Online analyser for heavy minerals grade control

Transcription

1 GAGNÉ, A., HORTH, D., RIVERIN, F., and BOURQUE, Y. Online analyser for heavy minerals grade control. The 7th International Heavy Minerals Conference What next, The Southern African Institute of Mining and Metallurgy, Online analyser for heavy minerals grade control A. GAGNÉ, D. HORTH, F. RIVERIN, and Y. BOURQUE Rio Tinto, Fer et Titane Inc., Sorel-Tracy, Canada In mining processes, the ore grade is a key parameter to control in order to reduce costs and to reach quality targets in subsequent processes. Although online analysers are readily available for fine material, analysers to measure the grade of a coarse ore such as massive haemo-ilmenite are uncommon and their reliability is low. A new analyser was developed at QIT-Fer et Titane that allows to continuously determine the haemo-ilmenite grade during transport on a conveyor, based on the density of the ore. The method consists of using a nuclear source and a level probe to measure the relative density of the ore, which is a function of its haemo-ilmenite content. This paper presents the theoretical concept of the analyser, the calibration technique, as well as the industrial results obtained on massive haemo-ilmenite during production. The accuracy of the analyser was estimated to be + 0.5%. This paper also discusses the potential application of the technique to mineral sand operations, for processes where total heavy minerals (THM) is the controlled variable. Simple modifications to the layout in place at QIT would allow for an accurate measurement of the THM content of sand fed to the concentrator as well as the THM grade of a heavy mineral concentrate and tail. Introduction Rio Tinto, Fer et Titane, operates a metallurgical complex treating massive haemo-ilmenite in Sorel-Tracy, Canada. The material fed to the complex is mined at the Tio mine in Havre-St-Pierre. The ore is transported by ship from the mine to the metallurgical complex where it is treated to feed the smelter furnaces. The massive haemo-ilmenite orebody has two predominant phases, haemo-ilmenite and silica. At the mine, the main control parameter is the haemo-ilmenite grade of the ore produced, which is measured with an online analyser. The readings of this analyser are critical as off-grade material would result in increased transportation and processing costs (energy) and difficulties in reaching quality targets for the concentrate at the mineral processing plant. For years, an Outokumpu Beltcon type analyser has been used to measure the haemo-ilmenite grade. However, with changes in technology, the company stopped manufacturing this type of analyser and replacement parts became almost impossible to find. This pushed Rio Tinto Fer et Titane to search for replacement alternatives to the Beltcon analyser. In the mid 1990s a trial was performed with a Quarcon analyser, but this did not give the expected results. In 2005, it was decided to go ahead and try a new home-made type analyser. This analyser uses the bulk density variations of the ore to measure its grade. Since the ore consists mainly of haemo-ilmenite and silica, a change in the haemoilmenite grade results in a significant change in the bulk density of the ore, which is related to its specific gravity. The approximate ore/silica ration of the mined ore is 2.3. Theory and model development The principle behind the analyser developed is not new. This method, with minor differences, has already been used in iron ore mines in Australia 1 to measure the hematite grade on a belt conveyor. The analyser developed consists of a cobalt 60 gamma-rays emitting nuclear source bombarding the material on a conveyor belt. On the other side of the belt, a detector measures the amount of gamma rays received. The signal attenuation (d/d o ) is directly proportional to the bulk density of the ore, the amount of material exposed and the material distribution on the conveyor belt (voids). As the specific gravity of the ore increases, the attenuation increases and as the level (amount) increases, the attenuation also increases. In order to eliminate the height factor, which affects the attenuation of the signal, a high precision level probe (laser type) is used. By dividing the attenuation of the source (d/d o ) by the level measurement (h), the attenuation per unit height is obtained. This ratio ((d/d o )/h) is called the d/h ratio and is directly related to the bulk density of the ore, assuming a constant size distribution (0 to 76 mm) and a constant void volume. Figure 1 shows an overview of the analyser principle. In order to get reliable results, the averages of the attenuation and the height are calculated over a 1 minute period. After 1 minute, the ratio is calculated to give the output signal of the analyser. In the calculation of the d/h ratio, the absorption value of the conveyor belt must be removed in order to get the real absorption value of the ore. By definition, the d/h ratio of the ore over a one minute interval is given by the following equation: It is well known that the grade of the ore mine is related to its specific gravity. Through the years, models were developed that relate mathematically the haemo-ilmenite grade of the ore to its specific gravity. Equation [2] shows a general relationship linking the specific gravity to the haemo-ilmenite grade. [1] ONLINE ANALYSER FOR HEAVY MINERALS GRADE CONTROL 29

