Predicting swine odour concentrations.

摘要

Confined animal feeding operations are known to produce odours, which often lead to public complaints. Efforts to reduce the odour of confined animal feeding operations focus on changes in housing, manure storage and animal diet, but evaluation of these efforts and the development of enforcement tools require precise odour measurements. This experiment attempted to determine if swine odour concentrations could be calculated using Artificial Neural Networks and the results of measurements of ammonia and hydrogen sulphide gases and the output of an AromaScan(TM) electronic nose. It was found that a recurrent network with, Symmetric Logistic activation of the hidden layer and Logistic of the output layer could be used to predict odour concentrations to account for 79% of the variation of the concentration measurements.