Submitted to: Transactions of the ASAE
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: February 23, 1998
Publication Date: N/A
Interpretive Summary: One of the most important factors in the storage and sale of cereal grains is moisture content. A rapid dielectric moisture sensor for flowing grain would have application in precision farming, grain elevator management, and numerous processing systems. An on-line moisture sensor's accuracy is affected by bulk density fluctuations of the grain as it passes through the esensor. This research describes a technique to sense the moisture content of wheat that is practically independent of grain bulk density variations. Grain moisture content is sensed by taking complex transmission measure- ments at frequencies from 1 to 350 MHZ of static wheat samples in a coaxial sample holder designed for flowing grains. Advanced multivariate analysis tools were used to find data outliers and to determine a few optimum measurement frequencies for potential use in a practical instrument. The results were calibrated against air-oven moisture content and validated with a standard error of 0.36% moisture content for four wheat cultivars ranging in moisture content from 9% to 21%, wet basis. This research demonstrated the potential for an inexpensive on-line dielectric moisture meter.
Technical Abstract: The moisture content of hard red winter wheat from 9% to 21%, with densities ranging from 0.66 to 0.83 g/cm**3, was predicted by an equation using the complex transmission coefficient S**21 at three frequencies, about 40, 240, and 320 MHZ. Measurements at these frequencies were selected from multiple linear regression analysis of data collected from 1 to 350 MHZ on static samples in a coaxial flow-through sample holder. Calibration was performed with 49 samples from four hard red winter wheat cultivars harvested in Nebraska in 1994, and validation was performed with 84 different samples of the same cultivars. The standard deviation of the differences (SEP) for the validation data, compared to standard air-oven moisture contents, was 0.36% moisture content, and the bias was - 0.06% moisture content. The frequencies deemed significant in this research, 40, 240, and 320 MHZ, were higher than those selected earlier, 2, 25, and 80 MHZ, in studies with a parallel-plate sample holder involving admittance measurements from 1 to 100 MHZ. The poorer sensitivity of the measuring system at the lowest frequencies was the most likely reason that no calibration frequencies were selected below 40 MHZ. Initial attempts to develop a moisture calibration from the intrinsic dielectric properties were unsuccessful. This was attributed to larger errors in the model used to calculate the loss factor values at the lower frequencies.