Rapid Detection of Mixed Odor Intensity Level in Composting Plants Based on E-Nose

Abstract

It is essential to be able to precisely identify the intensity level of the malodor in order to have a clear perception of and strong control over the pollution level of mixed malodor emissions. In this study, mixed malodors from a composting facility were utilized as test samples, and the malodor samples were analyzed and evaluated using data from the triangle odor bag method and the electronic nose instrument. Next to the extraction of the feature values from the mixed malodor response signals recorded by the electronic nose, dimensionality reduction was carried out using both principal components analysis (PCA) and linear discriminant analysis (LDA), and classification and identification were carried out in combination with support vector machines (SVMs), decision trees (Tree), and ${k}$ -nearest neighbors (KNNs). A good way to correlate malodor concentrations with human sensory olfaction is to use the correlation between the response signals of the electronic nose to mixed malodors and the results of the artificial olfactory dialectic method, according to the experimental results, which show that the training accuracy using KNN combined with LDA reaches 96.4% and the validation accuracy reaches 95.7%. Give them a more intuitive sense so the personnel at the cattle and poultry farms can better manage and contain the stench’s spread and give directions that work.