Non-Destructive Testing Of Apple Qaulity Based On Electronic Nose

With varieties of’Fuji’ and ’Pink Lady’ apples as test materials,application effects of the electronic nose based on stoichiometric method in classifying apples according to the different damage degree and quality prediction of apples which were stored in low-temperature,and the electronic nose combined with Gas Chromatography-Mass Spectrometer(GC-MS)in aroma quality evaluation of apples vhich stored in low-temperature were studied.It can provide theoretical reference for the better application of electronic nose in apple quality test.The results are as follows:(1)Discriminant models for classifying different damage degree of apples were built.Electronic nose in combination with stoichiometric methods was used to classify the apples with different damage degree.Linear discriminant analysis(LDA)turns good performance on classifying different groups of treated samples,and total contribution rate of LD1 and LD2 was 98.7%.The total discrimination accuracy(DA)of stepwise discriminant analysis(SDA),radial basis function neural network(RBFN),multilayer perceptron neural networks(MLPN)and back-propagation neural network(BPNN)models for the training data set was respectively 97.5%,78.8%,99.7%and 99.1%,and for the testing set was respectively 93.8%、66.3%.95.0%and 96.3%.The total DAs for training and testing data of RBFN were all quite low,while RBFN could do better in the discrimination of apples with higher damage degree(drop from the heights of 0.5 m and 0.8 m)and the DAs were all over 80.0%.The discrimination effects of the SDA,MLPN and BPNN were all turns well,and the DAs of them were all over 93.0%.In addition,the model of MLPN and BPNN turns better discrimination effect than SDA model.The quality of the apples with different damage degree in room temperature storage could be monitored by electronic nose.Electronic nose in combination with the LDA can distinguish between the apples with different damage degree in various storage periods.Electronic nose in combination with Fisher discriminant function could be used to predict storage time for the apples of different damage degree.In calibration sets,the overall prediction accuracy for the groups of apples dropped from 0.2 m,0.5 m and 0.8 m high damage,and control group was respectively 96.7%,97.5%,98.3%and 95.8%,while in the validation set prediction accuracy was respectively 97.5%,95.0%,97.5%and 92.5%.The prediction effect for the apples dropped from 0.5 m height turned best.These showed that electronic nose not only could be used to classify the apples with different damage degree,but also could be used to predict the room-temperature storage time for the normal apples and damaged apples.This would be helpful to apples grading in postharvest links,improve the value of apple goods in storage,reduce the loss of apples in postharvest links,and provide reference for more reasonable and timely processing and utilization of normal apples and damaged apples which were stored in room-temperature.(2)Prediction models for apple storage time and quality were built.Electronic nose in combination with MLPN was able to well predict storage time of ’Fuji’ and ’Pink Lady’apples.In training sets the prediction accuracy for storage time of ’Fuji’ and ’Pink Lady’ was respectively 99.3%and 98.5%,while in the testing sets prediction accuracy was 97.4%and 97.5%,respectively.Electronic nose combined with partial least squares(PLS)models could be used to predict firmness,SSC,TA and SSC/TA of apples during storage.For firmness,SSC,TA and SSC/TA of ’Fuji’ apples,the coefficients of correlation(R2)of calibration set was respectively 0.9103,0.8599,0.9184 and 0.8322,and R2 of prediction set was 0.9047,0.8302,0.8851 and 0.8139,respectively;for firmness,SSC,TA and SSC/TA of ’Pink Lady’ apples,the R2 of calibration set was respectively 0.9181,0.8776,0.8691 and 0.8480,and R2 of prediction set was respectively 0.8860,0.8258,0.8516 and 0.8452.The R2 of BPNN model which were built for predict the value of firmness,SSC,TA and SSC/TA of ’Fuji’ apples and ’Pink Lady’apples were all greater than 0.9000 and 0.8300,respectively.The results showed that PLS method and BPNN all could predict the physical and chemical quality of apple well.PLS method turned better prediction performance on firmness and TA than SSC of ’Fuji’ apples,it turned best prediction effect on firmness of ’Pink Lady’ apples,and BPNN turned better application in quality prediction of apples than PLS method.BPNN was used to build a general model to predict the physical and chemical quality of the two varieties of apples.The R2 of SSC prediction set was only below 0.8500,while others were all greater than 0.8500.So the application effect of this general model turned well,which provided a reference for expanding the scope of application of electronic nose in apple quality prediction.Electronic nose combined with PLS method could take a good performance on predicting crispness and resilience with the R2 greater than 0.8000,while there was a poor performance on predicting cohesiveness and chewiness.(3)Electronic nose in combination with GC-MS was used to evaluate the change of apple aroma in storage.75 kinds and 78 kinds of volatile components which were respectively released by ’Fuji’ and ’Pink Lady’ apples in low-temperature storage were detected by GC-MS.They were major of hydrocarbons,aldehydes,ketones,esters,alcohols,acids,terpenes etc.,and the relative content of esters and alcohols was quite high among these volatiles.The content of volatiles showed an obviously increasing or declining trend at the 90th day of storage and an obviously declining trend at the end of storage.Apples released more esters at the 90th day of storage and more alcohols at the 120th day of storage.The content of hydrocarbon,ketone,acid,phenol and terpenes which released by ’Pink Lady’ apples at the 90th day of storage was similar to them at the 120th day of storage.The S2,S7 and S9 sensor of Electronic nose played an important role in detection of apple aroma.Optimized sensors of electronic nose in combination with LDA method could be able to classify the apples in different storage time.The aroma of ’Fuji’ and ’Pink Lady’apples all turned significant changes at the 90th day of storage,which was basically the same with the results of GC-MS.In addition,the results of sensory evaluation for aroma of ’Fuji’and ’Pink Lady’ apples were also basically the same with the detection results of electronic nose and GC-MS.Therefore,electronic nose could be used to realize the accurate and rapid non-destructive detection for apples,and the test results were reliable.