【作者】 李阳；

【导师】 荆学东；

【作者基本信息】 陕西科技大学 ， 机械电子工程， 2015， 硕士

【摘要】 首先探讨了常用的测量不确定度的评估方法的基本原理及其适用范围,分析了这些方法的优势及不足之处,针对其中的不足提出将卷积原理应用到测量不确定度评定中,特别是针对传感器的测量不确定度评定。之后分析了传感器的测量不确定度来源,包括非线性度、迟滞以及重复性等;对于不同的传感器,各个参数的影响不同,分析了常用的校准方法,以对于影响较大的参数可以进行合理校准,减小其对传感器精度的影响。重点研究了基于卷积原理的传感器的测量不确定度评估方法,其理论依据是:当传感器的不确定度来源相互独立,且其概率密度函数已知时,理论上可以通过卷积求得合成分布的概率密度函数,由此再设法确定传感器的测量不确定度。基于此,本文提出了基于离散卷积的一种传感器测量不确定度评估算法。在该评估算法中,将传感器的每一种概率分布密度函数进行离散化,以获得输入序列,并可以根据输入序列的长度选择循环卷积或分段卷积以提高卷积的效率和质量;应用MATLAB实现了上述算法,由此解决了GUM评估方法中存在的利用经验估计置信系数的问题。在给定显著水平α的情况下,运用上述算法可以确定合成分布的置信系数k的值,为进一步计算扩展不确定度提供了依据。作为上述方法应用的实例,选取电阻应变式称重传感器和非接触式测量的电涡流位移传感器作为对象,分别运用GUM评定法、MCM评定法和基于卷积原理的测量不确定度评定方法对传感器进行不确定度的评定;综合分析比较三种方法的优缺点,结果表明:基于卷积原理的评估法比GUM评定法有更好的精度保证,MCM方法的仿真结果也证明了其实用性;因而,应用卷积方法对传感器测量不确定度评估具有一定的有效性。更多还原

【Abstract】 The measurement uncertainty evaluation methods available, including GUM method and Monte Carlo method, have been discussed in detail, especially their application fields, advantages and disadvantages; To compensate their deficiencies, the convolution approach is proposed to applied for measurement uncertainty evaluation,especially for that of transducers.After that, the sources of measurement uncertainty of transducers are analyzed detailedly, including the non-linearity, the hysteresis and the repeatability; and the effect of these sources on performance of transducers is also discussed; meanwhile some commonly-used calibration methods are also presented, which can be used to improve the accuracy of transducers.Based on the study above, the measurement uncertainty evaluation method on transducers is proposed on the basis of the principle of convolution; since if all the probability density functions of uncertainty sources are available, the probability density function of the combined distribution can be obtained via convolution approach; therefore the measurement uncertainty of the transducers could be managed to be determined if the confidence is specified. To achieve this purpose, the discrete convolution algorithm is adopted. With respect to this algorithm, each probability density function need to be discretized so as to obtain the corresponding sequence as input of the algorithm; moreover, the efficiency and quality of convolution could be improved by selecting circular convolution or piecewise convolution, depending on the length of input sequence; the algorithm is realized via MATLAB. If the confidence level is α specified, the value of confidence coefficient k can be determined with this algorithm; while the confidence coefficient k is determined by experience in GUM.To verity the effectiveness of the method above, the measurement uncertainty evaluation on a resistance strain sensor and an eddy current displacement sensor is carried out, whose results are compared with those of the GUM and MCM method. The results indicate that the convolution method is more accurate than the GUM evaluation methods and MCM method.更多还原