By C. Addison, R. Jones, F. Addison and F. Stanley, Alpha Natural Resources; and G. Luttrell and R. Bratton, Virginia Tech

Dense medium cyclones (DMCs) have become the workhorse of the coal preparation industry due to their high efficiency, large capacity, small footprint and low maintenance requirements. Although the advantages of DMCs make them highly desirable, field studies have shown that some industrial DMC circuits suffer from poor control due to fluctuations in the density cutpoint caused by inconsistencies in feed quality. To address this problem, a multi-stream monitoring system was designed and installed to simultaneously monitor the feed, overflow and underflow circulating medium from an industrial dense medium cyclone circuit. Algorithms were then developed to use these real-time measurements to predict and optimize the density cutpoint in response to variations in feed consistency. This article will describe the key features and benefits of this new approach to monitoring/control of industrial DMC circuits.

DMCs are used in the coal preparation industry to clean particles in the 50- to 0.5-mm size range. These high-capacity units use centrifugal forces to enhance the separation of fine particles that cannot be upgraded in static heavy medium separators. In the U.S. alone, DMC circuits account for an annual production of about 160 million metric tons (mt) of clean coal (36% of total U.S. washed coal) and represent an installed capacity approaching 75,000 t/hr. Due to the high tonnage, a small increase in DMC efficiency can have a large impact on plant profitability. Estimates suggest that a modest 2 percentage point increase in the DMC efficiency would produce 3.2 million tons of additional clean coal in the U.S. from the same tonnage of mined coal. At a market price of $68/ton, the recovered tonnage represents annual revenues of about $218 million for the U.S. coal industry or more than $850,000 per year for an average plant.
Most of the early expertise related to the design and operation of DMC circuits was developed by researchers working at Dutch State Mines under the direction of M.G. Driessen in the late 1940s and early 1950s. This renowned group developed a large body of test data that was not publicly disclosed for commercial reasons and was lost when this group ceased operations in the late 1960s. As a result, several follow-up studies were conducted in the U.S. and aboard to address the lack of operating and design guidelines for DMC circuits (Gottfried and Jacobsen, 1977; Napier-Munn, 1984; Rao et al., 1986, Chedgy et al., 1986; Davis, 1987; King and Juckes, 1988; Scott, 1988). Some of the most noteworthy studies were carried out by researchers at the Julius Kruttschnitt Minerals Research Centre (JKMRC) in Australia (Wood et al., 1987; Wood, 1990; Wood, 1997). In some cases, these researchers documented yield losses of as much as 15% due to problems with DMC circuits. Some of the most common troubles encountered in these studies included clean coal overload, excessive particle retention, and incorrect SG cutpoints. In many cases, these problems were corrected by simple low-cost modifications to plant circuitry or operating protocols.

Effect of Differential Cutpoints
DMCs are frequently installed in banks of two or more parallel units or in parallel with other separators (such as dense medium baths) in order to meet the production requirements of a given plant. Theoretical analyses show the clean coal yield from these parallel circuits is maximized when all of the separators are operated at the same specific gravity cutpoint (Abbott, 1981; Abbot and Miles (1990), Clarkson, 1992; Luttrell et al., 2000). This optimization principle is valid regardless of the desired quality of the total clean coal product or the ratios of different coals passed through the circuits. To illustrate the importance of this concept, consider a 500-tph circuit consisting of two identical DMCs operating in parallel. Both of the DMCs are capable of producing an 8% total ash product when they operate at the same cutpoint of 1.55 SG. The overall yield from these two DMCs is 69.6%. However, the two units can also produce a combined clean coal ash of 8% by operating the first DMC at 1.59 SG (which produces a 8.5% ash) and by operating the second cyclone at 1.51 SG (which produces a 7.5% ash). Although the combined product is still 8% ash, operation at a cutpoint difference of 0.08 SG units reduces the overall yield from the combined circuit from 69.6% to 68.2% (i.e., 1.4 percentage points less). If the cyclones are operated for 6,000 hrs per year, the annual revenue lost due to the cutpoint difference is $2.9 million annually (i.e., 1.4% x 500 ton/hr x 6,000 hr/yr x $68/ton = $2,856,000). As shown in Figure 1, the revenue loss becomes even more severe as the target quality of the clean coal is reduced from 8% to 6% ash for the combined circuit. In fact, at the 6% ash target, a cutpoint difference of just 0.04 SG units between the two DMCs results in lost revenues approaching $2 million annually for this particular coal. Therefore, it is very important that all heavy medium circuits (baths and DMCs) be operated at the same SG cutpoint to optimize total plant profitability.

