Rapid Identification Adulteration Beef By Electronic Nose Based On Data Processing Methods

Adulteration beef has been one of chronic illness in food safety. Different kinds of adulteration methods not only harm the interests of consumers, but also a threat to food safety.In recent years, many reports about the adulteration of beef have been published. For example,Adulteration beef was made of pork and beef flavor as well as bovine blood. Cheap meat such as pork, beef and chicken and even deteriorated meat were adulterated in beef.The adulteration beef was detected by E-nose which is one of rapid and non-destructive device combined with statistical theory. The average method and cluster analysis were used to extract the characteristic value. Discriminator analysis method results indicated that cluster analysis produced better accurate results. Linear regression analysis was used to quantitative analysis and multilayer perceptron neural networks were used for pattern recognition.(1) The electronic nose was used to detect adulteration beef adulterated with pork and adulteration beef which was made of pork and beef flavor. As can been seen from the line chart, The differences between samples is the second and seventh sensors. Samples which have beef and beef flavor have a significant respond about the second and seventh sensors.The results showed that electronic nose qualitatively detect the adulteration beef the second and seventh sensors. The second sensor has a wider detect scope, for example nitrogen oxides are the main objective. The seventh sensor detects the sulfide mineral, many sulfur-containing organic compounds and terpenes, pyrazines. Beef flavor contains pyrazines and Beef volatile components have many Sulfur-containing organic compounds.(2) As can been seen from the figure of discriminate analysis, the electronic nose can distinguish beef and adulteration beef.Using E-nose to detect the adulteration beef is feasible.By comparing the results of the discriminator analysis, cluster analysis is better method than area values to extract the feature data. Qualitative judgment of Linear discriminator analysis is better than qualitative judgment of the main component analysis of discrimination.(3) Electronic nose was used to detect adulteration in different ways under the quantitative analysis. Predictive model for the pork, beef flavor content in minced beef built by PLS, MLR and BPNN were obtained the R2 0.9 and the lowest prediction error within 5%. Neuralnetwork forecasting model is better than partial least squares and multiple linear regression analysis.(4) Multilayer Preceptor neural network was used to pattern recognition. Correct classification rate and Correct classification rate of the training set reaches 97%.In conclusions, the electronic nose based on the cluster analysis, linear discriminator analysis and neural networks has the capable to detect adulteration beef qualitatively and quantitatively.