Introduction

The roasting process produces a gas stream that contains heavy particulate loadings, strong SO₂, CO and NO_x, and significant quantities of mercury. An elaborate roaster gas cleaning system was installed as part of the project. The design of the system was driven by two key factors:

• Construction could not commence until environmental permits were received. In accordance with Barrick Gold Corporation Environmental Policy, a gas cleaning system was engineered to achieve very high collection efficiencies for particulate, SO₂, mercury, CO and NO_x. These design criteria also led to an expedited permitting timeframe.
• It was essential that the gas system perform reliably. A shutdown of the Roaster gas system could result in shutdown of the entire facility.

This paper describes the approach taken to design the gas system, the measures included to achieve the stringent performance and reliability requirements, and the actual performance of the system.

Description of Roasting Process

Figure 1 shows a flow diagram for the overall facility. .

Go to Figure 1

The ore is first crushed to –25 mm and stockpiled ahead of dry grinding. It is then fed to two double-rotator ball mills that dry the ore and grind it to 80% passing 74 µm. The fine dry ore is fed to the roasting plant comprised of two, two-stage oxygen roasters as shown in Figure 2.

Go to Figure 2

Each roaster was designed to process 6,000 tons per day of ore with 500 tons per day of oxygen. The flow of gases and solids is countercurrent. Dry feed is fed to the first stage that operates at a bed temperature of about 550ºC. Feed to the first roasting stage is exothermic and primary temperature control of the bed is by water quench. Coal can be added to the ore feed if the sulfide sulfur and/or organic carbon content are low. The ore is then transferred to the second stage where it contacts the concentrated oxygen gas stream. Process gas from the second stage is cleaned in cyclones and transferred to the first stage. The calcined ore leaving the second stage is quenched with water, neutralized with lime and then thickened prior to conventional carbon-in-leach processes.

TABLE 1 presents the off-gas conditions at the exit of the first-stage cyclones.

The off-gas has numerous characteristics that affect its handling and cleaning:

• Dust loads (about 10 tons/hr) are extremely high.
• SO₂ concentrations are both high and variable.
• The ore contains significant quantities of mercury that are volatilized in the roaster process and report to the gas system.
• CO loadings were normally expected to be fairly low, but some of the test work conducted showed very high levels of about 275 kg/hr.

One of the key characteristics of the roaster off-gas system is its variability. The roasters were installed to treat a number of different ores with different amounts of contaminants. Some are highly exothermic and require significant quantities of water to be sprayed into the top bed for temperature control, while others are endothermic and require coal additions to maintain temperature. Factors such as these resulted in variable contaminant concentrations as well as variable gas flows.

Environmental Design Objectives and Permitting Process

Go to Figure 3

Existing processing facilities on the Goldstrike property were in an air district where National Air Quality Standards were being met and where all the major pollutants were less than 250 tons per year (tpy) PSD limits. Staying beneath the 250-tpy ceiling was the starting point for permitting the roaster facility. Also taken into consideration was Barrick’s commitment toward protecting the environment. Engineering and equipment selection targeted the best and most efficient technology.

Barrick’s Environmental department managed the air permitting process by working in close association with Barrick’s technical staff, Hatch, and Air Sciences, Inc. who compiled all the information for the permit. Preparation of the permit started during the feasibility portion of the project, as early on it was recognized as the critical path. The application was submitted to Nevada’s Division of Environmental Protection, Bureau of Air Quality (BAQ) on August 01, 1997 just one month after basic engineering commenced. After a 60-day review of the permit to determine completeness by BAQ, the 12-month approval process started.

To enable detailed engineering to proceed, a procurement strategy based on two-stage purchase orders was used. Equipment specifications were prepared and issued for tender, and procurement recommendations were prepared. However, only Vendor equipment engineering was purchased initially. This enabled detailed engineering to proceed without committing Barrick to the full cost of the equipment. Upon receipt of the permit, the second stages of the purchase orders were issued.

