Comparison of total volatile basic nitrogen detection models in fishmeal based on electronic nose

Total volatile basic nitrogen (TVB-N) is one of main characteristics of fishmeal freshness. In order to discuss the feasibility of fishmeal freshness detection based on electronic nose, an electronic nose was developed for the detection of TVB-N of fishmeal. It mainly consists of SnO2 gas sensors, portable data acquisition device and data acquisition program based on LabVIEW. Different number of storage days of fishmeal were detected by electronic nose and semimicro-kjeldahl determination. The value of gas sensors response were analyzed by 2σ criterion and principal component analysis. Then principal component analysis model, multiple linear regression model and back propagation neural network model of TVB-N mass fraction was created. And they were validated by prediction set. Coefficients of determination, standard error of prediction, maximum relative error and mean relative error between predicted TVB-N and measured one of these models were 0.48, 10.25, 13.62%, 6.06%; 0.59, 9.14, 13.91%, 5.57%; 0.94, 3.64, 6.30%, 1.88%, respectively. And BP neural network was the best method comparing to PCA (principal component analysis) and MLR (multiple linear regression), and MLR was better than PCA. The results showed that it was effective for TVB-N mass fraction rapid detection in fishmeal based on electronic nose.