The application of fuzzy logic in programs for mastitis detection in dairy herds

An important factor influencing milk production is the health of the animals. In quality assurance programs for milk, the health of the cows is the most outstanding parameter. In the case of Automatic Milking Systems (AMS), the udder health becomes especially important due to the lack of visual control by the milking staff. On the other hand, AMS facilitate the opportunity to measure a lot of parameters of the milk and the milking process specifically for the single quarters of the udder. From this collected data, conclusions can be drawn towards possible udder diseases. The objective of the investigation was to elaborate an analysis system serving to supervise the udder health, using the concatenation of selected milk- and milking process parameters. Fuzzy-Logic was selected as an appropriate method of dealing with vague data and diffuse relations. Based on the data of a thorough investigation into a dairy farm with two AMS, three different methods of modeling were selected, analyzed and compared with one another: models of a single parameter (threshold models), models of indices and Fuzzy logic. While the threshold models showed higher values of specificity, the best results of sensitivity were calculated for the models of indices. However, the optimal result was obtained with Fuzzy logic. The corresponding modeling featured the smallest statistical probability of wrong classifications (5.9 percent of incorrect diagnoses). The results confirm the hypothesis that Fuzzy logic is a manageable and effective method to uncover the relations between parameters of milk and the milking process as well as the udder health.