Abstract
Milk powder is typically fortified during processing by the addition of several ingredients. At the powder plant studied, one particular ingredient showed significant concentration variation in the final product, of which the causes were not well understood. Hence, a Monte Carlo uncertainty analysis was used to predict the output concentration distribution based on the typical variations of processing conditions observed in the actual plant. The purpose was to find if the observed output variability could be attributed to either variation during processing, due to a poor analytical measurement procedure, or something else. The strategy was also used to quantitatively establish if changes in the operational procedure could further reduce the variation in final product quality. The variation in final quality parameter due to processing conditions was found to be of the same magnitude as the analytical technique, meaning that both contribute to the final product variation, and hence a reduction in processing variability will reduce the overall variation. Furthermore the model suggests that the product is more likely to be out of specification on the upper limit, than on the more important lower limit given typical process variation, which was also seen in the measured results. Thus, powder out of specification above the upper limit is likely to be caused by process variation; however powder out of specification on the lower limit is not. A concrete recommendation from this analysis is that ratio control of ingredient addition was shown to significantly reduce the final product variation, and could prevent out of specification on the upper limit.
