COALTECH

MAYA On-Line Analyzer Evaluation

G J de Korte

August 2007

1
Copyright COALTECH
This document is for the use of COALTECH only, and may not be transmitted to any other party, in whole or in part, in any form
without the written permission of COALTECH.
 Executive Summary

A new type of elemental on-line analyzer was developed by Laser Detect Systems (LDS)
of Israel. The new analyzer, called ‘MAYA’, uses Laser Induced Breakdown
Spectroscopy (LIBS) to determine the quality of coal on a moving conveyor belt. LDS
and Coaltech agreed to cooperate in testing the unit at Optimum Colliery. The results of
the test were favorable.

2
TABLE OF CONTENTS

1. Introduction............................................................................................................. 4
2. LIBS........................................................................................................................ 5
3. Maya....................................................................................................................... 6
4. Calibration .............................................................................................................. 6
5. Test on 15 August................................................................................................... 8
6. Conclusion............................................................................................................ 10
7. Comments ............................................................................................................ 10
8. References ........................................................................................................... 11

3
 1. Introduction
Laser Detect Systems (LDS), based in Israel, developed a new on-line elemental
analyzer called ’Maya’. The analyzer differs from other elemental on-line analyzers
presently in use in that it utilizes Laser Induced Breakdown Spectroscopy (LIBS) to
measure the quality of coal on a conveyor belt instead of Prompt Gamma Neutron
Activation Analysis (PGNAA) used by the other analyzers. The analyzer is much smaller
and lighter than the on-line analyzers currently in use and, most importantly, does not
use radioactive sources.

LDS and Coaltech agreed to co-operate to test the analyzer in South Africa in order to
evaluate its potential. The local agent for the Maya, World Focus, handled the
installation and operation of the unit and Coaltech assisted in the evaluation of the
performance of the unit.

It was decided to install the analyzer on the Wemco product conveyor at Optimum
Colliery. A Coalscan 9500X PGNAA analyzer is in operation on the same conveyor. The
conveyor is also equipped with an automatic sampler and two-hourly composite samples
are procured and analyzed by the Optimum laboratory routinely. The data from the
Coalscan and the laboratory analyses were used to provide a comparison with the
readings obtained from the Maya.

The Maya analyzer was installed at Optimum on 22 March 2007. It is worth commenting
that the installation was carried out in less than a day. A communication link was
provided which allowed the data to be downloaded in Israel and also to facilitate
adjustments to the calibration to be made from Israel. The unit was commissioned on 26
March and has been in operation since. Figure 1 shows the Maya analyzer at Optimum.

Figure 1: Maya on-line analyzer

4
The analyzer has now been in operation for 4 months and the calibration is now fully
established. It was therefore decided to evaluate the performance of the Maya as of 1
August 2007. The data from the Coalscan and the Optimum laboratory were compared
to the data from the Maya. A special test, during which time sample increments were
taken at 5 minute intervals, was also carried out to compare the performance of the
Maya over a shorter time period. The results obtained are given in this report.

2. LIBS
The following explanation of the principle used by LIBS to derive the basic elemental
composition of materials is taken from the website of Applied Photonics Ltd., based in
the United Kingdom. (http://www.appliedphotonics.co.uk/index.htm)

Laser-Induced Breakdown Spectroscopy (LIBS) is a form of atomic emission

spectroscopy in which a pulsed laser is used as the excitation source. The basic
principle is illustrated in Figure 2 below. The output of a pulsed laser, such as a Q-
switched Nd:YAG, is focused onto the surface of the material to be analyzed. For the
duration of the laser pulse, which is typically 10 nanoseconds, the power density at the
surface of the material can exceed 1 Gigawatt per cm2 using only a compact laser
device and simple focusing lenses.

Figure 2: LIBS principle

At these very high power densities, a fraction of a microgram of material is ejected from
the surface by a process known as laser ablation and a short lived but highly luminous
plasma with instantaneous temperatures reaching 10,000oC is formed at the surface of
the material. Within this hot plasma, the ejected material is dissociated into excited ionic
and atomic species. At the end of the laser pulse, the plasma quickly cools as it expands
outwards at supersonic speeds. During this time the excited ions and atoms emit
characteristic optical radiation as they revert to lower energy states. Detection and
spectral analysis of this optical radiation using a sensitive spectrograph can be used to
yield information on the elemental composition of the material.