The industrial application of cutpoint optimization is relatively straightforward for heavy medium baths. Baths tend to operate at a density cutpoint that is predictable based on the SG of the feed medium. On the other hand, the segregation of medium by the centrifugal field within a DMC makes it very difficult to estimate the true SG cutpoint for cyclones. As shown in Figure 2, the underflow medium from a DMC has a substantially higher SG than that of the overflow medium due to classification of the magnetite particles. The thickening of the medium tends to increase the SG cutpoint for the DMC above that of the feed medium SG. For the case shown in Figure 2, the actual cutpoint of the DMC is about 0.05-0.10 SG units higher than that of the measured SG of the feed medium. This “offset” between true and measured density can vary substantially depending on the feed medium density, extend of cyclone wear, and characteristics of the feed coal. As a result, the normal practice of on-line monitoring the feed medium SG using nuclear density gauges cannnot be used to accurately estimate the true cutpoint for DMCs. As discussed previously, this inability to estimate and maintain the SG cutpoint can result in coal losses that have a tremendous impact on plant profitability.

Estimation of SG Cutpoint
The separation performance of a DMC can be predicted using a partition model which assumes the partition curve for each particle size class passes through a common pivot point (Scott, 1988). The specific gravity (SG*50) corresponding to the pivot point can be estimated from an empirical expression given by Wood (1981):

SG*50 = 0.360SGfm + 0.274SGum + 0.532SGom – 0.205     [1]

where SGfm, SGum and SGom are the specific gravities of the feed, underflow and overflow streams, respectively. The SG*50 value represents the effective SG cutpoint of a large particle separated under a zero medium viscosity, i.e., a nearly perfect separation. The second defining term for the pivot point is obtained at a partition number that is numerically equal to the medium split to underflow (Su) given by (Restarick and Krnic, 1990):

Su = (SGfm – SGom)/(SGum – SGom)     [2]

Once the pivot point is identified, the specific gravity cutpoint (SG50) for each particle size class can be obtained using (Wood, 1990; 1997):

SG50 = SG*50 + 0.910Epln[(1-Su)/Su]     [3]

To use this expression, it is assumed the unknown Ep value for each particle size class can be estimated using various empirical expressions (Barbee et al., 2005).

Multi-Stream Monitoring & Control
Equations [1]-[3] indicate it is possible to predict and properly optimize the SG cutpoints for a DMC provided the values of SGfm, SGum and SGom are known. Unfortunately, only the feed medium density (SGfm) is typically measured in most industrial DMC circuits. Also, density for the feed medium (SGfm) is often measured with coal present so that the true medium density is not known. To overcome this limitation, an improved monitoring and control system was developed that uses multi-stream on-line measurements of the feed, overflow and underflow medium densities using low-cost nuclear density gauges and pressure transmitters. A schematic of the multi-stream monitoring system is provided in Figure 3.

