| Corn juices contain dietary fiber and essential iron, calcium, selenium, zinc, potassium,magnesium, manganese, phosphorus, glucose, amino acids, and so on. And corn juices alsohave characteristic aroma of corn, refreshing taste, appropriate sticky and convenience to eat.With the development of the corn juices market share, the quality control and sensory qualityhomogeneity garner more attention. At present, the detection of organoleptic attribute forcorn juices were accomplished by human which a large amount of manpower and materialresources were expended. Also physiological and psychological factors were influencefactors for the conclusion of sensory evaluation. With corn juices gradually startinglarge-scale production, the feedback of available detection methods is too long which isagainst real-time control on the production line. That, an automatic detection method whichis live, rapid and could adequately reflect all the sensory evaluation conclusion is urgentlyneed. In this paper, the taste sensor arrays and intelligent detection model based on fuzzy setswere built for corn juices taste and analog flow testing.1. A taste sensor array was built, which consists of2glass electrodes,7insoluble salt(solid) electrodes,3liquid membrane electrodes. According to the certain difference of themechanism and the component of sensitive membrane structure of the working electrode, theactivation parameters and stable state were determined. Signal of five basic taste substanceswere collected, which indicate sour, sweet, bitter, salty, umami. First three principalcomponents with89.915%cumulative variance contribution rate were extracted using theprincipal component analysis method. PNN network for the identification of basic taste wasbuilt with the first three principal components as the network input neurons. The taste sensorarray was good at recogniting basic taste substance. Basic taste identification rate was100%by trained PNN network.2. The response signal of corn juices was acquired based on the completed taste sensorarray. And the signal was analysed. Based on the completed taste sensor array, the responsesignal of corn juices was acquired and analysed. The principal component analysis methodwas used to reduce the dimension of the overall data. The first three principal componentswere extracted, the cumulative contribution rate of which was87.376%. Analysed the signalof corn juices by cluster analysis, the same corn juices can be clustered into one group whenthe average distance was3.44. Then the whole data was analysed by PNN. The first threeprincipal components were input neurons. And output layer neuron number was9based onthe type of corn juices. The identification results of the PNN showed that the classificationrecognition accuracy was95.06%. The sensor array was optimized based on the effect ofidentification for corn juices. The intrinsic link between the variables of sensors wasexpressed by factor analysis. Thus the sensors were classified as follows class, G1, S4, S5asthe first class, P1, S2as the second class, P3, S3as the third class, P2, S6, S7respectivelyfor three classes. Different sensor array combinations were acquired under the premise of containing all the category information. PNN was built with0.01smoothing parameter. Thesensor array which obtained optimal classification results was acquired based on thedifferences in accordance with the classification results. The sensors respectively named S4,S2, P2, P3, S6, S7were selected to build a new sensor array which was sharply reduced thenumber of sensor. The classification accuracy rate98.016%of optimized sensor array forcorn juices effectively improved the detection accuracy.3. For the complete representation of fuzziness and randomness of corn juices sensoryevaluation because of the sample differences, physiological and psychological factors, amathematical model which freely swithed between the concept of qualitative andquantitative values was built. A one-dimensional forward normal cloud model was used toexplain the various cross-sectional view of corn juices standard. And this kind of model wasimproved based on the actual situation. The descriptive language in rating standard wastransformed to a corresponding quantitative numerical representation. Sensory evaluationtest value was transformed to qualitative concept based on X reverse cloud model. And thecloud model was extended to multi-dimensional. The concept of qualitative characteristics(integrated expectations Exz, the integrated entropy Enz, entropy Hez) was used to reflectthe characteristics of the overall concept. The difference of sensitivity of corn juices tastewas analyzed based on the sensory evaluation conclusions qualitative and quantitative cloudconversion model. The mapping between the conclusions of the cloud model and sensorsignal was completed by fuzzy neural network. Fuzzy neural networks were built forprediction of fuzzy information in corn juices taste. The information for different aspectscollected from sensor array was input. The information from cloud model according tosensory evaluation was output. With training fuzzy neural network, adjustiong fuzzy layercenter value, the fuzzification layer node width values and fuzzy decision-making regulationparameters were obtained to determine the network structure. The forecast analysis showedthat the system allowed good effect with0.00243-0.0918error percentage in the process ofautomation evaluation of fuzzy information for corn juices.4. A detection device of corn juices was built. The device contained hoisting devices,horizontal rotary motion device and automatic control system. The sensor arrays couldimmerse in the fluid state of corn juices to collect data automatically. It can also resliseelution and continuous detection. The kinematic scheme of the transmission mechanism ofthe screw nut was selected in elevating linear motion. It contained screw, screw nut, thesupport mechanism, the coupling, stepper motor, the support guide plate and so on. The maincomponents of the rotating mechanism were support round table, rotating spindle, worm gearunits, couplings and stepper motor and so on. The control system was consisted of thetwo-phase hybrid stepping motor (57HS7630A4) and ancillary drive (MB450A). The motioncontrol chip TMC429was selected to control stepper motor. The detection sensor array forundiluted state of the flow pattern of the corn juice consisted of sensor S1, S3, S4, S5, S6and S7. Eluted effect of detection sensor array was analyzed. The difference of deviationbetween the the blank potential and initial equilibrium state scanning potential was from the range of0.008655to0.077657. The corn juice flow dynamic simulation test system wasdesigned and construction. The system showed good identification for dilutionless flowpattern corn juices with identification accuracy85%.In summary, the intelligent detection methods research based on sensor array provided atheoretical basis and technical support for the automatic control of the sensory quality of thecorn juice produciontion process. It was achieved that the high accuracy for corn juices. Andautomatic detection of taste was achieved in the fuzzy level. An automatic detection systemof corn juices in simulated flow dynamics was built. |
PDF Full Text Request