基于扩展卡尔曼滤波的岩石流变模型参数识别

PARAMETER IDENTIFICATION OF ROCK RHEOLOGICAL MODEL BASED ON EXTENDED KALMAN FILTER

针对岩石材料本身复杂性及观测条件引起的室内外试验数据中存在的不确定性,引入卡尔曼滤波方法,将岩石流变模型参数视为某随机过程的状态估计量,构造相应的卡尔曼滤波方程进行参数识别。以构皮滩水电站岩石工程黏性软岩的单轴压缩蠕变试验为例,对广义开尔文模型的有关参数进行识别,结果表明该方法具有很高的识别精度。同时探讨不同观测误差对滤波过程及参数识别结果的影响,数值试验表明该算法具有较强的抗噪声干扰能力,为解决岩石力学参数识别中的不确定性问题提供一个有效的理论方法。

The extended Kalman filter method is presented for parameters identification of rock rheological model considering the uncertainty of data due to the complexity of rock material and different observation conditions of laboratory or in-situ experiments.In this method the rheological mechanical parameters are improved iteratively as the state vector of a random process.The Kalman filter function is formulated to perform parameter identification.The method is applied to a kind of clayey soft rock from Goupitan Hydropower Station.Parameters of the generalized Kelvin model are estimated based on the results obtained from uniaxial compression creep tests.The estimated results show the good precision of the extended Kalman filter method.The influence of measuring errors is also discussed.Numerical experiments are carried out to verify the greatly antinoise capability of the current method.It provides an efficient theoretical tool for solving the uncertainty effects in the identification of rock mechanical parameters.