dc.description.abstract | Woody biomass is currently sold on a wet basis, a system that works because moisture
content of harvested trees normally
uctuates within a narrow range and wood from multiple
sources can be considered a uniform product. Wet wood, however, is not an optimal energy
feedstock and various methods have been proposed to reduce moisture content in woody
biomass prior to its sale. Once moisture content can vary outside its typical range, however,
the use of wet weight as a basis for selling timber products becomes untenable. Promoting
a robust bioenergy market will, therefore, require the development of reliable sensors for
measuring moisture content of wood in the eld. A logical place at which to measure moisture
content of woody biomass would be just prior to shipment for nal conversion. For biomass
harvested from pine plantations prevalent in the US South, that would most likely be between
the chipping and loading operations. To avoid handling the biomass an extra time, however,
a moisture sensor should work accurately on a moving stream of chips as they exit a chipper
and are blown into a van for hauling. This type of high-speed, non-contact moisture sensor
suitable for application in a rough environment is not currently available. Our goal in this
study was to develop a sensor for measuring moisture content of a rapidly moving stream of
pine chips that was robust, non-contact in operation, reliable, and accurate.
An 8-electrode electrical capacitance sensor was built for that purpose. The multiple
electrode system was used to characterize the distribution of permittivity within the circular
area (diameter = 16.5cm) enclosed by the sensor and surrounded by the electrodes (2.5x10
cm copper plates). The sensor was could be operated in two modes: a) static mode in which
the cumulative permittivity between each combination of plates was the variable correlated
with moisture (a bulk measurement), and b) a dynamic mode in which multiple sequential
readings of electrode combinations was used to sample a stream of chips moving through the
ii
sensor enclosure, the result being a sequence of 2-D maps of permittivity. This method is
referred to as electrical capacitance tomography (ECT).
The objectives of the research presented in this dissertation were:
1. Validate the accuracy of the sensor in static mode in predicting moisture content of
xed quantities of biomass, and investigate the sensor's limits of performance when
subjected to variations in such confounding variables as quantity, location, and size of
the material under test.
2. Develop the necessary hardware and software systems to perform dynamic (ECT) mode
characterization of moving streams of biomass, and establish the accuracy of moisture
content predictions.
Static mode calibration of the sensor was carried out on wood chips having uniform
moisture content. Tests were made on chips ranging in moisture content of from 4% to 140%
(dry basis), and the results compared with the widely-accepted near infrared spectroscopy
(NIR) method. For tests on individual wood chips (about 10 g each), the root mean squared
errors of prediction (RMSEP) were 13.52% and 48.43% for NIR and ECT, respectively. In
bulk measurements (multiple chips, 40-80 in number), RMSEPs were found to be 15.39%
and 11.65% for the NIR and ECT methods, indicating the main di erence between the two
methods: NIR predictions were independent of sample mass, the ECT estimates were not.
Two additional measurement procedures were developed using the capacitance sensor, one
using multiple excitation frequencies and the other assuming a xed sample mass based on
volume, to remove its dependency on sample mass. The RMSEPs for the two methods were
similar (13.0% and 10.9% for the xed volume and two-frequency methods, respectively) and
also comparable to the previous methods when sample mass was known. It was concluded the
capacitive approach was very comparable to NIR in its ability to accurately predict moisture
content of biomass using either approach (unknown or known mass), and the capacitive
sensor was superior in predicting moisture content of bulk materials. On quantities of wood
iii
near the detection limit of the capacitance sensor (about 10g), however, the NIR approach
was preferable.
Dynamic mode tests of the sensor were made in two steps, the rst involving simultane-
ous independent measurement of mass
ow using another sensor (plus moisture content from
the sensor in capacitive mode), and the second using the sensor in ECT mode to estimate
both total mass
ow and moisture content simultaneously. In the rst tests, an impact mea-
surement approach was used to estimate mass
ow. In the second, the ECT tomographic
mode was used to image the variation in permittivity of the
owing stream of wood chips.
The images also provided a means to estimate chip volume within the sensor, which, assum-
ing uniform density, provided a mass value. The RMSEP of moisture content predictions
using the impact and ECT mass
ow estimation methods were 11.86% and 17.71%, respec-
tively. It was concluded these tests proved the feasibility of using electrical capacitance
tomography as a means of measuring moisture content of moving biomass streams. Further
development of the techniques to increase the sampling rates above that achieved in these
tests will be required to make the system practical. | en_US |