fabrication and development of a machine olfaction system combined with pattern recognition techniques for detecting formalin adulteration in raw milk

adulteration in milk and other dairy products not only is a serious threat to human health but also leads to the economic losses in the dairy industry. utilization of the materials reducing microbial load is a common adulteration. in this study, a machine olfaction (electronic nose) based on 8 metal oxide semiconductor (mos) sensors was fabricated and developed and its capability to formalin detection in the raw milk was investigated. feature vector was then extracted from the sensors’ response and used as the inputs to pattern recognition models. based on the obtained results, principal component analysis (pca) with two first pcs (pc1 and pc2) could describe 93 % of variance within data. in the sensor array, mq4, fis, tgs822, and tgs2620 sensors had the highest loading coefficient values whilst tgs2602 devoted the lowest loading value. linear discriminant analysis (lda) revealed the classification accuracy as 80.1 %. support vector machine (svm) with three order multinomial kernel function showed the training and validation accuracy values as 100% and 90.91%, respectively. also, the full success rate was obtained for overall classification using the artificial neural network.