5
 3. Maya
The Maya analyzer uses a laser pulse to produce the luminous plasma on the surface of
the coal particles on the conveyor which is then analyzed to obtain the basic elemental
composition of the coal. Since the height of the coal layer on the conveyor changes
continuously, it is necessary to vary the focal point of the laser beam to ensure that it is
correctly focused at the precise moment that the laser fires. To achieve this, a height
adjustment mechanism, controlled by an ultrasonic sensor, is used to control the
distance between the coal and the lens of the laser. The frequency of the laser pulse can
be set at up to 20 pulses per second – at Optimum 5 pulses per second was found to be
optimal. The spectrum obtained from every single pulse can be analyzed and it is usual
to accumulate the average of a large number, typically more than 500, readings to
improve the statistics. The analyzer can be adjusted to report the analysis every minute
or at longer intervals as required.

4. Calibration
Although some preliminary calibration of the analyzer was carried out in Israel before the
unit was shipped, it was required to re-establish the calibration after installation. The
calibration took a relatively long time to perfect but this is not unusual, especially
considering that this is the first application of the Maya on coal. Initially the Maya data
showed a large degree of scatter. The calibration algorithm was, however, further
developed and adjusted and as from 1 August, very good agreement with both the
laboratory and Coalscan data was obtained. There was, unfortunately, a period of
almost two days when the Maya data were unavailable due to a problem with a
compressor. Figure 2 shows the comparative data between the Maya and the 2-hourly
laboratory results for the period 1 to 15 August. The average Maya results for the 2 week
period was 13,2% ash which compares favorably with 13,4% for the laboratory results
and 13,3% for the Coalscan. Figure 3 shows the Maya / Coalscan comparison.

6
 Optimum: Maya / Lab comparison (1 to 15 Aug )

21

19

17
% Ash content

15

13

11

5
39295 39297 39299 39301 39303 39305 39307 39309

Time value
Maya Lab

Figure 2: Maya and laboratory results compared

Optimum: Maya / Coalscan comparison (1 Aug to 15 Aug)

21

19

17
% Ash content

15

13

11

5
39295 39297 39299 39301 39303 39305 39307 39309 39311

Time value
Coalscan Maya

Figure 3: Maya and Coalscan results compared

Table 1 and Figure 4 shows the comparative normal distribution curves for the Coalscan,
the laboratory and the Maya results for the period 1 to 15 August.

7
Table 1: Comparison of ash values
Coalscan Laboratory Maya
Number of observations 2866 250 1015
Mean 13.35 13.43 13.17
Max 30.00 17.80 16.50
Min 10.14 10.00 11.10
Std Dev 0.672 1.064 0.827

Comparative normal distributions

0.7

0.6

0.5
Relative frequency

0.4

0.3

0.2

0.1

0.0
11 12 13 14 15 16
Ash %

Coalscan Lab Maya

Figure 4: Normal distribution curves

It can be seen from the data presented in Table 1 and also from Figure 4 that the mean
of the Maya readings is slightly lower than that of the Coalscan and the laboratory. The
standard deviation for the Coalscan is the lowest of the three determinations. This is due
to the higher number of values reported by the Coalscan and the fact that the unit
derives its values from the bulk of the material on the conveyor. The laboratory values
exhibit the highest standard deviation and that of the Maya is intermediate to the other
two.

5. Test on 15 August
A special test was conducted on 15 August to compare the performance of the unit over
a shorter time interval. The test comprised taking 30 samples from the conveyor, using
the automatic sampler, at 5 minute intervals. To do this, the complete primary sample
was re-directed to the by-pass chute in the sampling circuit where it could be collected in
a plastic bag. The sample was split, using a riffle chute, into two equal portions. One
portion was prepared and analyzed at the Optimum laboratory and the second portion by
Yanka laboratory. The results obtained from the laboratory analyses and the

8
corresponding data from the two on-line analyzers are summarized in Table 2. The
detailed results are appended in Appendix A.

Table 2: Summary of test results

Yanka Optimum Average of
laboratory laboratory laboratories Coalscan Maya
n 30 30 30 31 35
Mean 13.6 14.0 13.8 13.8 13.7
Std Dev 0.87 0.66 0.54 0.57 0.57

A graphical comparison of the data for the two laboratories and the two analyzers is
shown in Figure 5.

Maya test 15 August 2007 - all data

16.0

15.0

14.0
% Ash

13.0

12.0

11.0

10.0
0.47 0.49 0.51 0.53 0.55 0.57 0.59
Time (Hours)

Coalscan Optimum Lab Yanka Lab Maya

Figure 5: Test data

The Maya results are shown compared to the average values of the laboratory
determinations in Figure 6. It can be seen from Figure 6 that there is a relatively good
agreement between the data sets.

