PubAg

Main content area

Real-time modelling of individual weight response to feed supply for fattening pigs

Author:
Peña Fernández, A., Norton, T., Youssef, A., Exadaktylos, V., Bahr, C., Bruininx, E., Vranken, E., Berckmans, D.
Source:
Computers and electronics in agriculture 2019 v.162 pp. 895-906
ISSN:
0168-1699
Subject:
feed composition, feed conversion, finishing, genetic traits, models, monitoring, nutrients, prediction, regression analysis, swine, swine feeding, time series analysis
Abstract:
Precision feeding is a promising technique to achieve better feed efficiencies for livestock and realise production results closer to the genetic potential of the animal. Continuous and automatic monitoring of growth of individual finishing pigs is an essential element of a precision feeding system. For optimal pig growth, the feed nutrients need to be adjusted at different moments in time throughout the fattening period. In an integrated system, the key element is the prediction of the process output (weight) to a variation of the process input (feed supply). During the course of this study, three experiments were performed, with 80 pigs on average in each of them, to measure and model their growth responses to step changes in feed amount and/or feed composition along the production cycle. The individual dynamic responses of pig growth to feed changes are monitored by gathering daily weight and feed supply data for each individual pig. The time-series data is analysed using Transfer Function (TF) and Dynamic Linear Regression (DLR) models. On one hand, the average TF model fitting agreement is (RT2=94 ± 4)%. On the other hand, the Mean Relative Prediction Error (MRPE) of applying the DLR approach with a window size of four and seven days was found to be, MRPE = (1.0 ± 0.4)%, and MRPE = (3.3 ± 1.3)% for a forecasting horizon of one and seven days, respectively. Moreover, a parameter of the DLR model resembles the feed efficiency of the pig, exhibiting a coefficient of correlation, r = (0.9 ± 0.2). While the current models may need further validation, the approach seems promising to be implemented in an individual integrated pig feeding system.
Agid:
6449244