The permit was submitted with very little design and engineering being performed, and several changes to the process flowsheet occurred during basic engineering. Barrick issued a general revision to the permit in January 1998 without impacting the approval process. Ultimately, the permit for construction and operation was issued to Barrick in September 1998.

The overall permitting process required 17 months, which is a fairly short duration for a large project. The process was expedited by Barrick’s desire to maintain its reputation as an environmentally responsible corporation. The design criterion for the air emission control equipment was directly related to the air permitting strategy. Extremely low emission targets were selected for the facility, which required unit operations and equipment selection with very high operating efficiencies. TABLE 2 presents the emission limits committed to by Barrick, along with the inlet loadings for comparison.

As shown in TABLE 2, very high collection efficiencies were required to achieve the desired emission rates for many of the contaminants.

Description of Gas Handling System

Go to Figure 4

The two roaster gas systems are kept separate through particulate collection and are then tied together. Because of the very high dust loads, it was felt that attempting to combine the two systems prior to dust removal would compromise the reliability of the process. Particulate collection is achieved by multiple stages of wet cleaning. Given that roaster calcine is handled wet and the strong SO₂ present in the off-gas, a dry system was never a consideration. Gases are first cooled to saturation and partially cleaned in an open quench tower. The vessel consists of an open vertical chamber with a series of hydraulically atomized spray nozzles followed by a disengagement tank to separate the gas and entrained water/particulate. Because of the very large particulate loadings entering the quencher, it was prudent to use a fully open, low-velocity vessel to minimize problems associated with build-up and erosion. The arrangement of the vessel suits the roaster plant layout and minimizes the handling of hot, heavily loaded gas. Special attention was required at the interface between the hot gas and wet quencher to prevent heavy build-up and rapid corrosion that often occurs here.

About 90% of the incoming particulate is collected in the quencher. Removing the bulk of the particulate, especially the coarse fraction, in the low energy quench stage is important to minimize problems with the subsequent scrubbing stage. After reviewing the options with respect to scrubbing equipment, a variable throat venturi was selected. The scrubber operates at a 10 kPa pressure drop and includes alumina tiles to protect against erosion. Scrubbing water is introduced through open bore pipes enabling partial recirculation of the water. The key reasons for selection of the venturi are the absence of internal components, which are prone to build-up and erosion, and its well-proven performance.

The gases leaving the scrubbing stages are saturated at 65-80ºC. A gas-cooling step was required for a number of reasons:

• The gas temperature had to be reduced to about 40ºC to enable subsequent mercury collection (a vapor pressure issue).
• Cooling the gas resulted in substantial amounts of water being condensed. Collecting the water here enabled it to be reused, while minimizing water balance problems in downstream gas absorption systems.
• Condensation of the water reduced the actual flow of gas substantially, along with the size of downstream fans and ductwork.

Numerous equipment options were available including open spray chambers, packed or tray towers, and shell- and tube-type coolers. Two-stage cooling was also considered. A single shell and tube-type condenser was selected. The decisive reason for this was mercury; a significant portion of the metallic mercury present in the gas was expected to condense as the gas was cooled to 40ºC. The single shell and tube-type condenser did not require a ‘dirty’ heat exchanger to cool contact water with the potential for mercury accumulation. Fouling of the tubes was the main concern during the design stage but has not been a problem due to the effective gas cleaning upstream and the large volumes of condensate providing continuous washing of the heat transfer surfaces. Figure 4 also shows how water is managed at the front end of the system. Water condensed at the cooler is returned to the quenching/scrubbing stage. The quencher removes the bulk of particulate, including the coarser fractions, and uses spray nozzles so particulate-free water is used. With the bulk of the particulate removed and the absence of spray nozzles, recirculation is used at the venturi. However, the solids content of the circuit is kept below 5% wt. This circuit has enabled reliable operation while minimizing water usage and the subsequent impact on overall water balance.

