Ordinary least squares method: A strategy to reduce the mass measurement uncertainty

Abstract

This research was motivated by technical-economic challenges imposed by mass metrology, in particular on issues concerning calibration methods of non-automatic weighing instruments (i.e.: scales). In the contemporary world it is vital to ensure the reliability of measurement results in a test, testing and research laboratory. These results must be reproduced by institutions –—National and International— considering the uncertainties associated with the process. Two of the factors that strongly affect the metrological reliability of the mass measurement processes are: (i) non-frequency of calibration of scales; (ii) analytical methods for the treatment of experimental data obtained in the calibration. In this context, and maintaining the principles for the calibration of a measuring instrument in which it must be performed under the same operating conditions of the instrument, metrology laboratories are in need of knowing metrological reliability (i.e.: errors and uncertainties) for each calibration scenario. In this order of ideas, this study sought to evaluate the incident factors in the measurement of mass and calibration of non-automatic weighing instruments. The applied methodology (ordinary least squares method) allowed estimating the metrological reliability of a Digital Scale (Max. Cap.: 30 kg; Res.: 0.001 kg) at 0.012% (k = 2). The results confirmed that the applied method allowed to reduce the adjustment uncertainty up to 92.0% from the use of a fourth degree polynomial when compared to a model widely used by conventional metrology laboratories, i.e.: first degree polynomial to from a simple linear regression.