Neural network modelling and sensitivity analysis of a mechanical poultry catching system

Mechanical poultry catching equipment has been under development for a number of years. Producers on the Delmarva Peninsula have begun to move towards mechanical catching equipment. For this study, the Anglia Autoflow mechanical catching system was observed on 11 occasions from January 2002 to July 2002. The data collected were used to create a predictive time artificial neural network (ANN) of the catching process. In the study, an ANN model with a network with seven input nodes, one hidden layer of four nodes and one output node was used. The model was used to determine the sensitivity of the process to changes in the system. The system was most sensitive to changes in house configuration and least sensitive to changes in the mechanical catching system operating characteristics. (c) 2005 Silsoe Research Institute. All rights reserved Published by Elsevier Ltd.