basic parameters of the saturated sand liquefaction in the fuzzy inference through training with the practical engineering data.Thus,the inference system can be optimized.The fuzzy rules for the lique-faction of the saturated sand were set up with the fuzzy neural network system.The factors having little effect on the liquefaction of the saturated sand were eliminated and some factors having more effect on the liquefaction of saturated sand were given prominence to.Some parameters of the saturated sand,such as the relative density,the blow count of SPT,the overlaying effectiveness stress,the vibration force value,were used as the indicators to judge the liquefaction of the saturated sand and to forecast the possibility of the liquefaction of the saturated sand under the action of vibration force.The system was used in 35saturated sand samples of the practical engineering to forecast the liquefaction,and the results were of high consistency with the practical engineering.