Resumen

In the experimental calibration processes, several error of measure can be obtained. Usually, these values are called as outliers. The outliers impact in the instruments’ metrological reliability and, consequently, in the industrial processes’ quality control. In specific terms, for mass metrology, the outliers are extremely critic, due to sensibility of technology used for this industry, i.e.: non-automatic weighing instruments (scales). The EURAMET/cg-18/v.02 is the international standard that defined procedure and tests for to assessment the metrological reliability of scales, however, a statistical treatment for outliers is not defined. In order to propose robust alternatives for outliers analyses, this work to evaluate parametric techniques for the elimination of outliers in the calibration of a Bernalo non-automatic weighing instruments (Capacity: 30.000 kg; Resolution: 0.001 kg). Three methods were applied in the analysis of experimental data: Dixon, Grubbs and Chauvenet. The results confirmed a reduction in the expanded uncertainty associated to mass measurement (k=2) up to: 17.6% (Dixon); 19.9% (Grubbs) and 35.2% (Chauvenet). According to the results and considering the contribution of work to applied mass metrology, the authors propose the inclusion of analyzes and evaluation of outliers in the EURAMET/cg-18/v.02. Thus, it is expected an increase in metrological reliability in mass measurement processes at the scientific and industrial sector.