First industrial PhD student at DeLaval defending her thesis

On 5 February, our first ever Industrial PhD Dorota Anglart, publicly defended her PhD thesis titled “Indicators of mastitis and milk quality in dairy cows, data, modelling, and prediction in automatic milking systems”.

Dorota Anglart has worked at DeLaval for ten years and has during the past five years been analysing big data from automatic milking systems in order to improve mastitis detection. She defended her thesis at the Swedish University of Agricultural Sciences, Uppsala.

The PhD project is a collaboration between DeLaval and Swedish University of Agricultural Sciences (SLU) and presents interesting results. By using machine learning methods with input from data generated by DeLaval automatic milking machines, predictive models could be constructed. These findings could contribute in applications that could  help farmers to detect possible health issues.

The thesis is available here: https://pub.epsilon.slu.se/21244/



Summary of the thesis

Method

The method used in the thesis was to collect data from DeLaval milking robots, which was used as input to machine learning models. The models were trained to predict somatic cell counts (SCC) and clots in milk, two important and well-established indicators of udder health and mastitis. Previous studies have used data from sensors to predict sick cows or bad-quality milk, but few tried to predict SCC.

 

Results

Somatic cell count could be predicted with a low prediction error, and improved even further if the cow’s earlier level of SCC was known. Using system data from three days was sufficient to make the predictions, meaning that the cows does not have to be milked over a long time to predict the SCC. By using models based on data generated daily, applications could be created where the SCC could be displayed by the system without sampling the cows and as often as the user prefers. Prediction values of SCC could also be included in mastitis prediction models to improve the prediction performance.

The occurrence of clots at quarter level among cows milked in automatic milking systems was investigated for the first time within the framework of this thesis. It was discovered that longer milking intervals increased the risk of having clots in the milk, as well as the occurrence of clots in a quarter at a previous milking. Cow milkings free of clots were correctly predicted with a high certainty by their models, while the occurrence of clots was harder to predict, especially milder cases.

 

Future research

Dorota and her team hope to combine the knowledge of this research to create applications where the user can easily obtain predicted SCC for individual cows, and hopefully combined with the occurrence of clots in milk.

The research was partly funded by the Swedish Foundation of Strategic Research.