AE 469/569 TERM PROJECT DEVELOPING CORRECTION FUNCTION FOR TEMPERATURE EFFECTS ON NIR SOYBEAN ANALYSIS

1
AE 469/569 TERM PROJECTDEVELOPING CORRECTION
FUNCTION FOR TEMPERATURE EFFECTS ON NIR SOYBEAN
ANALYSIS

• Jacob Bolson
• Maureen Suryaatmadja
• Agricultural Engineering
• Iowa State University
• May 4, 2005

2
Introduction

• NIR instruments play an important role in
  predicting chemical composition and biological
  properties of food and agricultural material.
• NIR spectroscopy measures the wavelength and
  intensity of the absorption of near infrared
  light (800nm 2.5µm) by a given sample.

3
Introduction

• NIR is primarily used for the detection of C-H,
  N-H and O-H bonds, which relate to concentration
  of oil, protein and moisture.
• The advantages of NIR
• a non-destructive procedure
• minimal sample preparation
• fast analytical techniques (less than 1 minute)
• Two important factors in NIR analysis
• a spectrum
• reference values

4
Introduction

• The development of calibration model on NIR
  instrument consists of two procedures
• develop a base calibration
• add the samples to the base calibration for
  instrument and temperature stabilization
• Temperature stabilization, collect at grain
  temperatures from -150C to 450C.
• (Rippke et all., 1996)

5
Problem Statement

• The method to include some hot and cold samples
  does not work well and quite inconsistent.
• NIR spectra of liquid component shift on
  wavelength axis as temperature changes, predicted
  results become less accurate.

6
Problem Statement

• Researchers have proven that NIR spectra of
  liquid components shift on the wavelength axis as
  temperature changes
• The bands corresponding to hydrogen-bonding
  groups (N-H, O-H bands) are expected to be highly
  influenced by temperature (Miller, 2001)
• Temperature influences the spectra, the increase
  of temperature allows liberating a part of fixed
  water meat measurement (Corbisier et all.,
  2004)

7
Objective

• To determine whether a temperature adjustment
  function could improve the accuracy of NIR
  analysis at conditions other than room
  temperature.

8
Materials and Methods

• Soybean sample temperature set from ISU Grain
  Quality Lab (20 samples)
• Run in three conditions cold, room, and hot
  using Omega G 6110 Analyzer with temperature
  compensation calibration (already exists).

9
Materials and Methods

• Recalculate the results using no temperature
  compensation calibration.
• Calculate the slope from every prediction values
  of moisture, protein and oil using Excel
  function.
• Calculate the average and standard deviation of
  the slopes.

10
Materials and Methods

• One of the samples was discarded because of its
  extreme values.
• Test the slopes on the original samples using
  this formula
• Corrected value Measured value (m (250C
  measured temperature))
• Calculate the differences between the corrected
  values at non-room and room temperature.
• Test the slopes (m, msd, m-sd) on the new seven
  soybeans samples using the same previous
  procedure.

11
Result
12
Result
13
Results
  Moisture Protein Oil
Average (m) (/0C) 0.0164 -0.0048 0.0063
Std Dev (/0C) 0.0083 0.0128 0.0042
msd (/0C) 0.0247 0.0079 0.0105
m-sd (/0C) 0.0081 -0.0176 0.0021
Range (/0C) 0.0165 0.0255 0.0084
14
Correction Function

• M corrected M measured (0.0164 (250C- T
  measured))
• P corrected P measured (- 0.0048 (250C - T
  measured))
• O corrected O measured (0.0063 (250C - T
  measured))
• M Moisture
• P Protein
• O Oil

15
Results (19 Samples)
  Moisture Differences Non room and Room Temperature Moisture Differences Non room and Room Temperature Moisture Differences Non room and Room Temperature
  Without Temperature Using Correction With Temperature
  Compensation Function Compensation
Average () 0.384 0.208 0.338
Std Dev () 0.210 0.137 0.177
16
Results
17
Results (19 Samples)
  Protein Differences Non room and Room Temperature Protein Differences Non room and Room Temperature Protein Differences Non room and Room Temperature
  Without Temperature Using Correction With Temperature
  Compensation Function Compensation
Average () 0.479 0.474 0.457
Std Dev () 0.309 0.312 0.309
18
Results
19
Results (19 Samples)
  Oil Differences Non room and Room Temperature Oil Differences Non room and Room Temperature Oil Differences Non room and Room Temperature
  Without Temperature Using Correction With Temperature
  Compensation Function Compensation
Average () 0.200 0.154 0.154
Std Dev () 0.145 0.113 0.115
20
Results
21
Moisture Correction Function(7 samples)

• M corrected M measured (0.0164 (250C- T
  measured))
• M corrected M measured (0.0247
• (250C - T measured))
• M corrected M measured (0.0081 (250C- T
  measured))

22
Results (7 samples)
  Moisture Differences Non room and Room Temperature Moisture Differences Non room and Room Temperature Moisture Differences Non room and Room Temperature
  Without Temperature Using Correction With Temperature
  Compensation Function Compensation
Average () 0.297 0.128 0.132
Std Dev () 0.153 0.089 0.103
23
Results
24
Results
25
Results
26
Protein Correction Function(7 Samples)

• P corrected P measured (- 0.0048 (250C - T
  measured))
• P corrected P measured (0.0079 (250C - T
  measured))
• P corrected P measured (- 0.0176 (250C - T
  measured))

27
Results ( 7 samples)
  Protein Differences Non room and Room Temperature Protein Differences Non room and Room Temperature Protein Differences Non room and Room Temperature
  Without Temperature Using Correction With Temperature
  Compensation Function Compensation
Average () 0.450 0.388 0.261
Std Dev () 0.343 0.321 0.187
28
Results
29
Results
30
Results
31
Oil Correction Function

• O corrected O measured (0.0063 (250C - T
  measured))
• O corrected O measured (0.0105 (250C - T
  measured))
• O corrected O measured (0.0021 (250C - T
  measured))

32
Results (7 samples)
  Oil Differences Non room and Room Temperature Oil Differences Non room and Room Temperature Oil Differences Non room and Room Temperature
  Without Temperature Using Correction With Temperature
  Compensation Function Compensation
Average () 0.136 0.105 0.118
Std Dev () 0.090 0.072 0.079
33
Results
34
Results
35
Results (7 samples)
36
Conclusion

• A temperature adjustment function
• M corrected M measured (0.0164
  (250C- T measured))
• P corrected P measured (- 0.0048
  (250C- T measured))
• O corrected O measured (0.0063
  (250C- T measured))
• M Moisture
• P Protein
• O Oil
• can be used to improve the accuracy of NIR
  predicted values at conditions other than room
  temperature.

37
Conclusion

• The correction function applied to soybean
  moisture and oil was more consistent than to
  protein.
• The implementation of a correction function is
  less time consuming than developing temperature
  compensation calibration because a slope
  correction can be recalculated for new
  calibrations.
• The future work should implement the correction
  function into the soybean calibration development
  and test with other NIR instruments.