Sources of Uncertainty: Process and Measurement Variability

Suppose that we have a production process which is monitored by taking samples and measuring the parameters of interest. The results of these measurements are not constant, they are subject to uncertainty. Two main factors are responsible for this uncertainty: changes in the process itself (process variability) and measurement errors (measurement process variability) . Suppose, for example, that we are interested in controlling the magnesium content in steel rods. In the normal course of rod production, the magnesium content will vary due to variations in the chemical content of raw materials, “normal” deviations in the parameters of the technological process, temperature variations, etc. So, even an “ideal” laboratory which does absolutely accurate measurements would obtain variable results. In real life there is no such thing as an “ideal” measurement laboratory. Suppose that we prepare several specimens which have practically the same magnesium content. This can be done, for example, by crushing the steel into powder and subsequent mixing. The results of the chemical analysis for magnesium content will be different for different samples. The variability is introduced by the operator, measurement instrument bias and “pure” measurement errors, i.e. the uncertainty built into the chemical measurement procedure.