2 Figure 1. Operating principle of the analyser [2] The analyser measures the bulk density of the ore on the conveyor belt. Therefore, an equation must be derived to relate the bulk density measurements of the analyser to the specific gravity of the ore. Equations [3] and [4 ]show the relation between the bulk density and the specific gravity for given haemo-ilmenite grades (a and b) where VolumeOccupied is the volume occupied by the ore for a fixed volume. [3] In the model developed, the assumption is made that the size distribution of the ore is a constant independent of the grade. This assumption has not been verified as variations may occur due to the difference in hardness of silica and ilmenite. However, since the operating range of the analyser (+ 10%) is narrow and based on industrial results, this assumption was found to be reasonable for the application. For better precision of the analyser, it should be tested. With this assumption, it is then possible to say that the volume occupied by an ore of grade a is equal to the volume occupied by an ore of grade b as depicted by Equation [5]. Equation [6] is obtained by substitution of Equations [3], [4] and [5]. [5] To solve Equation [6], it is necessary to know the specific gravity of the ore for a given grade (say b) as well as its bulk density on the conveyor. When these two parameters are known, this equation allows us to predict the specific gravity of an ore of grade a from the bulk density reading of the analyser. From experiments performed, it is known that the bulk density of the ore on the conveyor belt is directly proportional to the analyser reading (d/h) and can be represented by the following equation: [7] In this equation, it is possible to eliminate the b 2 constant as when the bulk density reaches 0, the d/h reading should be 0. Thus Equation [7] can be rewritten as [8]. [8] Substituting Equations [2], [6] and [8], it is possible to obtain an equation relating the haemo-ilmenite grade to the analyser reading, d/h, where all the parameters can be determined easily, which is Equation [9]. [4] [6] Parameters setting The AbsorptionConveyorBelt parameter in Equation [1] is set by running the conveyor without material for a fixed period of time. The average absorption during this period corresponds to the absorption of the conveyor belt. As the conveyor does not change significantly in time, this parameter is a constant and only need to be calibrated periodically. For increased precision of the analyser, it is possible to program a routine that detects when the conveyor runs empty. When it is detected as empty, the absorption readings are recorded to calculate the average and the parameter is actualized in the formula. The m 1 and b 1 constants need to be determined with laboratory studies by the development of a specific gravity versus ore grade curve. These parameters are fixed for a given ore and do not change with time, unless the mineralogy varies significantly. This relation is not necessarily proportional and varies for different ores. The SpecificGravity b constant is the specific gravity of the ore for a given grade (b in this case). It can be taken directly from the specific gravity versus ore grade curve. This parameter is fixed and does not vary with time. The m 2 parameter needs to be calibrated periodically and is specific to a given nuclear source. This parameter, relating the d/h ratio to the bulk density of the ore, needs to be determined by exposing material of known bulk density to the analyser. Although the instinctive method to determine it would be to expose ore of known grade to the analyser, this approach appeared impossible to apply. The main reason is that the measurement area of the analyser on the conveyor belt is small (about 1 cm x 10 cm). For a coarse material, it is very difficult to get a uniform grade and size distribution over the measurement area, which is crucial for the calibration. Alternatively, smaller size samples are taken, which are more uniform in size distribution and grade over a small area. Figure 2 shows the results obtained for the exposition of samples. For the tests done, samples of different ore grade crushed to 1040 microns (constant size distribution), pure metal powders and metal plates (aluminium, copper and steel) were used. These materials were chosen mainly as they were easy to procure, although other materials such as graphite and titanium, with densities closer to that of ilmenite, could have been used for the plate samples. It shows the d/h measured as a function of the bulk density of the samples. The curve, although not forced to zero is a [9] 30 HEAVY MINERALS 2009