The multi-stream monitoring system uses four nuclear density gauges to simultaneously monitor medium density throughout the entire circuit. The first density gauge (P), which was already installed at the plant, was used to monitor and control the density of the circulating medium that was fed to the DMCs. Unfortunately, since this stream also contains coal particles, the reading from this density gauge does not necessarily represent the true density of the circulating medium. To determine the true medium density, a small slipstream from the DMC feed line was passed across a small sieve screen. The underflow from the sieve screen reported to another nuclear density gauge (F). Two additional nuclear density gauges (U and O) were installed in the return medium streams from the drain sections of the overflow and underflow drain-and-rinse screens. These four measurement points made it possible to monitor the density of the feed medium (with and without coal present), underflow medium and overflow medium. Data from the electronic sensors was continuously logged online using a PLC data recorder. In principle, the real-time data from these sensors can be passed through a mathematical algorithm to estimate the “true” SG cutpoint for the DMCs (see Equation [1]). As such, this information makes it possible to fully optimize DMC cutpoints under conditions of changing coal types and feed blends.

Equipment Setup
The nuclear density gauges were mounted in custom fabricated portable racks. A vertical feed pipe above the gauge was used to ensure a high velocity flow that prevented any settling of the magnetite through the system. The vertical feed pipe was fitted with an overflow at the top. Flow to the rack and density gauge was set to provide an overflow stream at the top of the vertical feed pipe to ensure a full feed pipe and to eliminate any air bubbles in the medium passing through the gauge. The medium that passed through the nuclear density gauge and from the overflow was routed back to the DMC feed sump.The density gauge rack for the underflow medium sample was installed, along with the associated sampling points and piping, to receive medium flow from either the clean coal or refuse drain-and-rinse screens.

After the installation of the density gauges, manual density (Marcy) cup measurements were taken and flows were established to insure that the flow through the gauges was an accurate representation of the actual medium flows around the DMC. The next step involved energizing the three nuclear density gauges, checking the electrical connections, setting the proper configuration parameters, and then standardizing the gauges with clear water. Circulating medium was then routed through the gauges for the calibration procedure. A pulp density scale, calibrated using a two-point procedure at 1 and 1.6 SG, was used to obtain an accurate medium SG for the density gauge calibration. The process signals from the density gauges representing the SG for the streams were connected to the plant PLC control system via 4- 20 mA analog inputs.

Data Logging
A dedicated data monitoring and logging system was developed to retrieve the relevant process information from the plant PLC control system during the sampling periods. The system provided a means to monitor the process data while collecting samples and logged all the data to a computer text file. The process data was logged every five seconds, with a time and date stamp, and included:
•    Plant Feed Rate (tph)
•    Clean Coal Rate (tph)
•    Plant Secondary DMC Feed Medium Density (used for circuit SG control)
•    Project Secondary DMC Feed Medium Density (from sieve screen underflow)
•    Project Secondary DMC Overflow Medium Density
•    Project Secondary DMC Underflow Medium Density
•    Differential Pressure Cell Transmitter
•    Calculated SG Cutpoint
The differential pressure (d/p) cell transmitter, noted in the above list, was installed on the vertical feed pipe for one of the density gauges and the signal was intended to be logged in an attempt to correlate that data with the density gauge data. Problems were encountered in obtaining reliable data from the d/p cell transmitter, so this data was not reported in this article. The calculated SG cutpoint for particles was based on the mathematical model described by Equation [1]. While the accuracy of this equation was not verified in the present work, it was believed to provide a relative indicator of the expected cutpoint densities for particles passing through the DMC circuit.

System Testing
Three series of test runs were conducted at low, medium and high SG setpoints using the multistream monitoring system. In each run, the values for the feed, underflow and overflow medium were recorded using density gauges “F”, “U” and “O”. The density readings from these gauges were used in Equation [1] to estimate the expected SG cutpoint (SG*50) for the DMC circuit. The reading from the existing plant density gauge (“P”) was also recorded. At the midpoint of each test run, the feed coal to the circuit was intentionally switched from a low-ash feed coal containing a low amount of reject rock to a high-ash feed coal containing a large amount of reject rock so the effects of coal type on control system performance could be established. The results obtained from all three sets of test runs are summarized in Table 1.
For the low-reject feed, a relatively constant value of 1.33 SG was obtained by both the plant gauge (P) and the slipstream feed gauge (F). The density values for the overflow and underflow streams were found to be about 1.21 and 1.57 SG, respectively. Based on these values, the SG*50 value was estimated to be about 1.35 SG for this particular set of operating conditions, which was nearly identical to the reading from the plant gauge (P). However, when the plant switched to the high-reject feed, the reading from the slipstream feed gauge (F) dropped by about 0.02 SG to about 1.31 SG. The reason for the drop is that the plant gauge (P) misinterpreted the extra rock in the feed as high density medium. In response, the plant control system added more water to drop the true density of the circulating medium. Under this new condition, the densities of the overflow and underflow streams changed to 1.21 and 1.50 SG, respectively. These new values inadvertantly reduced the SG*50 for the particles from an estimated value of 1.35 SG down to 1.32 SG, a net change of about 0.03 SG units. The split of medium to overflow (calculated using Equation [2]) did not change appreciably.

