Developing an electronic portable device based on dielectric power spectroscopy for non-destructive prediction of date moisture content

Moisture content is one of the main quality parameters for evaluating and grading of dates. This study was intended to design and develop an electronic portable device based on a dielectric power spectroscopy for nondestructive and accurate prediction of the moisture content of dates in a noncontact parallel-plate mode. The device consists of the main, monitor, and feeding boards along with two elliptical parallel electrodes. The preliminary tests were first carried out with spectrum analyzer and function generator to acquire the dielectric power spectra of the dates in the radio frequency range of 1-100 MHz through noncontact parallel plates. The moisture content predictive models were then developed by adopting different data analysis procedures including multiple linear, principle components, and partial least squares regressions. Following the satisfactory results obtained from the proposed models, as many as six effective frequencies (SEFs) which displayed the maximum correlations with the moisture content were selected and defined for the microcontroller of the developed electronic portable device. Afterward, the output voltage of the device were collected from the new sample set and used to predict the moisture content by means of artificial neural network (ANN) analysis. The obtained ANN with a topology of 6-5-1 enabled excellent prediction of moisture content with the coefficient of determination in prediction (R-p(2)) and standard deviation ratio (SDR) values of 0.934 and 3.62, respectively. The results showed strong potential of the developed device for accurate prediction of the moisture content of dates through the noncontact parallel-plate mode. (C) 2016 Elsevier B.V. All rights reserved.