6
achieving a Root Mean Square Error (RMSE) of 4.46%.
The classification model, on the other hand, demonstrated
impressive accuracy, reaching 76.92% for Category 2
(TIC =70%–80%). Nevertheless, the overall accuracy of
the classification model stands at 56%, which is primar-
ily attributed to challenges encountered by the classifier in
accurately categorizing instances within Category 1 (TIC
70%).
CONCLUSIONS
Two models, regression, and classification, were evaluated
for the data acquired using Stellarnet NIR ADK portable
spectrometer. The regression model achieved a RMSE of
4.46% and accuracy of R2 equal to 0.60. The classifica-
tion model demonstrated a high accuracy of 76.92% for
Category 2 (TIC =70%–80%), however, its overall accu-
racy of 56% requires further analysis with the main dif-
ficulty encountered during the classification within the
grouping of Category 1 (TIC 70%).
FUTURE DEVELOPMENTS
Further refinement and optimization of the models for
Stellarnet NIR ADK are in progress including:
1. Refinement of spectra correction techniques,
2. Optimization of smoothing methodologies, and
3. Identification of machine learning algorithms best
suited to the spectral data.
ACKNOWLEDGMENT
The authors would like to acknowledge the Coal Services
Trust for funding the research through the Project Number
20633 entitled “System Demonstrator of a Portable NIR
spectrometer for Rapid Stone Dust Compliance Testing”
and the mining companies for providing their coal samples
used in this study.
REFERENCES
[1] Harris ML, Sapko MJ, Varley FD, Weiss ES (2012),
“Coal Dust Explosibility Meter Evaluation and
Recommendations for Application,” National Institute
for Occupational Safety and Health, Pittsburgh,
Pennsylvania, USA.
[2] Wu HW, Gillies S (2010), “Evaluation Of A New
Instantaneous Coal Dust Explosibility Meter For
Use In Mine Airways,” The Australian Coal Industry’s
Research Program, Brisbane, QLD, Australia.
[3] Wedel DJ, Belle B, Kizil MS (2015), “The
Effectiveness of Rapid Stone Dust Compliance
Testing in Underground Coal,” Coal Operators’
Conference, Wollongong, NSW, Australia.
[4] Recongnised Standard 5 (2023) [Online]. Available:
https://www.rshq.qld.gov.au/__data/assets/pdf
_file/0006/240378/recognised-standard-05.pdf.
[5] Ching R (2017), “Measuring Ash Content in Coal
using Portable X-Ray Fluorescence,” Brandon
University, Brandon, Canada.
[6] Kasbekar RS, Ji S, Clancy EA, Goel A (2023),
“Optimizing the Input Feature Sets and Machine
Learning Algorithms for Reliable and Accurate
Estimation of Continuous, Cuffless Blood Pressure,”
Nature Portfolio, pp. 1–13.
[7] Xie LH, Tang SQ, Wang XQ, Sheng ZH, Hu SK,
Wei XJ, Jiao GA, Shao GN, Wang L, Hu PS (2023),
“Simultaneously Determining Amino Acid Contents
Using Near-Infrared Reflectance Spectroscopy
Improved by Pre-processing Method in Rice,” Food
Science and Technology, pp. 1–20.
[8] Schmid M, Rath D, Diebold U (2022), “Why and
How Savitzky–Golay Filters Should Be Replaced,”
ACS Publications, pp. 185–196.
[9] Gong M, Wang K, Sun H, Wang K, Zhou Y, Cong
Y, Deng X, Mao Y (2023), “Threshold of 25(OH)
D and Consequently Adjusted Parathyroid Hormone
Reference Intervals: Data Mining for Relationship
between Vitamin D and Parathyroid Hormone,”
Endocrinological Investigation, pp. 2067–2077.
Figure 6. Confusion Matrix for NIR Classification Model
(Test Set)
achieving a Root Mean Square Error (RMSE) of 4.46%.
The classification model, on the other hand, demonstrated
impressive accuracy, reaching 76.92% for Category 2
(TIC =70%–80%). Nevertheless, the overall accuracy of
the classification model stands at 56%, which is primar-
ily attributed to challenges encountered by the classifier in
accurately categorizing instances within Category 1 (TIC
70%).
CONCLUSIONS
Two models, regression, and classification, were evaluated
for the data acquired using Stellarnet NIR ADK portable
spectrometer. The regression model achieved a RMSE of
4.46% and accuracy of R2 equal to 0.60. The classifica-
tion model demonstrated a high accuracy of 76.92% for
Category 2 (TIC =70%–80%), however, its overall accu-
racy of 56% requires further analysis with the main dif-
ficulty encountered during the classification within the
grouping of Category 1 (TIC 70%).
FUTURE DEVELOPMENTS
Further refinement and optimization of the models for
Stellarnet NIR ADK are in progress including:
1. Refinement of spectra correction techniques,
2. Optimization of smoothing methodologies, and
3. Identification of machine learning algorithms best
suited to the spectral data.
ACKNOWLEDGMENT
The authors would like to acknowledge the Coal Services
Trust for funding the research through the Project Number
20633 entitled “System Demonstrator of a Portable NIR
spectrometer for Rapid Stone Dust Compliance Testing”
and the mining companies for providing their coal samples
used in this study.
REFERENCES
[1] Harris ML, Sapko MJ, Varley FD, Weiss ES (2012),
“Coal Dust Explosibility Meter Evaluation and
Recommendations for Application,” National Institute
for Occupational Safety and Health, Pittsburgh,
Pennsylvania, USA.
[2] Wu HW, Gillies S (2010), “Evaluation Of A New
Instantaneous Coal Dust Explosibility Meter For
Use In Mine Airways,” The Australian Coal Industry’s
Research Program, Brisbane, QLD, Australia.
[3] Wedel DJ, Belle B, Kizil MS (2015), “The
Effectiveness of Rapid Stone Dust Compliance
Testing in Underground Coal,” Coal Operators’
Conference, Wollongong, NSW, Australia.
[4] Recongnised Standard 5 (2023) [Online]. Available:
https://www.rshq.qld.gov.au/__data/assets/pdf
_file/0006/240378/recognised-standard-05.pdf.
[5] Ching R (2017), “Measuring Ash Content in Coal
using Portable X-Ray Fluorescence,” Brandon
University, Brandon, Canada.
[6] Kasbekar RS, Ji S, Clancy EA, Goel A (2023),
“Optimizing the Input Feature Sets and Machine
Learning Algorithms for Reliable and Accurate
Estimation of Continuous, Cuffless Blood Pressure,”
Nature Portfolio, pp. 1–13.
[7] Xie LH, Tang SQ, Wang XQ, Sheng ZH, Hu SK,
Wei XJ, Jiao GA, Shao GN, Wang L, Hu PS (2023),
“Simultaneously Determining Amino Acid Contents
Using Near-Infrared Reflectance Spectroscopy
Improved by Pre-processing Method in Rice,” Food
Science and Technology, pp. 1–20.
[8] Schmid M, Rath D, Diebold U (2022), “Why and
How Savitzky–Golay Filters Should Be Replaced,”
ACS Publications, pp. 185–196.
[9] Gong M, Wang K, Sun H, Wang K, Zhou Y, Cong
Y, Deng X, Mao Y (2023), “Threshold of 25(OH)
D and Consequently Adjusted Parathyroid Hormone
Reference Intervals: Data Mining for Relationship
between Vitamin D and Parathyroid Hormone,”
Endocrinological Investigation, pp. 2067–2077.
Figure 6. Confusion Matrix for NIR Classification Model
(Test Set)