Vendor column: If you aren’t controlling your process, your process is controlling you

Skip to Navigation


  • Published: Nov 29, 2016
  • Author: Daniel Shiley
  • Channels: Infrared Spectroscopy / Chemometrics & Informatics
thumbnail image: Vendor column: If you aren’t controlling your process, your process is controlling you

Daniel A. Shiley – Senior Application Chemist, SummitCAL Solutions Team, ASD Inc.

This is the fourth and final segment in the series on near-infrared for the mining industry.

In the three previous articles I have been discussing the use of near-infrared spectroscopy (NIR) and its importance in the mining industry. From early exploration using remote sensing, to ground-based studies, to the measurement of drill core and finally to making quantitative measurements of minerals and metallurgical properties in the lab or over a conveyor belt, NIR provides useful, non –destructive rapid analyses throughout the mine development cycle.

The key problem

Many mines do not test enough or they utilize analytic tests that are so time-consuming that the results cannot be used effectively for control of the mine. This lack of rapid, relevant information causes lost productivity, expense and inefficient plant operation. Without timely data the mine essentially becomes reactive rather than proactive. The process effectively controls them, rather than mine operators controlling the process! Addition of NIR technology to the mine control system can provide the time-sensitive data needed to keep the mine operating at peak efficiency to meet shareholder expectations.

All rocks are not created equally

Although we certainly would like to measure every rock with NIR, many rock-forming minerals such as feldspars and quartz have no spectral features in the NIR. As I have previously discussed, many important minerals, such as alteration minerals and clays can be successfully measured in the NIR. These are minerals that are used as pathfinders to the mineralization and provide the exploration geologist with essential information to vector in on the areas that may contain metal. NIR use during exploration is focused on qualitative mineral abundance as the presence or absence of certain minerals is sufficient to provide useful information. During the production process, clays often influence operations causing lost productivity or add extra expense or upsets in the processing plant. The quantitative measurement of clays allows the mine to maintain peak productivity

Metals, elements and metallurgical response

Measurement of metal or elemental analysis is not generally recommended using NIR, again due to the lack of spectral features in the NIR region. However if the metal always occurs in combination with other minerals that do have spectral features, then the presence of metal can be inferred by the presence of the indicator mineral. This is exactly the common use of NIR during the exploration process, but this is an indirect rather than direct relationship between NIR and metals or elements and is not quantitative. We believe that combination of NIR technology with X-ray fluorescence (XRF) analysis to measure elements and metal content is a great partnership of technology. NIR and XRF each provide a part of the information needed to economically operate a mine. Certain metallurgical responses, such as acid consumption can be effectively measured using NIR. Acid consumption is a measure of the acid neutralization characteristics of the ore and is influenced by the mineralogy. In fact many companies measure mineralogy so they can estimate the acid consumption. With ASD’s mineral analyzers, it is possible to measure acid consumption of the ore directly.

What is success anyway?

People often believe that it is important to have a high correlation coefficient to have a successful calibration model with R-squared values greater than 0.95. Certainly we all would like to have an R-squared value like that for every constituent; however this should not be an absolute requirement and actually is a rather poor indicator of model performance. In order to be useful, the model must be capable of producing values that can be used to control the process. Surprisingly, this can mean that even a model with an R-squared value of 0.75 may be useful. It all depends upon the level of error that is allowed to make a positive change in the way the plant operates. Also the range of the constituent values affects the R-squared statistic; the R-squared value is simply not the most important factor. The statistics that best describe the calibration error are the standard error of cross validation (SECV) and the standard error of prediction (SEP). SECV is produced during the modeling process and is an estimate of what the error of the model predictions will be on new samples, whereas SEP is measurement of actual error on an independent test set. Various chemometric programs may have slightly different names for these statistics but they are the measurement of error that can be most helpful in assessing a model.

Figure 1 contains two plots comparing the percentage of spectrally active minerals in ore. In these plots we see the values of kaolinite, chlorite, biotite and muscovite reference values on the left compared to ASD NIR predicted mineral values on the right. In general, we see a very good relationship between the reference values and the predicted values and even though there were other minerals present that had no spectral features, such as quartz and feldspar, we still had predicted values from the NIR that were quantitative. These results would be very useful to help control a process.

Figure 1. Plot of reference mineralogy (left) and NIR predicted mineralogy (right)

To obtain quantitative results like those shown in Figure 1, we need to examine a representative portion of the ore. Samples sent to the laboratory can be crushed so the sample presented to the ASD TerraSpec® 4 Standard-Res mineral analyzer is homogeneous and representative of the ore. Very large amounts of uncrushed ore sample may also be measured to obtain a representative spectrum for the material. This can be accomplished by either rotating a large portion of the sample under the sample optic head, such as a pan of blast chips, or by installing the QualitySpec 7000® spectrometer over a conveyor belt. In either case the spectra of rock directly beneath the optic head is averaged with other rock from the conveyor or pan to obtain the representative spectrum needed for quantitative measurement.

Never before capabilities!

Through the use of the QualitySpec 7000 analyzer measuring blast chips or over the conveyor belt, mining companies now have a method to determine mineralogy or metallurgical parameters in real-time. This is a much more rapid measurement than in the lab because there is no need for any sample preparation. This gives the mine operations team a tool that can be used to help them optimize their processes to make the best economic use of the ore that is going through the plant.

This mineralogy can be used to aid in the construction of the leach heap so that certain minerals do not occur in such abundance as to cause problems with permeability of the heap. Real-time mineralogy can also provide an indication when certain problem minerals are entering the floatation circuit. This information enables the operators to add certain chemicals to the cell to mitigate the effect of these minerals. Without such knowledge the purity and quality of the concentrate can be reduced by these problem minerals. Acid consumption values are used to determine how to efficiently manage the leach heap.

Onward to the future

Typically in the past, mines have either relied on slow, expensive reference analyses that were not suitable for plant control, or they simply have done no testing at all until they detected a problem. Because of the variable nature of ore, the problem might not even exist by the time the reference data had been generated and reported. Mines now have the opportunity to create proactive control strategies rather than trying to be reactive. ASD’s QualitySpec 7000 analyzer is the tool that can be used to propel mines from the old-school reactive mentality into the efficient, economical, knowledge-based decision making era.

Mining A-Z with NIR

As I have discussed during these last four articles, NIR has an important place in fulfilling the analytic needs of mining programs. ASD’s NIR systems are used from early exploration utilizing a remote-sensing approach to qualitatively identify areas of alteration. They are also used to aid in mapping and creation of initial block models by qualitatively measuring drill core. Mine site laboratories use ASD systems to provide quantitative measurement of pulverized materials. Finally the QualitySpec 7000 system can provide quantitative measurement of mineralogic and metallurgical parameters on the conveyor belt. Contact ASD with questions on how this technology can best be implemented to solve your most difficult analytic needs for the mining industry.

Article by Daniel Shiley

The views represented in this article are solely those of the author and do not necessarily represent those of John Wiley and Sons, Ltd.

Social Links

Share This Links

Bookmark and Share


Suppliers Selection
Societies Selection

Banner Ad

Click here to see
all job opportunities

Copyright Information

Interested in separation science? Visit our sister site

Copyright © 2017 John Wiley & Sons, Inc. All Rights Reserved