Detecting cow anomalies with precision dairy farming techniques: application to some case studies in the Po Valley

The increasing interest in precision livestock farming (PLF) for dairy cows allows researchers to collect data from milking robots, collars and pedometers. This makes it possible to develop numerical models based on big data. This work presents two numerical methods for detecting daily milk production anomalies, applied to 15 different case study farms located in the Po valley area. Both algorithms take as input daily resting time and daily milk yield. These methods utilize statistical calculations based on data from pedometers and automatic milking systems. Data are recorded respectively by the pedometers worn by the cows and by the automatic milking system of the barns. The first method considers two indicators: Difference in Relative Production (DRP) and Daily Rest Time (DRT). DRP measures the relative difference in daily milk yield compared to an ideal trend. An anomaly is identified if both DRP and DRT meet specific conditions for a specific cow and day. The second method improves the reliability of the lactation curve by incorporating robust statistics and the Wood function, the most used model for cow production. By reducing the impact of outliers, a more accurate representation of individual cow performance is obtained. These numerical techniques provide valuable tools for detecting milk yield anomalies and supporting effective animal management practices.