In recent decades, human activities have introduced various pollutants into the environment at significant levels. Among these pollutants, oil pollution poses considerable threats to both human health and the environment. Therefore, the detection and remediation of oil pollution have become critical tasks. Numerous methods for detecting oil pollution in soil have emerged, including gas chromatography, mass spectrometry, and electronic nose technology.
In this study, we utilized an electronic nose system equipped with eight metal oxide semiconductor sensors from the MQ and TGS sensor series to detect and assess soil samples contaminated with crude oil. The soil samples were obtained from the agricultural fields at the University of Tehran's Faculty of Agriculture and Natural Resources. The samples were deliberately contaminated with crude oil at four concentrations: 200 ppm, 500 ppm, 1000 ppm, and 5000 ppm. Notably, the 1000 ppm concentration represents the permissible limit, while the 5000 ppm concentration leads to soil destruction. We collected two samples from each concentration, totaling eight samples, which were stored and tested over a 14-day. The sensor responses were transformed into voltage signals and normalized using the differential method. The pattern recognition methods yielded results of 92% accuracy for Principal Component Analysis, 100% for Linear Discriminant Analysis, and 100% for the Support Vector Machine method. Additionally, we compared the electronic nose system's results with those of the gas chromatography system, demonstrating the efficiency and speed of our method in classifying and distinguishing different concentrations of oil pollution. Also, in this research, a method was used, with the help of which the sensors can be examined separately and the accuracy of each can be obtained using machine learning algorithms. The factorial experiment was conducted with two main factors, day and time, and for analysis, 80% of the data were selected for training and 20% for testing, and decision tree regression, support vector regression, enhanced gradient regression, and random forest regression algorithms were used. After obtaining the results, it was found that the MQ3 sensor has the highest accuracy and the lowest error rate, With the best performance with an average of 95.5%, they show the detection rate of pollution levels. Therefore, the selected sensors from the results obtained from the gas chromatography system research on soil contaminated with crude oil have had an acceptable sensitivity, and in general, the electronic nose system can be replaced by other methods as a cheaper and easier method.