At this point in the system, the gas is cold, essentially particulate-free and could be handled reliably without concerns over build-up. The two roaster systems were, therefore, combined to minimize project cost. A booster fan was included for each roaster to facilitate draft/flow control and reduce the draft in downstream equipment. The fans were selected with as low a fan speed as practical (1,800 rpm) to maximize reliability. Chevron-type mist eliminators are included upstream of each fan to collect condensate that forms in the duct runs. This minimizes the amount of condensate to be drained via the fan casings.

The combined roaster gases are then cleaned in a wet electrostatic precipitator (wet ESP). Wet ESPs are efficient collection devices for very fine particulate and mist, and play a key role in all metallurgical acid plants. The wet ESP was not part of the original flowsheet but was included to serve several functions:

• Complete collection of particulate.
• Collect fine condensed forms of mercury and mercury compounds.
• Remove acid mist (SO₃ + H₂SO₄), which would cause opacity problems at the stack.

Mercury removal is then completed by scrubbing with mercuric chloride, HgCl₂ (Boliden-Norzinc process). The process is well established in base metal smelters but collects only vapor-phase, metallic mercury, so it is important to collect other forms in the upstream equipment. Mercury is absorbed by the recirculating solution as follows:

HgCl_2(aq) + Hg_(v) = Hg₂Cl_2(s)

The solid mercurous chloride (Hg₂Cl₂) is separated in settling cones. A portion of the collected solid is bottled in special flasks, which is then shipped to mercury producers. The remaining is chlorinated with Cl₂ gas to regenerate HgCl₂ for return to the scrubber. The absorption tower is a packed column with two levels of mesh pads for mist elimination. This was required to minimize mercury emissions as a result of mist losses and the associated dissolved and suspended mercury compounds. All the mercury handling equipment was installed in a separate building to isolate mercury from the rest of the process equipment. A provision was also made for installation of an electrolytic cell to produce metallic mercury as product while regenerating the chlorine gas.

The next unit operation is SO₂ scrubbing. Selection of SO₂ removal technologies is always difficult because of the large number of available processes. A concentrated-mode, dual-alkali scrubbing process was selected for the following key reasons:

• Sulfuric acid production was considered but dismissed on the basis of the small tonnage of acid, the wide range of sulfur contents in the ore, and the impact this variability would have on plant operation. In particular, the need to keep the SO₂ concentration above its autothermal point would result in the acid plant driving the roaster.
• The high concentrations of SO₂ (2-10% on a dry basis), together with the high collection requirements, necessitated a solution-based scrubber. Direct lime or limestone processes are not suitable for these process conditions.
• A concentrated-mode, dual-alkali process using soda ash and lime is well suited for the process conditions and performance requirements. It offers the benefits of solution scrubbing while using lower cost lime as the key reagent.

Figure 5 shows a flow diagram for the system, which was designed in conjunction with Dr. R. D. Lunt.

Go to Figure 5

SO₂ is scrubbed from the gas in a seven-tray plate tower. Absorption occurs in a recirculating stream of sodium salts:

Na₂SO₃ + SO₂ + H₂O = 2NaHSO₃

A bleed stream is regenerated using a slurry of slaked lime:

Ca(OH) ₂ + 2NaHSO₃ = Na₂SO₃ + CaSO₃• ½H₂O_(s) + 3/2 H₂O

The regenerated sodium sulfite is separated from the precipitated solids and returned to the scrubber.

Two regeneration reactors are included to enable pH adjustment and particle growth. The precipitated calcium sulfite is separated in a thickener and pumped to tailings while thickener overflow is returned to the scrubber. The regeneration system was designed to operate with gravity flow to simplify its operation and maximize reliability. Sodium make-up is provided in the form of a soda ash solution. A novel feature of the system is the absence of a filter; because the gas entering the SO₂ scrubber is water saturated, the system has almost no evaporative capacity. As a result, the system water balance forced a bleed stream to be provided - thickener underflow was used for this. While there are some sodium losses attributable to this, they are minor in comparison to the simplification and improved reliability associated with elimination of the filter.

