Iron Ore Sample Mackay Test Rig & Feild Running Data Measurement Results

The purpose of this test was to determine the suitability of the AllScan PGNAA Elemental analyser for use on iron ore quality measurement applications. The report outlines the sample preparation, presentation thereof to the analyser and concludes with the measurement results compared to an independent accredited ore analysis laboratory. The results are presented for accuracy, repeatability and the analyser response to belt loading.

Sample Preparation.
A set five of iron ore samples were obtained from a mine in Western Australia’s Pilbara region for this measurement test. The samples were supplied in 200L drums (one drum per sample). Each drum contained approximately 150kg of material. The samples were sent to a local laboratory for drying and homogenisation as per RTI specification. Each sample was then weighed into 15kg lots, each of which were bagged in a LDPE (Low Density Polyethylene) tube bag and heat sealed at 1m in length. The final dimensions of each of the bags were approximately 1000x200x50mm. Each of the bags was labelled with its sample and bag number. The samples was bagged in this manner to match the strict criteria of the calibration process.

Mackay Test Rig Configuration.
The Mackay test rig consists of an AllScan elemental analyser mounted to a small section of conveyor belt. This conveyor belt has been constructed to allow for easy presentation of samples to the analyser tunnel. The configuration of the test rig used for this sample test work was as follows.

  • Source – 30μg of Cf252
  • Detector – Single 4x4x16 inch NaI Detector
  • Belt – 1200mm

Sample Test Procedure
Three tests were undertaken on the samples, namely:

  • Static Accuracy tests
  • Static repeatability tests
  • Belt loading Vs Accuracy tests

Static Accuracy Test
Five iron ore samples were loaded onto the belt and presented separately to the analyser. These samples were representative of a wide range of ore grades. Measurements were taken by the analyser on a minute basis and collated over an hour period for each sample. The measured results from the analyser were then compared to the laboratory elemental results and the RMSD was calculated for a range of elements of interest. The calculation equation for the RMSD was:

The one-hour analyser result is plotted against the laboratory result for each element of interest. The X-Y plots for laboratory Vs analyser are presented below for Fe, SiO2, Al2O3, TIO2, P and MnO.

The RMSD values for the elements of interest were calculated for the five samples that were presented to the AllScan and compared with the laboratory results. These measurements were recorded by the AllScan analyser for an optimal belt loading of 105 kg/m. (The minimum operational belt loading is 75 kg/m). The resultant RMSD for each of the elements of interest is presented below.

Static Repeatability Test
The test belt was loaded with 7 bags of ‘sample 2’ to an optimal bed depth of 105 kg/m (for iron ore). This material was then placed in the analyser and the results recorded every 1 minute. The repeatability for a 1 minute and 10-minute period were then calculated and are presented below.

Belt Loading Vs Accuracy test
The belt was loaded progressively from a base loading of 60 kg/m to a maximum loading of 360kg/m. Data was collected over a one-hour period for each of the loadings below. This simulated a full range of expected belt loadings. The loadings used for each sample were;

  • 15 kg/m (1 off 15 kg bags)
  • 30 kg/m (2 off 15 kg bags)
  • 45 kg/m (3 off 15 kg bags)
  • 60 kg/m (4 off 15 kg bags)
  • 105 kg/m (7 off 15 kg bags)
  • 120 kg/m (8 off 15 kg bags)
  • 150 kg/m (10 off 15 kg bags)

During each test the required number of 15kg sample bags were stacked into the analyser tunnel on the conveyor belt.
The sample data was collected for a period of 1 hour for each loading. A sample period that consisted of an empty belt was also collected.
Data was collected on a 1-minute basis and was aggregated into a 1-hour basis for final calibration work.

Calibration Process.
Using the data from the different loadings of each samples the site elemental calibration was adjusted for the effect of this material on the background components of the model. The adjusted model was then applied to the collected data for each period and results for each element calculated and averaged for the 1-hour sample periods. It is important to understand how the resulting PGNAA spectrum is influenced by loading variations, which is explained in the
following set of graphs;

AllScan – modelling for belt load variations

Figure 2 – Belt load modelling

AllScan – modelling for belt load variations

  • Note the existence of a significant empty belt spectrum that has nothing at all to do with the sample. This is the result of neutrons interaction with the analyser structure w/supports, the conveyor belt, idlers, detector, etc.
  • The addition of a sample will result in a mixed spectrum of both sample and the surroundings, and will not just be a pure sample spectrum
  • Increasing from low numbers of bags we see the expected increasing sample response (increased count rates with increased sample mass). But this effect reaches a maximum response at 7 bags of sample and then the recorded spectrum begins starts to decline in activity. This was due to the self-shielding effect of the material on the belt.
  • After a certain point adding more sample material becomes self-defeating!

AllScan – modelling for belt load variations

  • Spectral difference with iron ore loading – each difference represents the addition of additional bags of identical material to the analyser
  • A naive interpretation of spectral linearity would propose that these three difference spectra should all be identical. Clearly, they are not.
  • Thus, just defining a simple model where a certain % change in spectral count corresponds to a certain % change in sample mass will in fact be biased and incorrect if applied over a wide range of sample loading

In other words, this is more complex than the non-linear spectral scale factor with sample mass – not only can the same changes in identical sample result in different scales of spectral response, the issue of whether these responses are linearly related by channel energy is now explored.

If the linearity hypothesis was correct, all these curves should be coincident with each other, and clearly they are not.
Measuring Mass Loading from the Spectra

  • The single strongest peak in the PGNAA spectrum is a low energy peak found at approximately 500 keV. This peak is predominantly produced above the belt in the source holder.
  • This peak is very similar in energy to a Cs-137 source (662 keV) which has found significant use in radiation-based belt scales.
  • These scales work as the attenuation of the gamma rays in this energy range is related to the density and thickness of the material, but not significantly related to the composition.
  • Using the 500 keV peak in this manner we can derive the mass loading on the belt.

AllScan – modelling for belt load variations

From the model used for (B) “the scaling coefficients” with loading are quantified in (C). This shows how a proportional change in sample loading results in a significantly non-linear response in the scale of the change in spectra.
By applying a non-linear spectral response for belt loading to the spectra the resultant spectra can be mass loading adjusted in order to result in a uniform spectral response for a given sample of material. Allying this corrected uniform spectral response to a linear decomposition algorithm then results in a decomposed elemental result that is not dependent on belt loading.

Iron Ore Test Results for Belt Loading

Dynamic Calibration Data
Site 1

Standard Error = 0.37% Absolute at 1 SD

Standard Error = 0.43% Absolute at 1 SD

Standard Error = 0.20% Absolute at 1 SD

Site 2

Standard Error = 0.48% Absolute at 1 SD

Standard Error = 0.38% Absolute at 1 SD

Standard Error = 0.18% Absolute at 1 SD

The test results for this material are very pleasing for the elements of interest. This is to be expected as the PGNAA analysis method is particularly suitable to measuring Fe, as it has a large neutron capture cross section and strong peaks in the spectrum. Typical iron ore material has high concentrations of Fe and this makes measurement of Fe simpler when utilising the PGNAA technique; however, with the high a concentration of Fe the signals the other elements are slightly harder to measure as a proportion of the total spectrum which is why the errors for the other elementals are generally slightly higher than could be expected.