论文部分内容阅读
目前样本数目对岩土体参数联合概率分布模型识别精度的影响还缺少研究。该文提出了样本数目对岩土体参数联合分布模型识别精度的影响分析方法,给出了基于蒙特卡洛模拟的统计量AIC值变异性模拟步骤,定义了描述岩土体参数联合概率分布模型识别精度的正确识别概率,采用蒙特卡洛模拟方法分别研究了样本数目对岩土体参数最优边缘分布函数和最优Copula函数识别精度的影响规律。结果表明:基于有限岩土体参数数据估计的边缘分布函数和Copula函数的AIC值存在较大的变异性。岩土体参数样本数目对最优边缘分布函数和Copula函数的识别精度具有重要的影响,边缘分布函数和Copula函数的正确识别概率随样本数目的增加而增大。岩土体参数变异系数对最优边缘分布函数的识别精度影响相对较小,岩土体参数间相关系数对最优Copula函数的识别精度影响较大。此外,岩土体参数二维分布模型识别比一维边缘分布模型识别需要更多的数据。因此,为了提高岩土体参数联合概率分布模型的识别精度,建议尽可能多地收集岩土体参数试验数据。
The impact of the current number of samples on the identification accuracy of the joint probability distribution model of rock and soil parameters is still lacking. This paper presents a method to analyze the effect of the number of samples on the identification accuracy of the joint distribution model of rock and soil parameters, and presents the simulation step of the statistical AIC variability based on the Monte Carlo simulation, and defines the joint probability distribution model The correct recognition probability of recognition accuracy, Monte Carlo simulation methods were used to study the effect of the number of samples on the identification accuracy of the optimal edge distribution function and the optimal Copula function of rock and soil parameters respectively. The results show that there is a large variability between the edge distribution function and the AIC value of the Copula function, which is estimated based on the data of the rock and soil parameters. The number of rock and soil parameters sample has an important influence on the optimal edge distribution function and the recognition accuracy of Copula function. The correct recognition probability of edge distribution function and Copula function increases with the number of samples. The coefficient of variation of rock and soil parameters has a relatively small influence on the identification accuracy of the optimal edge distribution function. The correlation coefficient between rock and soil parameters has a great influence on the recognition accuracy of the optimal Copula function. In addition, the identification of two-dimensional distribution model of rock and soil parameters requires more data than the one-dimensional edge distribution model identification. Therefore, in order to improve the identification precision of the joint probability distribution model of rock mass parameters, it is suggested to collect the test data of rock mass parameters as much as possible.