IntroductionIn the contemporary measurement theory, the new methods of accuracy determination are still investigated. It is possible, that the theory of uncertainty is a universal tool for the estimation of accuracy [1,2]. This theory allows for the analysis of random interactions, even if the measurement model is strongly nonlinear. According to the theory of uncertainty, the standard uncertainty plays the main role in the accuracy determination.The problem of determination of uncertainty components in thermovision measurement was described in [3][4][5][6]. In this paper the analysis of the measurement method accuracy, based on the coverage interval idea was presented. In the simulation research, the model described in [4,5,7] was assumed. It can be formulated as a function of five input variables:where: εob -object emissivity, Tatm -the temperature of atmosphere, K, T0 -ambient temperature, K ω -relative humidity, d -distance between infrared camera and the object, m.In the research of components of combined standard uncertainty, by means model (1), Monte Carlo simulations were used. The expanded uncertainty, with the assumed level of confidence was calculated using the method for the propagation of distributions.
The method for the propagation of distributions and Monte Carlo simulationsThe goal of the use the propagation of distributions is the determination of uncertainty, using Monte Carlo simulations. The fundamental purpose of computational procedure is to obtain a statistical coverage interval of the measurement model output variable. It is necessary to emphasize, that the results of calculations are correct, even if the model is strongly nonlinear and the probability density functions of output are asymmetric. This situation takes place in the thermovision measurements.The Common Commitee for Basic Issues In Metrology took this into consideration and worked out "Supplement No.1" entitled "Numerical methods of propagation of distributions" [8]. "Supplement..." deals with evaluation of precision in indirect measurements, with particular emphasis on strongly nonlinear and/or complicated measurement models, like e.g. the processing algorithm of measurement path of an infrared camera. The method of propagation of distributions makes it possible to give a correct estimation of a measurement precision, in particular in the following cases [9]:• the partial derivatives are unavailable, • the distribution of the output variable is not Gaussian, The method for the propagation of distributions evaluates the uncertainties using Monte Carlo method. The principal aim of the computational procedure is to evaluate the statistical coverage interval at a specified confidence level. It is worth to emphasize that the procedure gives correct results even for strongly nonlinear functional relationships of measurement models as well as for assymetric probability dennsity functions of input random variables. The following steps can be distinguished in the evaluation of uncertainty: The method for the propagation of distributi...