Implementation of improved fat standardization using statistical process control

Sammanfattning: The aim of this project was to apply statistical process control (SPC) and measure the variation of fat content in milk in order to improve the standardization, so that the fat content does not change more than 0.03 percentage units from target. Recommendations of how to adjust the standardization should also be developed. The standardization takes place together with pasteurization in one of the three pasteurizers. Thereafter, the milk goes to a common product tank with all the pasteurizers. Samples from the three pasteurizers and the product tank were collected and analyzed on MilkoScan FT2 and the fat content was plotted into Shewhart and cumulative sum (CUSUM) charts. Sampling on the pasteurizers from startup showed that samples should be taken after about 20 minutes, because then the variation is in general smaller. The data from the product tank showed a smaller variation than the pasteurizers. Because the milk from all the pasteurizers is transported into one product tank, it is impossible to know which pasteurizer that is out of control and need to be adjusted. Therefore, the conclusion is that samples should be taken after the pasteurizer and plotted into Shewhart and CUSUM charts. Action limits were achieved from the Shewhart and CUSUM charts, respectively. These are the limits that should be used to determine when adjustments of the pasteurizers are needed, and not the brand limits that are considerably wider. If the measurements fall outside the second limit in the Shewhart chart (three times the standard deviation) or outside the limits (H) in the CUSUM chart, the standardization before the pasteurizer in question should be considered. It is not known if using SPC will improve the fat content to be within 0.03 percent units from target, because the recommendation has not been applied in the process yet, but it going to be that soon.