3 d/h Bulk Density (g/cm 3 ) Figure 2. Relationship between bulk density and d/h ratio Figure 3. Effect of increasing the particle size on the bulk density of the ore for a given measurement volume straight line with a very high correlation coefficient. However, these results are strongly affected by the two points located to the extreme, with bulk densities around 8 and 11 g/cm 3. It is also possible to observe that the variation around the line is wider in the small bulk density range, which corresponds to the particulate samples. This can be explained by the packing of the samples, which affects the bulk density measurements (more or less voids) from one operator to another. An important parameter also affecting the precision of the curve is the level which is difficult to measure precisely on particulate samples. This means that a very precise calibration curve is very difficult to obtain with granular type samples. Calibration of the apparatus with metal plates appeared to be the most precise and effective way to obtain the calibration curve for the analyser. Figure 3 helps to understand the reason. For a given volume (the square area on Figure 3), if the particle size is increased, the particles start to expand outside the fixed volume. As a result, the volume filled by voids is decreased. As the particles get infinitely large, the volume available gets fully filled by the sample and as a result, the bulk density of the material exposed becomes equal to the specific gravity. As the analyser has a fixed measurement volume, exposition of a metal plate is equivalent to exposing infinitely large particles of known height and bulk density, which is equal to its specific gravity in this case. So, for instance, exposure of an aluminium plate of a given height with a specific gravity of 2.7 g/cm 3 is equivalent to the exposition of an ore of known bulk density (independent of its particle size and grade) of 2.7 g/cm 3. Also, since the height (h) of a plate is easy to measure, the error associated with the level measurement is significantly reduced. This allows one to determine very precisely the relation between the bulk density of the ore and its d/h reading, for a given nuclear source. For calibration, three aluminium plates, three steel plates and three copper plates are used. Each of them has a known height and specific gravity. Each plate is exposed three times to the analyser to get the absorption readings of each plate independently. The absorption (d), excluding the belt, is divided by the known height (h) and this gives the d/h reading for each plate. The measurements are plotted as a function of the bulk density (equivalent to specific gravity) with the curve forced to intercept the y-axis at zero and a very precise value is obtained for the m 2 parameter. After all the previous parameters have been set, the remaining parameter to determine is the BulkDensity b. This parameter represents the bulk density of a material of grade y, but for its actual disposition on the conveyor belt. Therefore, measurement of this parameter in a laboratory is not possible. This parameter needs to be adjusted through manual sampling. The ideal way to do it is to take a fixed period of time where it is possible to get the analyser readings and a representative sample of the ore at the same time. Referring back to Equation [9], since the average ore grade and d/h readings are known, the only unknown variable is the BulkDensity b, which is determined by ONLINE ANALYSER FOR HEAVY MINERALS GRADE CONTROL 31

4 solving the equation. For greater precision and confidence in the parameter, sample QIT on four different periods and solve to determine the value of the parameter by minimizing the sum of square residuals between the analyser readings and the samples grade measurements. It is important to note that if changes are made to the distribution of the ore on the conveyor belt, this parameter needs to be recalibrated. Industrial results The availability of two ore analysers on the same conveyor, using two different measurement principles (new analyser uses absorption and the old analyser uses reflection), allow one to validate easily the readings of the developed analyser. Figure 4 shows the measurements of both analysers for a one-week period. It shows that measurement tendencies follow each other very closely, meaning that both analysers give good measurements as they react similarly, unless both analysers are wrong. Figure 5 shows the comparison of the measurements from both analysers for a 10-hour period. This figure allows one to observe the tendencies which follow each other very closely again. The noisy appearance of the signals from both analysers may be attributed to the non-uniform distribution of the material on the belt surface (voids distribution). A solution to that problem would be to install a scraper on the belt to level out the material upstream of the online analyser. These two figures showed that the new analyser is a robust alternative to the old analyser with a strong advantage over it; the theory behind it is simple and well understood. This makes troubleshooting easy. To validate the reliability of the calibrated analyser to predict grade, the analyser was tested on four different mt lots, corresponding to the size of a ship load. For each lot, the production period was known at the mine so that the average haemo-ilmenite grade read by the analyser for each ship was known. When the ship was discharged at the metallurgical complex, about 30 samples of the mt were taken (standard deviation shown on graph), prepared in the laboratory and analysed for grade. Figure 6 shows the results comparing the analyser reading and the chemical analysis measured by the laboratory. The standard deviation between the analyser and the laboratory measurements is 0.8% on the haemo-ilmenite grade. It is important to note that this is not the precision of the analyser, but is rather the average difference that is expected between the analyser reading and the laboratory measurement at ship unloading on a mt lot. The difference includes the analyser error (including size distribution variations), the sampling error, and the laboratory error. The precision of the analyser to predict the grade of a mt lot would therefore be smaller. Date Figure 4. Comparison of both analysers measurements over a 1-week period Old Analyser Haemo-ilmenite Grade Measurement (%) New Analyser Measurement (d/h) Old Analyser Haemo-ilmenite Grade Measurement (%) New Analyser Measurement (d/h) Date Figure 5. Comparison of both analysers measurement over a 10-hour period 32 HEAVY MINERALS 2009