The density data for the test run was conducted using an intermediate SG setpoint. In this case, the plant density gauge (P) indicated that the circulating medium was 1.50 SG. The feed density (F) measured without coal showed a slightly higher value of 1.51 SG when running the low-reject feed coal. The switch to the high-reject feed coal sharply reduced this value from 1.51 SG down to 1.49 SG. Once again, the existing plant density gauge (P) and control system misinterpreted the higher rock content in the high-reject feed coal as too much medium and reduced the density. As such, the SG*50 for particles in the DMC circuit changed from about 1.54 SG to 1.51 SG when switching from a low-reject to high-reject feed coal. This unexpected change was not apparent in the readings from the plant density gauge (P) which remained relatively constant at 1.50 SG during the entire test period. The large change was noted in the medium split to overflow, which shifted from about 0.62 to 0.57 when the characteristics of the feed coal changed.

Finally, the density values were obtained for the test run performed using a very high SG setpoint. While more variability was observed in the plant density gauge (P) readings during this particular run, the data still showed a strong dependence between coal type and true medium density. For the low-reject feed coal, the true medium density reported by the feed gauge (F) was significantly higher than that from the plant gauge (P). In contrast, the trend was exactly opposite when running a low-reject feed, i.e., the true medium density was significantly lower than the plant gauge reading. According to the measured SG values, the cutpoint density for the cyclone estimated using Equation [1] shifted from 1.81 SG to 1.79 SG during the course of this test run. The medium split to overflow also shifted from 0.61 to 0.66 in response to the changing feed coal type.

The difference between the apparent density of the circulating medium (obtained from gauge P) and the true circulating medium density (obtained from gauge F) is shown in Figure 4. When running a low-reject feed coal, there is little observed difference between the true medium density and that reported by the plant density gauge (P). On the other hand, the true medium density appears to be about 0.014 to 0.026 SG units lower than the plant gauge reading when running a high-reject feed coal. As described in the introduction to this article, these differences make it difficult to properly optimize dense medium circuit performance in cases where the plant feed coal characteristics routinely change throughout the production period.

Figure 5 shows SG*50 estimated from Equation [1] as a function of the apparent density of the circulating medium. The data shown in this plot indicates that the plant density gauge provided a good indication of the SG*50 when operating under conditions of low-to-medium density with high-reject feed coals. However, in the high SG range, a difference of 0.052 SG units was observed between the plant SG reading and the estimated particle SG cutpoint. A higher SG*50 was also observed for all cases when running low-reject feed coals. This difference increased steadily as the plant density was raised (i.e., 0.02 SG units for low density, 0.04 SG units for medium density, and 0.09 SG units for high density). These results again demonstrate the importance of knowing the true densities of the circulating medium streams that pass through the DMC circuit as feed, overflow and underflow. Without these values, proper control and optimization of DMC circuits does not appear to be realistically possible when processing feed coals with highly varying characteristics.

