Estimation of Moisture Content in Paper Pulp Containing Calcium Carbonate Using Fringing Field Imped

1
Estimation of Moisture Content in Paper Pulp
Containing Calcium Carbonate Using Fringing Field
Impedance Spectroscopy

• Kishore Sundara-Rajan, Leslie Byrd II, and Prof.
  Alexander Mamishev
• Sensors, Energy, and Automation Laboratory
• Department of Electrical Engineering
• University of Washington
• Seattle, WA, USA
• kishore_at_ee.washington.edu
• http//www.ee.washington.edu/research/seal

2
Outline

• Introduction
• Experimental Results
• Data Analysis
• Validation Tests
• Conclusion

3
Motivation

• Annual worldwide paper production is nearly 312
  million tons ? Huge application market.
• Machine controlled using feedback systems ?
  Stable, but slow.

Wet End
Dewatering Section
Finishing Section
Existing Sensors

• 10 sec delay on a 2000 m/min machine leads to
  over 0.2 miles of bad quality paper !!
• Solution Incorporate Feed Forward Control

4
Fringing Field Interdigital Sensor

• For a semi-infinite homogeneous medium placed on
  the surface of the sensor, the periodic variation
  of the electric potential along the X-axis
  creates an exponentially decaying electric field
  along the Z-axis, which penetrates the medium.

5
Experimental Setup

• Pulp is blended using a blender to a consistency
  of a suspension.
• Sensor is attached to the outer side of the base
  of an acrylic tray.
• A guard plane is placed underneath the sensor
  electrodes to provide shielding from external
  electric fields.

6
Experimental Setup

• Sensor Used
• Spatial Wavelength 40 mm
• Finger Length 160 mm
• Penetration Depth 8 mm
• Wall thickness of the tray 5 mm
• RCL Meter (Fluke Manufactured, Model PM6304)
• Single Channel Measurements
• One Volt RMS Sinusoidal AC Voltage
• 50 Hz to 100 kHz Frequency Range

7
Experimental Results
8
Data Analysis

• 3 unknown variables, of which 2 are independent.

• X, Y, and Z are measured electrical parameters.
• m11 to m33, and C1 C3 are constants.
• p, t, and w respectively are the estimated fiber,
  additive and moisture concentrations.

9
Parameter Selection Algorithm

• Automatic selection of parameters and constants
  based on training data set.
• The accuracy of the estimation is dependent on
  the quality of the training data set.
• Two interlinked algorithms operating in parallel
• Learning Algorithm
• Estimation Algorithm

10
Learning Algorithm
11
Learning Algorithm
12
Learning Algorithm
13
Learning Algorithm
14
Estimation Algorithm
Start
Loads the information on best fitting Model,
Parameters and Loadings as determined by learning
algorithm.
Load the Best Fit Information from File.
1
Make Measurements Using IFEF Sensor.
Real-time online measurements.
Estimate Physical Parameters of the Pulp.
A
15
Estimation Algorithm
16
Estimated Values

• Mean of residuals 0.032166667
• Standard deviation of residuals 0.007359801

17
Estimated Values

• Mean of residuals 0.235766667
• Standard deviation of residuals 0.124537764

18
Estimated Values

• Mean of residuals 0.238071667
• Standard deviation of residuals 0.124058865

19
Validation Tests

• Measurement Validation
• Repeatability Test
• Ability to repeat the measurements for the same
  sample
• Reproducibility Test
• Ability to reproduce the measurement for similar
  samples
• Estimation Validation
• Blind Test
• Ability to estimate for untrained data points

20
Repeatability Test Results

• Pulp composition 90 moisture,7.5 fiber, and
  2.5 CaCO3
• Standard deviation is 4 orders of magnitude
  lesser than the mean.

21
Reproducibility Test Results

• Pulp composition 90 moisture,7.5 fiber, and
  2.5 CaCO3
• Standard deviation is 3 orders of magnitude
  lesser than the mean.

22
Blind Test Results
23
Summary

• Advantages
• Non contact measurement
• Static sensor array
• Very high measurement speeds
• Simultaneous estimation of multiple components
• Accuracy better than state-of-art
• Inexpensive
• Disadvantage
• The accuracy is highly dependent on the training
  data set

24
Acknowledgements

• A special thanks goes out to

• Sponsors
• Center for Process Analytical Chemistry, UW
• National Science Foundation
• Electric Energy Industrial Consortium, UW
• Undergraduate Research Assistants
• Abhinav Mathur
• Nick Semenyuk
• Cheuk Wai-Mak
• Alexei Zyuzin

25
Questions?