Application Of Smartnose And Smartongue Based On SVM Algorithm In The Edible Vegetable Oil’s Quality Detection

There are many methods for the detection of edible vegetable oil, such as physical and chemical method, chromatography, NMR spectroscopy, and so on. These methods are basically for the detection of a single substance in the sample, it is difficult to make a systematic and scientific comparison with human senses although it has a good application in the detection of edible vegetable oil. Now the factory main detection method of edible vegetable oil is sensory evaluation, it is difficult to popularize in the quality detection of edible vegetable oil due to the cultivate of professional sensory personnel is hard and the detection result is unstable. What we need is a simple、effective detection method which can be promoted easier. Smartnose and Smartongue are simple and efficient instrument which can be detected the integral feature of the test samples. This work is used the Smartonse and Smartongue dased on SVM, cooperated with the Luhua company, to distinguish and identificate edible vegetable oils from the smell and taste. The works get the following conclutions.(1) The Smartnose based on SVM algorithm:change the cleaning gas of Smartnose from the filtered air to standard air; increase the water bath temperature of edible vegetable oil from 30℃ to 90℃. Optimization of these experimental conditions increases the discrimination of edible vegetable oil. In the edible vegetable oils detection of Smartnose, SVM classification model accuracy rate is 99%,and significantly higher than the detection accuracy rate of 86% of LDA model. In the detection of different quality peanut oils, the predicted error of SVM regress model was lower than 5%, which was significantly lower than PLS model, and the detection of unknown quality of peanut oils are about 5% too. From the results, the Smartnose based on SVM algorithm can be applied to the detection of edible vegetable oil, which provides a basis for the quality control of edible vegetable oil’s production process.(2) The application of saponification make Smartongue applied to the detection of edible vegetable oil. After optimization of the saponification conditions, Smartongue not only can distinguish different kinds of edible vegetable oils, but also can distinguish different quality of peanut oils. In the edible vegetable oils detection of Smartongue, SVM classification model accuracy rate is 94.58%, although higher than the detection accuracy rate of LDA model(81.25%), but not good as Smartnose SVM class model(99%); In the detection of different quality peanut oils, the prediction error of SVM regress model was corresponding lower than PLS model but higher than SVM regress model of Smartnose. The Smartongue of SVM class model is applied to camellia oil, olive oil adulteration test, with 80% limit, SVM class model can detected if the adulteration rate is higher than 20%. From the results, the Smartongue based on SVM algorithm can be applied to the detection of edible vegetable oil.(3) The data fusion of Smartnose and Smartongue, weakened the advantage of them. Detecting edible vegetable oils with the combination of Smartnose and Smartongue, the accuracy of SVM class model is 69%, the predicted error of SVM regress is higher. Because of different detection mechanisms of Smartongue and Smartnose, leading the material act on the sensor is different, the superposition of the two kinds of data may cover the characteristics of the original data.In conclusion:Smartnose and Smartongue based on SVM models have a good application in the identification and quality control of edible vegetable oil. The advantage of Smartnose and Smartongue is rapid, simple and intelligence, so that it can get rapid promotion in the society especially in the factory. If we can instead of sensory evaluation method with Smartnose and Smartogue, and application to edible vegetable oil production process, will save a lot of manpower and material resources, improving the economic benefit.