常用测量不确定度评定方法概述与比较

Overview and Comparison of Common Measurement Uncertainty Evaluation Methods

对测量不确定度理论诞生以来的理论与应用研究进行了系统梳理和回顾。首先,总体介绍了测量不确定度理论的历史沿革。其次概述了几类主流测量不确定度评价方法的基本原理、最新研究应用及其局限性,例如最早发布的GUM方法,该方法主要针对线性或者可近似为线性的测量模型,采用基于标准不确定度传递的方法,是目前最常用的评定方法;基于蒙特卡洛的测量不确定度评定方法及其衍生而来的拟蒙特卡洛法和自适应蒙特卡洛法在处理复杂模型时具有更广泛适用性;以贝叶斯(Bayes)为基础的测量不确定度评定方法在小样本测量中可以充分发挥先验数据的价值,并有良好的表现;此外,还讨论了一些非统计学方法的测量不确定度评定方法,如灰度评定、模糊评定、最大熵和神经网络法等。最后简要总结了各种评定方法,认为随着人工智能技术的发展,在复杂的测量模型和测量环境中应用支持向量机和神经网络等方法前景广阔。

This article provides a systematic review of the theoretical and applied research on measurement uncertainty since the inception of the theory.Firstly,it gives an overall introduction to the historical development of measurement uncertainty theory.Secondly,the basic principles,latest research applications,and limitations of several mainstream measurement uncertainty evaluation methods are summarized.For example,the earliest published GUM method,which is mainly applicable to linear or approximately linear measurement models,adopts the method based on the propagation of standard uncertainty and is currently the most commonly used evaluation method.The Monte Carlo-based measurement uncertainty evaluation method and its derivatives,the quasi-Monte Carlo method and the adaptive Monte Carlo method,have wider applicability when dealing with complex models.The Bayesian-based measurement uncertainty evaluation method can fully exploit the value of prior data in small-sample measurements and has good performance.In addition,the article discusses some non-statistical methods for measurement uncertainty evaluation,such as the grey evaluation method,fuzzy evaluation method,maximum entropy method,and neural network method.Finally,the article briefly summarizes the various evaluation methods and suggests that with the development of artificial intelligence technology,methods such as support vector machines and neural networks have broad prospects for application in complex measurement models and measurement environments.