修正D-P模型在三峡二期围堰分析中的应用

Application of modified D-P model to analysis of the second stage cofferdam in Three Gorges Project

为了考虑中主应力及材料剪胀性的影响,并验证堰体填筑材料参数,运用遗传神经网络对堰体填料进行参数反演,引入修正D-P模型进行围堰施工模拟.修正D-P模型采用非相适应的流动法则,可以模拟材料的各向同性硬化/软化及中主应力的影响,并可以考虑流变;遗传神经网络具有很强的非线性映射能力及全局优化功能.在运用遗传神经网络对堰体材料进行参数反演的基础上,利用修正D-P模型对三峡二期围堰填筑过程进行了模拟.实例分析表明,遗传神经网络能准确反演材料的力学参数;修正D-P模型适合分析围堰材料的力学性能;两者结合分析土石围堰时,计算结果与实际情况相符.修正D-P模型克服了Duncan-Chang模型不能模拟中主应力及材料剪胀性的缺点;遗传神经网络反演参数可以使计算参数与实际情况更接近,两者有机结合可提高分析精度.

In order to consider the effect of mediate principal stress and shearing - dilation of materials, and verify the material parameters of cofferdam, parameters of the materials are analyzed in reverse based on artificial neural network with genic arithmetic （GA -ANN）, construction simulation of cofferdam is carried out based on modified D - P model. Non - associated flow laws are used in D - P model so that the characteristic of materials to harden and/or soften isotropically and the effect of mediate principal stress can be simulated, and the creep characteristic of materials can be considered. The capability of GA - ANN is strong in non - linear mapping and optimizing in the whole field. After parameters of the materials are analyzed in reverse based on GA - ANN, filling process of the second stage cofferdam of Three Gorges Project is simulated based on modified D - P model. Analysis of an example shows that mechanics parameters of materials can be analyzed in reverse exactly and mechanics characteristic of materials can be simulated with modified D - P model. Moreover, the results of cofferdam construction of soil and rock simulated with two methods combined agree with those of the fact. In conclusion, without mediate principal stress and shearing - dilation of material in Duncan - Chang model, these factors can be considered in modified D -P model, and calculating parameters can be close to the practice with GA - ANN, thus analysis precision can be improved with the combination of the two methods.