| Popis: |
A prototype real-time system was developed for the control of broiler growth and nutrition intended for commercial use. All of the experimental work was carried out on a commercial broiler farm with eight identical modern houses of 30 000–40 000 birds. Each house was fitted with a commercially available broiler nutrition management system, which automatically measured the feed intake and bird weight, and provided nutrition control by blending two feeds according to the specifications given by the human operator. A semi-mechanistic growth model was developed, based on established models and principles, in which growth is predicted from feed intake and feed composition. The controller first attempts to improve the prediction of the growth model using feedback from past data from the house it is controlling. It does this by optimising a common digestibility parameter. The controller then determines the nutrition for the remainder of the growing period. It optimises feed composition, and optionally the required feed intake, to minimise the root mean square error (RMSE) between the target and predicted growth curves. The model adaptation procedure gave excellent results for healthy birds. For example in a typical validation experiment, which included a step change in nutrition, the range of RMSE for the eight houses was 16·1–109·2 g before adjustment and 15·4–34·3 g after. When operated as a closed loop, off-line controller, updating the control variable three times per week, with a 24-h lag due to the logistics of the operation, results comparable with a human manager were attained. A fully automatic product has now been developed and is undergoing commercial testing on several farms. |
Full Text from ScienceDirect