The final unit operations are for CO and NO_x, both requiring elevated operating temperatures. The main fan was located prior to these steps, which enabled a smaller fan to be installed due to the lower temperature and corresponding lower actual flow rate. Again, a relatively slow speed (1,800 rpm) was selected to maximize reliability. The weak CO and any trace hydrocarbons are destroyed in a recuperative thermal oxidizer. Incoming gas is preheated in a shell and tube heat exchanger using combustion chamber exhaust gases. It then enters the combustion chamber where a burner increases temperature to about 800ºC. The gas leaves through the heat exchanger at a controlled temperature of about 350ºC in preparation for NO_x reduction. An ammonia-based selective catalytic reduction unit (SCR) using a zeolite catalyst was installed. Aqueous ammonia is injected ahead of the catalyst and reacts with the NO_x compounds. SCR is well proven technology and, most importantly for this application, the zeolite catalysts are very resistant to poisoning by heavy metals. This is not the case for the metal oxide catalysts used for CO and hydrocarbon oxidation and is the fundamental reason that a catalytic approach was not considered for the oxidizer.

Figure 6 shows the overall layout for the system. The roaster gases start off at a very high elevation, so the system was laid out to take advantage of this. The gases are moved downward through the particulate cleaning stage, eliminating horizontal duct runs that are prone to build-up. Fans were located at grade where they could be readily serviced. .

Go to Figure 6

System Performance

It is apparent that the Roasting plant started up and achieved design tonnages very quickly.

TABLE 3 presents the performance of the roaster gas system in terms of gas cleaning.

The actual data are results from Barrick’s annual stack sampling. The data in Figure 7 and TABLE 3 show that the roaster gas system has enabled Barrick to meet production targets while providing a level of gas cleaning that exceeds the requirements of a very stringent permit. .

Go to Figure 7

The gas system has been very successful, but there have been some problem areas. During early operation, the SO₂ scrubber experienced calcium scaling of the scrubber trays, which is very unusual for a concentrated mode scrubber. This was aggravated by problems with performance of the main fan; the higher-than-expected pressure drop, together with the shortfall in fan performance, resulted in positive pressure in the wet ESP. The fundamental cause of the scaling was found to be solids carryover from the regeneration system resulting in the build-up of calcium sulfite and calcium sulfate scale in the tower. This scale accumulated on the trays leading to the high pressure drop. Once this was diagnosed, operating conditions for the scrubber and regeneration circuit were modified. Some of the changes included putting fresh seal water on the scrubber recirculation pumps, and scrubber pH adjustments which, over a period of time, enabled the accumulated scale to dissolve. The trays have remained clean ever since.

Another key problem area has been the thermal oxidizer. This unit has suffered from poor temperature control, excessive pressure drop and numerous integrity problems. Many factors have contributed to these, but the major one has been the variability of CO concentrations generated by the roasting process. The pilot work conducted as part of the project indicated that most of the CO generated from the ore would combust in the Roaster freeboard. This has not occurred to the extent expected and, as a result, CO loads to the oxidizer have been higher and more variable than expected. This has been particularly problematic because the oxidizer is designed to operate without external combustion air (i.e., it uses oxygen in the flue gases for combustion). These factors have combined to cause stability problems leading to loss of the flame and tripping of the oxidizer. It has also caused very large temperature swings that have led to both overheating and poor CO destruction. Barrick has addressed much of these by careful blending of incoming ores but is planning extensive modifications to improve operating flexibility.

Conclusion

From an operational perspective, engineering a dependable process was the key objective. Each unit operation was selected to accommodate the full range of process conditions using technology proven in comparable applications. Flexibility was also important to enable the roasters to operate to accommodate Barrick’s mine plan, and not the gas system. Even though the unit operations selected were proven, it was uncertain how the overall system would perform; once the system was commissioned, there was no doubt as to the performance. There were several challenges encountered during the first few months of operation, but all were successfully solved. A few nagging problems remain but are being addressed, which will only further improve the system’s reliability.

The success of this project substantiates the premise that a mineral processing plant can be designed while balancing stringent environmental regulations with a reliable, efficient and cost-effective gas cleaning system.