Transactions of the ASAE vol:47 issue:5 pages:1757-1764
Controlling the growth trajectory of broiler chickens is a possibility to reduce the negative growth responses related to fast growth, such as increased body fat deposition, a decrease of reproduction capacity, metabolic diseases, a high incidence of skeletal diseases, and sudden death syndrome. In previous research, a growth control algorithm for broilers was developed based on a compact adaptive growth prediction model, assuming a linear relationship between cumulative food intake and animal weight. Because the growth response of broilers to cumulative food intake is a non-linear process, this article investigates the validity of the assumption of linearity in an adaptive modeling approach in terms of prediction accuracy. The dynamic growth response of the broiler chickens was modeled and predicted using a time-variant parameter estimation procedure. A recursive non-linear model was used to estimate the model parameters and to predict the growth response every 24 h based on a fixed number of actual and past measurements. Tests were performed on 43 data sets. A comparison was made with the prediction accuracy of the recursive linear modeling approach. The non-linear modeling approach made it possible to predict the growth of the broiler chickens up to 7 days ahead with a mean relative prediction error of 5%, or less. This non-linear model reduces the prediction error significantly with a maximum of 1.5% in comparison to the linear modeling approach. It can be concluded that for growth control purposes the dynamic growth response of broiler chickens is slightly better modeled assuming non-linear dynamics in a short time window.