9
 Maya compared to average results of laboratories

16.0

15.0

14.0
% Ash

13.0

12.0

11.0

10.0
0.47 0.49 0.51 0.53 0.55 0.57 0.59
Time (Hours)

Laboratories Maya

Figure 6: Maya data compared to mean laboratory values

The data presented in Table 2 show that the sample means and standard deviations
derived from the laboratory results and the Maya determinations are quite close.
Student’s t-test was employed to determine whether the means of the distributions are
statistically similar. The procedure for the test is given in Appendix B. The outcome of
the test was positive and confirms that there is a 95% probability that the two sample
distributions were derived from the same population.

6. Conclusion
The Maya analyzer has been proven to be able to accurately determine the quality of
coal on a moving conveyor belt in real time. The performance of the unit was evaluated
and the results were found to be statistically valid.

7. Comments
A disadvantage of the Maya is the fact that it measures only the surface of the coal
particles. On a conveyor transporting a single product, for example Wemco floats as at
Optimum, this is not a problem but in cases where cyclone product is placed on the
conveyor first and the Wemco product loaded on top, only the Wemco product will be
monitored. One possible means of getting around this problem would be to install the
Maya after a transfer point so that the mixed coal is presented to the analyzer.

The fact that the Maya analyzes the coal surface, on the other hand, provides a potential
advantage. PGNAA (and dual-gamma monitors) require a minimum thickness of coal,
typically 100 mm, on the conveyor to be able to effectively analyze the bulk of the

10
material. In many coal preparation plants, especially when low product yields are
experienced, the layer of coal on the product conveyor is sometimes only one or two
particles thick. Under these conditions, existing on-line analyzers can not provide
accurate analyses but the Maya could well. This will, however, require testing in practice
to prove.

The Optimum test only evaluated the performance of the Maya when determining the
ash content of coal. The unit, being an elemental analyzer, is capable of measuring
other parameters of coal as well – calorific value for example. Further work will have to
be done by LDS to evaluate this potential.

The Maya calibration for Optimum coal has been established and proven. It is
anticipated that the calibration will be easy to adjust when the unit is employed to
measure the quality of other South African coals.

8. Appreciation
The author wishes to express a special word of thanks to Claudia Lopes and the staff of
the Optimum laboratory as well as the staff of Yanka Laboratories for their assistance in
the evaluation of the Maya analyzer.

9. References
1. Applied Photonics Ltd. Website. http://www.appliedphotonics.co.uk/index.htm
(accessed on 20 August 2007)

2. Solid mineral fuels – evaluation of the measurement performance of on-line

analysers. International Standard ISO 15239

3. Cass, T. Statistical methods in Management. Cassell Ltd. London.

11
APPENDIX A

OPTIMUM COLLIERY
LABORATORY REPORT

For: Drum LAC 5 Minute Samples

Date: 15/08/2007

Air Dry Yanka Optimum Coalscan Yanka Opt Coalscan

Coalscan ID and Moisture
Lab ID and Time Sampled Ash
Time Analysed in Ash Content (%)
Sample Weight Content
Sample (Air Dry)
(%) Dry)
(%)
DRUM_PROD_NO1_11:30 15/08/2007 11:31 4.1 14.35 17.05 37.55 14.7 13.1 13.5
DRUM_PROD_NO2_11:35 15/08/2007 11:36 4.2 13.05 16.5 33.1 12.6 14.9 13.5
DRUM_PROD_NO3_11:40 15/08/2007 11:42 3.9 14.45 17.7 35.07 12.8 14.4 13.8
DRUM_PROD_NO4_11:47 15/08/2007 11:47 4.1 8.85 8.7 23.88 14.5 12.9 13.9
DRUM_PROD_NO5_11:57 15/08/2007 12:00 3.8 7.35 6.55 12.78 14.9 13.8 15.0
DRUM_PROD_NO6_12:02 15/08/2007 12:05 3.9 10.85 14.5 30.18 13.8 13.8 13.7
DRUM_PROD_NO7_12:07 15/08/2007 12:11 3.6 12.95 10.7 26.75 13.6 14.2 13.4
DRUM_PROD_NO8_12:13 15/08/2007 12:16 3.9 15.25 11.3 26.97 13.9 13.0 13.7
DRUM_PROD_NO9_12:18 15/08/2007 12:22 3.8 4.3 3.95 12.76 14.3 13.3 13.4

