Détection de problèmes techniques à partir de données individuelles d’abreuvement de truies gestantes

International audienceBesides farmer observations, new technology such as connected troughs allows for individual and automatic monitoring of gestating sows. The aim of this study was to determine whether technical malfunctioning in the gestation room can be detected from records of individual water intake. Data were collected for 52 gestating sows during two types of induced stress events: thermic stress (hot and cold) and feeding stress, due to the closing of an automatic feeder. Data were analyzed using a smoothing method in R software (“fda” package) that can smooth data to varying degrees to detect potential disruptions. A specificity/sensitivity analysis was then conducted. Initial results showed that gestating sows drank a mean of 8.8 L of water/d, at a frequency of 14.5 visits/d and for a global duration of 9.5 min/d. One to four perturbations per gestation were detected by the smoothing methods. The specificity analyses highlighted that the event-detection method was more than 95% specific and thus able to detect nearly all true negative individuals. But the sensitivity was very low (9.4%). In conclusion, too few technical malfunctions can be detected from records of individual water intake. The method needs to be more accurate; one solution could be to apply it simultaneously to more variables