Test data collected in the current study indicate that optimization of dense medium cyclone (DMC) performance cannot be realistically achieved for cases in which only the feed medium density is monitored in the presence of coal. This problem appears to be created by incorrect density readings which interpret the presence of large amounts of high-density rock as overdense medium. To avoid this error, it is recommended that plant circuits be designed with a means to monitor the true density of the circulating medium in the absence of feed coal. Ideally, a multistream monitoring station is recommended to simultaneously measure the overflow, underflow and feed (recombined overflow and underflow) medium steams just below the drain-and-rinse screens. This layout requires that the plant be designed with sufficient headroom for the multistream monitoring station between the drain-and-rinse screens and the DMC feed sump. As an alternative, a correct medium sump and pump arrangement could be incorporated into the DMC circuitry. Data from the monitoring station can also be used to determine the medium split around the DMC circuit, which is a useful indication of whether any major equipment problems have occurred with the DMC units (e.g., apex excessively worn or blown out).
This article was adapted from a paper Bratton presented at Coal Prep 2009.

•    Abbott, J., 1982. The Optimisation of Process Parameters to Maximise the Profitability from a Three-Component Blend, 1st Australian Coal Preparation Conference, April 6-10, Newcastle, Australia, 87-105.
•    Barbee, C.J., Luttrell, G.H., Wood, C.J., and Bethell, P.J., 2005. “Simulation of Heavy Medium Cyclone Performance,” Minerals & Metallurgical Processing, Vol. 22, No. 1, pp. 38-42.
•    Chedgy, D.G., Watters, L.A., and Higgins, S.T., 1986. Heavy Medium Cyclone Separations at Ultra-Low Specific Gravity, 10th International Coal Preparation Congress, Edmonton, 60-79.
•    Clarkson, C.J. & Wood, C.J., 1991. A Model of Dense Medium Cyclone Performance, Proceedings of the 5th Australian Coal Preparation Conference, Australia, 65-79.
•    Davis, J.J., 1987. A Study of Coal Washing Dense Medium Cyclones, PhD Thesis, University of Queensland, Australia.
•    Gottfried, B.S. & Jacobsen, P.S., 1977. A Generalized Distribution Curve for Characterizing the Performance of Coal Cleaning Equipment, USBM. Report 8238.
•    King, R.P. & Juckes, A.H., 1988. Performance of Dense Medium Cyclone when Beneficiating Fine Coal, Coal Preparation: An International Journal, 5: 188-210.
•    Luttrell, G.H., Catarious, D.M., Miller, J.D., & Stanley, F.L., 2000. An Evaluation of Plantwide Control Strategies for Coal Preparation Plants, Control 2000, J.A. Herbst (Ed.), SME, Littleton, CO, 175-184.
•    Luttrell, G.H., Barbee, C.J., Wood, C.J., and Bethell, P.J., 2003. “Operating Guidelines for Heavy-Media Cyclone Circuits,” Coal Age, Vol. 108, No. 4, April 2003, pp. 30-34.
•    Napier-Munn, T.J., 1984. The Mechanism of Separation in Dense Medium Cyclones, Ph.D. Thesis, University of London, England.
•    Rao, T.C., Vanagamudi, M., & Sufiyan, S.A., 1986. Modelling of Dense Medium Cyclones Treating Coal, International Journal of Mineral Processing, 17: 287-301.
•    Restarick, C.J. & Krnic, Z., 1990. Effect of Underflow/Overflow Ratio on Dense Medium Cyclone Operation, 4th Samancor Symposium on Dense Medium Separations, Cairns.
•    Scott, I., 1988. A Dense Medium Cyclone Model Based on the Pivot Phenomenon, Ph.D. Thesis, University of Queensland, Australia.
•    Wood, C.J., Davis, J.J. & Lyman, G.J., 1987. Towards a Medium Behavior Based Performance Model for Coal-Washing Dense Medium Cyclones, Australian IMM, Brisbane, 247-256.
•    Wood, C.J. 1990. A Performance Model for Coal-Washing Dense Medium Cyclones. PhD Thesis (unpublished), University of Queensland, Australia.
•    Wood, C.J., 1997. Coal Preparation Expertise in Australia: In-Plant Issues and the Potential Impact of Broader Applications, Proceedings of Coal Prep ’97, Lexington, Kentucky, 179-198.