12
DRUM_PROD_NO10_12:24 15/08/2007 12:27 4.1 10.15 9.95 24.35 11.5 13.7 13.9
DRUM_PROD_NO11_12:28 15/08/2007 12:32 4.0 14.55 13.7 31.69 12.7 13.9 13.3
DRUM_PROD_NO12_12:34 15/08/2007 12:37 3.9 15.65 13 29.64 14.2 14.9 13.5
DRUM_PROD_NO13_12:39 15/08/2007 12:43 3.7 14.95 17.05 33.51 12.9 13.8 13.6
DRUM_PROD_NO14_12:45 15/08/2007 12:48 3.7 14.3 16.95 35.59 13.2 13.0 13.5
DRUM_PROD_NO15_12:50 15/08/2007 12:54 4.1 13.9 16.7 34.77 13.1 14.0 13.4
DRUM_PROD_NO16_12:55 15/08/2007 12:59 4.0 12.45 14.55 27.77 13.0 14.9 13.7
DRUM_PROD_NO17_13:00 15/08/2007 13:00 3.9 12.2 16.1 5.89 13.4 13.7 14.1
DRUM_PROD_NO18_13:06 15/08/2007 13:05 3.6 14.05 18.95 34.62 12.5 13.9 13.9
DRUM_PROD_NO19_13:11 15/08/2007 13:11 3.6 12.55 12.1 35.47 13.6 14.6 13.1
DRUM_PROD_NO20_13:22 15/08/2007 13:16 3.9 12.05 15.2 3.69 14.5 14.9 11.7
DRUM_PROD_NO21_13:27 15/08/2007 13:25 3.7 13.7 12.35 32.07 13.4 13.9 14.8
DRUM_PROD_NO22_13:32 15/08/2007 13:30 3.8 15.95 19.1 30.17 13.9 13.3 13.6
DRUM_PROD_NO23_13:38 15/08/2007 13:35 4.1 13.6 14.85 33.16 13.6 15.0 13.9
DRUM_PROD_NO24_13:43 15/08/2007 13:40 4.0 15.75 16.05 33.33 14.7 14.4 14.0
DRUM_PROD_NO25_13:48 15/08/2007 13:44 3.9 15.7 15.8 30.78 13.0 15.0 13.9
DRUM_PROD_NO26_13:54 15/08/2007 13:50 3.7 16 17.75 38.25 15.1 14.6 13.6
DRUM_PROD_NO27_13:59 15/08/2007 13:54 3.7 15.4 23.9 32.39 14.9 13.8 14.1
DRUM_PROD_NO28_14:04 15/08/2007 14:00 4.0 17.55 14.15 40.32 13.5 14.0 14.3
DRUM_PROD_NO29_14:09 15/08/2007 14:05 3.8 13.1 16.05 32.47 12.7 13.1 14.5
DRUM_PROD_NO30_14:15 15/08/2007 14:10 3.9 14.95 15.4 34.16 14.3 14.5 14.4
15/08/2007 14:15 35.39 14.2

13
Maya data 15 August 2007
Time Ash %
11:25:41 13.43
11:30:08 13.68
11:34:17 14.19
11:38:38 13.02
11:43:13 14.26
11:56:44 15.01
12:01:13 13.68
12:05:46 13.37
12:10:14 14.15
12:14:31 13.51
12:20:49 13.51
12:25:34 13.89
12:29:49 13.23
12:34:05 14.82
12:38:07 13.95
12:42:42 13.19
12:46:51 13.44
12:51:20 12.85
12:55:21 14.22
13:00:06 14.13
13:04:16 13.43
13:08:22 13.23
13:20:06 13.90
13:24:35 13.25
13:29:25 13.54
13:33:53 13.00
13:38:19 14.28
13:42:33 14.09
13:47:05 13.39

14
13:51:51 12.68
13:55:58 13.86
14:00:30 13.77
14:04:48 14.90
14:09:11 14.29
14:13:56 13.24

15
APPENDIX B

Student’s t-test for unequal sample sizes

1. Null hypothesis
The 2 sample distributions have the same population means

2. Alternative hypothesis
The distribution means differ significantly

3. Significance level
0.05

4. Data
Name Maya Labs
Number of observations (n) 35 30
Mean 13.73 13.82
Standard deviation 0.57 0.54

5. Determine t value

t = -0.67695

6. Degrees of freedom
Degrees of freedom = (n1 + n2 -2) = 63

16
7. Obtain theoretical limiting t value from tables
tth = 2.0

8. Evaluate values
If MOD(t) > tth then reject null hypothesis otherwise accept null hypothesis

9. Outcome of test
t < tth therefore accept null hypothesis -> There is a 95% probability that the means of the distributions are the same

17