Fast Detection Freshness Of Fish Meal Based On Eletronic Nose

Fishmeal is an important animal protein feed, which is rich in protein and vitamins and is a major raw material of animal feed. Its quality directly affects the health of animals and the safety of their products, further impacts on human health and safety. Because of the high content of protein and fat, fishmeal is easy to spoilage and oxidative rancidity, which leads nutritional value decreased and affects the safety of feed products. Freshness of fishmeal is an important indicator of metamorphic grade. The current method of testing the freshness of fish meal has been unable to meet the requirements of rapid tests, so to carry out the study on freshness rapid detection of fish meal is particularly important.The electronic nose measurement system for freshness detection of fishmeal was developed based on the gas composition in the progress of fishmeal spoilage and virtual instrument technology. On this basis, the impact parameters of sensors response and pattern recognition were studied, and the regression models between volatile basic nitrogen (TVB-N) and acid value content and electronic nose data were established, as well as freshness discriminant models. Major research results were as follows:1. TGS822 of organic solvent vapors, TGS825 of hydrogen sulfide, TGS826 of ammonia and TGS832 of chlorofluorocarbons assembled to sensor array of electronic nose based on the gas composition in the progress of fishmeal spoilage. Signal pick-up circuit and other hardware components of measure system were designed. Information acquisition of fishmeal smell, data storage and signal display processing software based on virtual instrument were developed. Different freshness fishmeal samples were acquired by electronic nose. The results showed that sensors'response changed with the freshness of fishmeal.2. The optimum characteristic value, optimum sample weight and acquisition time respectively were determined to steady-state value of sensor response, fifty gram and five minutes based on principal component analysis. The affect of sample temperature and moisture on the sensor response were carried out. The results showed that sample temperature and sample moisture content have a significant impact on the sensor response(P<0.05).3. Different freshness samples of fishmeal were acquired by electronic nose and volatile basic nitrogen (TVB-N) and acid content were determined by chemical method. Models of PCR (Principal Component Regression), MLR (multiple linear regression), BP-ANN (back propagation network) and SVR (support vector regression) between electronic nose data and chemical indicators of freshness were established. And their were validated by prediction set. The coefficient of determination R2 between predictions of models and measured value were 0.48,0.59,0.94,0.91 (TVB-N) and 0.69,0.76,0.97,0.97 (acid value). The results show that BP neural network and SVR model were better.4. Principal component analysis (PCA), BP neural network and support vector machine (SVM) discriminant models of fishmeal freshness were established. The results showed that principal component analysis was more difficult to distinguish fresh and spoilage samples of electronic nose, while the recognition accuracy rates for fresh and spoilage samples of BP neural network and support vector machines respectively were 96.77% and 100%. Keywords:electronic nose; fishmeal; principal component analysis; BP neural network; support vector machine...