含相关性的测量不确定度拟蒙特卡罗评定方法  

Quasi Monte Carlo method for the measurement uncertainty evaluation considering correlation

作　　者：凌明祥[1] 李会敏[1] 黎启胜[1] 李明海[1] 

机构地区：[1]中国工程物理研究院总体工程研究所,绵阳621900

出　　处：《仪器仪表学报》2014年第6期1385-1393,共9页Chinese Journal of Scientific Instrument

基　　金：国家重大科学仪器设备开发专项(2011YQ130047)资助项目

摘　　要：针对测量不确定度评定中存在相关性分量以及蒙特卡罗法收敛速度和精度较低的问题,提出能处理不确定度源相关且兼顾均匀性和随机性的复杂系统测量不确定度拟蒙特卡罗评定方法。该方法通过产生均匀分布的独立随机化Sobol序列,并基于Cholesky因子线性-非线性变换产生具有任意边缘分布和相关系数的多维随机抽样序列,建立测量模型进行概率分布传播得到被测量的估计值和包含区间。算例和实际测量表明,该方法具有比蒙特卡罗法更高的精度和效率,而且除保持一般拟蒙特卡罗法的优良低偏差性外更能真实反映测量不确定度传播的随机过程,尤其适用于测量不确定度源服从非正态分布且相关的情况。对采用拟蒙特卡罗法评定测量不确定度的一些问题进行了讨论和总结。Aiming at the problems that there exist correlation components and the convergence speed ＆ precision of Monte Carlo method （MCM） are low in measurement uncertainty evaluation, a quasi Monte Carlo method （QMCM） is proposed for the measurement uncertainty evaluation of complicated system, which can deal with correlated uncertainty sources and consider both uniformity and randomness. Through generating independent randomized Sobol sequence with uniform distribution, this method generates multi-dimensional random variable sampling sequence with arbitrary marginal probability distribution and target correlation coefficient based on improved linear-nonlinear transforming technique with Cholesky factor. The measurement model is built, the probability distribution of the measurand is propagated and calculated, and the estimation and coverage interval of the measurand are obtained from its sampling sequence. The example and actual measurement illustrate that the proposed method has higher precision and efficiency compared with MCM ; besides keeping the excellent low deviation performance of QMCM, this method can really reflect the random process of measurement uncertainty propagation. The proposed method is especially suitable for the cases where the uncertainty sources are correlated and have non-normal distribution. Some issues about using QMCM for measurement uncertainty evaluation are also discussed and summarized.

关 键 词：测量不确定度 蒙特卡罗法 相关性 随机变量抽样 Sobol序列 

分 类 号：TB9[一般工业技术—计量学]