Design Of Bionic Electronic Nose And Its Application In The Chicken Quality Detection

The quality of food is the most important issue concerned by the food industry and consumers. China is a big country in chicken production and consumption. With the growing attention of food safety, meat quality has got more attentions. However, the current main evaluation methods of chicken quality are still sensory testing and chemical detection. Sensory testing wastes time and effort, and the result can easily be impacted by subjective factors. There are many chemical analysis methods. But the main problem is that they are cumbersome, tedious and expensive. And also the results sometimes can not be accurate enough. In recent years, the development of electronic technology makes it possible to exactly test the material properties of smell and qualitatively and quantitatively evaluate the intrinsic nature of smell vector by using the sense of smell bionic technology. As China's growing meat production-scale, there is the objective need to extract smell information and get the final result rapidly and steady through simulating the developed olfactory system of human being during testing meat quality.Based on this background, the paper carried out the research about the quality detection for the chickens with the Bionic Electronic Nose. The technology of the Electronic nose is relatively mature in the food detection, but it does little research to the detection of the qualities of the chickens. This paper has built an electronic nose system and used it to detect the quality of the chickens. Through forecasting the different stored times for numbers of fresh chicken samples, the better method of the neural network recognition was obtained. As the change of storage time, the relations between the electronic nose response signal and meat volatile base nitrogen were discussed, and the chicken corruption threshold and the neural network methods for chicken fresh classification also were established.Bionic electronic nose made of five parts, which contains gas room, sensor array, signal conditioning circuit, data acquisition system, computer, was designed. Using the software CATIA the bionic model of human being's nose was built. The characteristics, connectivity and learning rules of a neural network were analyzed and the RBF neural network was taken as a pattern recognition approach. Through training, the ideal parameters were determined.In this paper, the chicken breasts provided by DEDA Ltd. in Jilin Provience were taken as the research subject. The test found that a single sensor can not distinguish the different chicken. However, the sensor array showed the different response and sensitivity between fresh chicken and rot chicken. And the output signal in the near chicken corruption showed the larger stage of the overall volatility. Through analysis of sensor's skewness, peakness, coefficient of variation and correlation coefficient on the basis of experimental results, a better sensor array was established.Network training adopted a 10% cross-certification training program, the expected result was shown as the binary with 3 digits, the number of test training was set for 5000, and the permitted error range was about 0.001. Taking the average from 29,999 numbers of each set, 360 results as sample data were received.By taking the advantage of RBF net we discussed the limits of rotted chicken, detected the freshness of five teams chicken in two adjacent days. Then you can easily separate the net by forecasting when is the lowest error point come. This mean the limited value turn out when the biggest difference produced.Physical and chemical indicators verified the correctness of the experimental guess that the volatile odors from the rotting chicken will be dramatic changes. At the same time fresh chicken and rotting chicken were classified with an above 75% accuracy rate, and the classification of chicken with the different freshness degree was made with a more than 90% accuracy rate.