Development of novel sensors for online milk quality and cow health monitoring

In Flanders, dairy production is an important segment of the agricultural production, representing about 15% of the total market value. Over the last 50 years, the average milk production per cow and per lactation has increased drastically as a result of genetic selection and improved feed and management practices in dairy farming. However, due to the intense focus on high milk production, modern dairy cows are prone to production-related disorders. As the production of milk is a dominant factor in the metabolism of dairy cows, involving a very intense interaction with the blood circulation, the extracted milk contains valuable information on the metabolic status of the cow. Therefore, regular analysis of the produced milk is considered the most efficient way to monitor cow health. Today, online measurements of the milk production, conductivity and color is already common practice in dairy farming. Nevertheless, these milk parameters are influenced by many factors next to the health status of the cow. For this reason, there is a need for monitoring parameters which have a more direct link with the cow’s health. The formation of milk components, such as milk fat, protein and lactose, is the immediate result of the cow’s feed uptake and metabolic status, as well as the health of the milk producing cells in the udder. As a result, regular analysis of the basic milk components for each individual cow can give valuable information on its udder health and metabolic and nutritional status. In this project, an online milk analyzer prototype will be built and implemented on an automatic milking system. This sensor will monitor the milk quality on the level of udder quarters to be able to eliminate the systemic effect of the animal through inter-quarter comparison, supporting a high performance warning system for identifying mastitis. Furthermore, the robustness of this sensor will be evaluated and improved by applying different calibration strategies. Finally, the variation of the sensor measurements will be studied in relation to the cow’s health and combined with advanced data-processing techniques to obtain an early-warning system.