M.A. Thesis:

nalysis the zucchini osmotic dehydration process using signal processing and data mining

One of the non-destructive tests developed to determine some food attributes is sound Analysis and signal processing resulting from impact. In this research, from the recording system of the acoustic response resulting from the impact during evaluation of the process of juicing zucchini using the osmotic method, different shapes of zucchini samples (cylinder, cubic, cuneal) in different ratio of zucchini to the osmotic solution (1/2,1/4,1/8) were used in osmotic solution containing sucrose(5%,10%,15%w/w). for this purpose, the sound of hitting the container was recorded in three stages(0,4,8h). also, in order to record and extract sound characteristics (non-destructive test), different directions of blows to the container containing sucrose solution(down, center, up) were used as variables.Variance analysis of water loss, solid gain, weight reduction and moisture content data showed that all the independent variables used, except impact, had a significant effect on the researched characteristics. Regarding the data analysis of the sound signals resulting from blows to the container containing the osmotic solution, it was determined that the linear and interactive effect of the sample shape, impact time, concentration, the ratio of zucchini to the osmotic solution, the impact location at a level of more than 99/9% on most of measured signals were significant.The results of signal processing and Data mining Indicated That, The division tree and artificial neural network respectively, succeeded in correctly predicting the shape of 770 and 899 samples out of 1380 samples respectively. The decision tree and the artificial neural network were able to correctly predict the shape used in the search with 68 and 77percent accuracy.the scatter matrix obtained by applying decision tree and artificial neural network to predict the ratio and impact location used in the search, respectively succeeded in correctly predicting 879 and 1052 samples, 922 and 1069 samples out of 1380 samples. In addition scattering matrix and evaluation criteria were able to predict the ratio and impact location with 76%and 84% , 78%and 85% accuracy according to the features extracted from sound signal with decision tree and artificial neural network.and finally The scatter matrix obtained by applying the decision tree to predict the impact time was 705 samples out of 1380 samples. In addition, scattering matrix and evaluation criteria could predict the impact time with 67% accuracy acording to the features extracted from sound signals.

Keywords: Keywords: analysis, drying process, osmosis, analysis, data mining