5 However, due to the difficulty in sampling directly at the mine, a rigorous R&R study could not be performed to determine this error. Application to the heavy minerals industry This analytical technique could be applied to any industry dealing with separation of minerals with differences in their specific gravity. The Mandena ilmenite deposit in Madagascar 2 is an example where this type of technology could be applied. It separates heavy minerals from silica to produce ilmenite. As the total heavy minerals (THM) of the ore contains only 14% of non-valuable minerals, the plant is designed to recover the totality of the heavy minerals (HM) with no distinction between valuable and nonvaluable HM. Therefore, the analyser described in this paper could be used to measure the THM grade continuously for different streams of the plant as specific gravity of the material is directly related to the THM grade. The uniformity and small variations of the size distribution through time would make the BulkDensity b parameter very stable, increasing the accurary of the apparatus compared to coarse ores. Although this analytical technique cannot be applied directly to pressure pipes, small modifications to the flowsheet of a spiral plant would allow the application of this technique to the heavy minerals industry as depicted on Figure 7. The idea would be to take the rejected portion of already installed online samplers and direct it to a small filtering conveyor to remove excess water. Even if the water could not be entirely removed, the use of the filtering conveyor would bring the moisture content of the cake to a relatively constant level, which would bring a constant bias to the analyser readings that can be included in the calibration. The analyser would be installed on the filtering conveyor to continuously monitor the THM grade. As a sample of the ore exposed to the analyser would still be taken to the laboratory for complete chemical analysis, continuous calibration would be possible by adjusting the BulkDensity b automatically based on the control laboratory results. This would allow one to eliminate the effect of moisture, size distribution and mineralogy variations, which could change over a long period of time and affect the readings of the analyser. Continuous actualization of the parameters would make the analyser very accurate in monitoring the THM online. The main use of this technology in the heavy minerals industry would be for spiral circuit grade control. As proposed by A. Steinmüller 3, the three main factors affecting the efficiency of a spiral separator are the heavy minerals feed grade, the feed flow rate and the feed density (TPH). Generally, feed flow rate and density are well Chemical Analysis Haemo-ilmenite Grade (%) New Analyser Haemo-ilmenite Grade(%) Figure 6. Chemical analysis haemo-ilmenite grade measurement versus new analyser grade measurement Figure 7. Application to the heavy minerals industry ONLINE ANALYSER FOR HEAVY MINERALS GRADE CONTROL 33

6 controlled, which makes the main variable affecting the performance of a spiral separator to be the feed grade. Although this parameter does not usually change rapidly, mining of different zones brings in small feed grade variations which, over time, affect significantly the grade and recovery of a plant. Either feed back control, by monitoring the concentrate grade of a plant, or feed forward control, by monitoring the feed grade to the plant could be performed with the analyser measurements. The measurements could be used to control the density or volumetric flow rate through the different stages of the circuit in order to regularize the performance of the spirals. The advantage over the technique described in 3, which consists of using an optical method to monitor the position of the heavy minerals band on one spiral, would be that slimes would not affect the measurement; it would be possible to analyse the entire circuit as opposed to one spiral, giving a real picture of the circuit; and finally, as measurement of the THM grade of the feed would be possible, application of feed forward control to the circuit would be possible. Conclusion A new online analyser was developed and successfully tested by Rio Tinto, Fer et Titane. The operating principle is simple and robust, the theory behind it is well understood and its application to various processes would be very simple. Plant experience and sampling showed that it is accurate to predict the haemo-ilmenite content of an ore. Adaptation of this technique to the heavy mineral industry would be simple and could allow for feed forward or feedback control of spiral circuits to improve significantly the control on concentrate grade and heavy minerals recovery on ores where THM is the controlled parameter. Reference 1. HOLMES, R.J., McCRACKEN, K.G., and WYLIE, A.W. Analysis utilising neutron irradiation, Australian patent Appl. PA9294, DUMOUCHEL, J., GIROUX, J., MEAD, M., and YULE, W. The QIT Madagascar minerals deposits geology, mining & mineral processing, Heavy Minerals Conference 2005, STEINMÜLLER, A. Development and application of simple process models for spiral concentrators, Heavy Minerals Conference 2005, A. Gagné Process Engineer, Rio Tinto Fer et Titane, QIT Madagascar Minerals Division Worked for QIT Fer et Titane for 2 years in Canada as a process engineer were I worked primarily in plant control, quality control and development projects. Since 2007, I have been working as a process engineer on the commissioning of QIT Madagascar Minerals as the hot commissioning leader for the mineral separation plant including the feed preparation plant, dry ilmenite production plant, wet zircon plant and dry zircon plant. 34 HEAVY MINERALS 2009