In this study, adulteration in tomato paste was investigated using a gas sensors array. For this purpose, olfactory machine system was evaluated based on five gas sensors to investigate pumpkin, potato, and starch adulteration (0, 5, 10, 15, and 20%). Principal component analysis (PCA), linear discriminant analysis (LDA), support vector machine (SVM), and partial least square (PLS) methods were used for classification and analysis of sensors response. The use of PCA method in the classification of pumpkin, potato, and starch adulteration has explained 99% of the variance between the data. Also, TGS2610, MQ3, TGS2620, and TGS2600 sensors are widely used for classification. LDA method precision for pumpkin, potato, and starch adulteration classification was 79.07, 83.87, and 92.22%, respectively. In SVM method, C-SVM function as polynomial function and Nu-SVM as radial base function were used for pumpkin classification with 98.84 and 88.14% precision, respectively. In the classification of potato adulteration, the polynomial function had 90.32 and 83.87% precision in C-SVM and 96.77 and 86.02% precision in Nu-SVM. Finally, the two TGS2610 and MQ3 sensors had the best results for classification by PLS. Among the methods of classification, PCA method for categorizing pumpkin adulteration and PLS method in the classification of potato and starch adulteration had the best results.
Practical applicationsIn this study, an electronic nose system was examined based on five metal oxide semiconductor sensors to recognize pumpkin, potato, and starch adulteration in tomato paste in combination with pattern recognition methods including PCA, LDA, SVM, and PLS.According to results, it is recommended to use TGS2610 and MQ-3 sensors combined with PCA and SVM method in the design and development of the electronic nose system. Also, potential of the electronic nose system uses for food adulteration monitoring.