Detecting the different blends of diesel and biodiesel fuels using electronic nose machine coupled ANN and RSM methods

In this study, various biodiesel fuels were blended with 2, 5, 10, 20, and 80 volumes of petroleum diesel. The results were collected using an electronic nose including 8 metal oxide semiconductor (MOS) sensors. The collected data were then analyzed by the artificial neural network (ANN) and response surface method (RSM) techniques. According to the results, ANN and RSM methods were able to classify and discriminate the pure biofuels with an accuracy of 100 and 92.4%, respectively. Also, the ANN method was capable of identifying and classifying, six types of biodiesel fuels into the pure category while categorizing various types of blended fuels into another (impure) with an accuracy of 96.5%. Discrimination and identification of different blended fuels of B20 (20% biodiesel +80% diesel), B5, and B2 were done by the ANN method at the accuracy of 100%, 98.8%, and 98.8% respectively. Based on average functional parameters of the models, the ANN model exhibited better discrimination performance than the RSM model with a mean accuracy, sensitivity, and specificity of 98.8, 98.5 and 99.5, respectively.

Automated summary

This study used an electronic nose, artificial neural network, and response surface method to analyze various biodiesel and petroleum diesel blended fuels. The results showed that the artificial neural network method achieved a 100% accuracy in classifying and discriminating pure biodiesel fuels, while the response surface method had an accuracy of 92.4%. The artificial neural network method also demonstrated high accuracy in identifying and classifying different blended fuels, with accuracies ranging from 96.5% to 100%.