Madsen : Monitoring Daily Gain in the Slaughter Pig Production Monitoring Daily Gain in the Slaughter Pig Production Unit Using Kalman Lter Techniques

This contribution discribes a new method for monitoring productivity in the slaughter pig production unit. The method is implemented with a graphical user interface. By calculating average daily gain each time a group of slaughter pigs is delivered to the slaughterhouse, a time series is made. Among other features, the application allows the user to monitor the time series graphicaly. Monitoring data from various herds has shown some uctuation in the series, therefore the user may optionally smoothen the graph by means of a Kalman lter technique. Other production traits like temperature, humidity and stocking density may be selected interactivly for display in the same window as daily gain. By comparison it is, in some cases, possible to detect the reason for a decline in the estimated daily gain. 1 Objectives The traditional way of calculating average daily gain is to deene a valuation period (usually 3 months) in which all animals that enters or leaves the barn are registered by date, weight and action (enter or leave). Based on these registrations and a valuation weighing at the begining and at the end of the period, the average daily gain is estimated. This method has some disadvantages, though. First of all, it implies some considerable manpower to weigh every animal in the herd. Secondly, the usual valuation period is 3 months, which means that a decline in average daily gain is detected 1 3 months after its actual occurrence. The delay of information means that this way of estimating daily gain is diicult to use as a tool for herd management. Finally, a 3 month average could, in some cases, cover up considerable uctuations in productivity. More information can be obtained by reducing the valuation period length. This does, however, require more frequent (labor demanding) valuation weighing. Furthermore, a shorter valuation period implies higher eeect of measurement errors.