Quality-grade evaluation of petroleum waxes using an electronic nose with a TGS gas sensor array

In this paper, the potential of an improved electronic nose to discriminate the quality of petroleum waxes based on their volatile profile was analyzed. Two datasets at 25 and 50 degrees C were collected from an experiment in order to compare influence by temperature. More fine-grained levels were further labeled for classification to meet various purposes. As petroleum waxes with lower odor levels are more difficult and important to identify than those with higher odor levels, we focus on the discrimination task for low-level ones. Principal component analysis was used for dimensionality reduction and data visualization. k-nearest neighbors, support vector machine, and multilayer perceptron were employed to classify among different qualities of petroleum waxes. The leave-one-out cross-validation method was employed due to the small sample sizes. Results showed good performance on both datasets, and at a temperature of 50 degrees C all pattern recognition methods showed improved classification rates. The improved electronic nose can potentially be applied to discriminate the quality of petroleum wax.