Process control challenges and opportunities in mineral processing

Variations in feedstock hardness, grade and rock size lead to fast and frequent unmeasured disturbances, calling for the best of technologies and techniques.

By Greg McMillan and Stan Weiner, PE

Greg: Mining may well have been the second of humankind's earliest endeavors – granted that agriculture was the first. The two industries together were the primary industries of early civilization. Milestones in human history are marked by mankind's ability to mine new minerals, and use them to advance technology.

Stan: Today, mining is prevalent in more than a 100 countries and directly employs 3.7 million people and even more indirectly. The global mining industry contributes about 45% to the world Gross Domestic Product (GDP) on a direct and indirect basis, from iron ore and copper that serve as building blocks of modern cities, to coal and uranium that keep the world's power plants running, to potash, an essential ingredient in fertilizer manufacturing for the agricultural industry that feeds the world's population. Everything around us is either mined or grown in some way. For the source of this data and more information see miningfacts.org, "How many jobs depend upon the mining industry," and the July 2012 Mining Weekly article, "Global mining drives 45%-plus of the world GDP."

Greg: We are long overdue for a close look at this industry, so this is the first of a series of columns to catch up. I have asked Michael Schaffer, president of Portage Technologies, to provide us a solid understanding of the characteristics and uniqueness of this industry and the advances that have been made. Portage has operations in Canada, Mexico, Chile and South Africa.

Stan: What makes mineral processing difficult to start with?

Michael: The raw material (ore) hardness, grade and rock size vary greatly. There is no homogenization of the feed and no direct measurement of raw material characteristics. This results in fast and frequent unmeasured disturbances. Variability is minute by minute. There can be a huge change in 30 seconds and the operators have to stay on their toes to keep on top of them. Some applications provide the extra challenge of being in difficult to reach places. An issue with a cyclone can take even the best operator 10 to 15 minutes to evaluate and respond. The operator is constantly playing catch up. What they can see by way of measurements is sketchy at best. It is like they had on a pair of glasses with one lens missing and the other lens with the wrong prescription.

Greg: In most of the plants where I worked there were well-mixed volumes that attenuated the changes in the process. Feed tanks blended out variations in raw materials that were minimal to begin with, and storage tanks blended out variations in the product. The attenuation was so significant that in some cases it was difficult to justify better automation to reduce variability. For the most part, the true step load disturbance seen in the control literature did not exist when control loops replaced on-off actions. The exception was the plants that used the extruders and sheet lines for the final product (e.g., clear plastic interlayer for safety glass). Fast disturbances seen in chemical processes can be slowed down by better automation system design, as outlined in the 10/18/2013 Control Talk Blog, "Disturbance Dynamics Recommendations Tips".

Stan: What are some of the challenges within the process?

Michael: It doesn't matter how small you grind a particle. Each is unique. You almost always have two phases (solids and water) and in many cases there is a third phase with air creating foam and froth. All fluids are non-Newtonian. The mineral content can double in a matter of minutes, changing the flow characteristics dramatically. The processes are complex, with many circulating loads. The operator not only has to think of what is coming, but also what is coming back. Grinding typically has two recirculating streams and flotation can have as many as 10. There can be a building up of inventory within the circuit that forces a cutback in production. A greater recirculating load ultimately means less fresh feed as the capacity is consumed by the circulating load. The problem escalates as the material keeps coming back at you. The best operators recognize when the inventory is building and find a way to pull it out.

Greg: Recirculating or recycle streams presented a challenge in even less difficult processes due to a "snowballing effect" where an imbalance in material and contaminants build up. The solution almost always requires some added intelligence, with optimization often offering a greater opportunity than seen in once-through processes. How do you stop a problem from the start?

Michael: The first step is ensuring that you are setting realistic targets. As grade and hardness change, your expectations need to follow. If the operators are asked to deliver a recovery or concentrate grade that is not aligned with the ore potential, they can chase their tails (literally – the waste from the process is known as tailings), increasing the circulating load in an attempt to hit metallurgical targets that are not realistic. Beyond setting reasonable goals for operations, the operators themselves are tasked with evaluating the efficiency in the plant and pressing for results. For example, in grinding, an operator has a number of parameters that can be evaluated in real time and will provide an indication of how the mill is performing. If we start from the basics, ore hardness is quoted as the amount of power required to grind a mass of ore from one size to another…this is expressed in kilowatts per ton. Ultimately, this depends on a combination of the feed size of the ore, the hardness of the ore and the product size of the ore. The operator monitors the load in the mill and the power and evaluates how they are performing together. They should trend the same.

If you have more or less load in the mill, then you should be using more or less power (if the hardness of the ore is the same). If the trends diverge, then something has changed. Now the operator must determine what is happening and take a corrective action. The best way to illustrate this principal is with a classic overload in a SAG mill. In this case, the load in the mill hits a point where the center of gravity in the mill shifts. The momentum of the material in the mill drives the rotation and the motor no longer needs to work to spin the mill. This is referred to as "freewheeling." The greater issue, though, is that grinding stops, as the material in the mill is no longer tumbling and is plastered to the walls of the drum. In this case, the load in the mill increases rapidly and the power drops precipitously. If a correction is not made immediately the mill can fill to the point where it must be "dug out," leaving the plant down for hours if not days.

Another example pertains to the feed to the mill itself. The mill draws on a stockpile, often with two to four feeders. The stockpile contains ore from different parts of the ore body. It is a natural classifier with the fines reporting to the center and the coarser material flowing to the outside. This phenomenon opens up a control degree of freedom for the operators but also complicates matters as the ore, as variable as it already is in a particular location, is now coming from various parts of the ore body at once...yet another unmeasured disturbance.

When there is a problem, the natural response of the operator is to cut back on the feed. You can imagine the amount of product loss due to continual problems, the amount of off-spec product due to poor particle size control, and the amount of energy used for grinding and recirculation of recycled material rather than for fresh feed.

Stan: What are the unit operations?

Michael: A company may start out with a fully autogenous mill, hoping that the rocks can grind themselves during the spinning. Most find they need a semi-autogenous (SAG) mill, where steel is added to provide more grinding than what you get from just rock on rock. The grinding is achieved by impact, creating coarser particles, pinching or nipping giving smaller particles, and abrasion generating fines. Grinding is often finished with a ball mill, which differs from a SAG mill, as it relies on the steel alone for grinding. A SAG mill may have 10% steel where a ball mill will typically have 30%. Classification is normally carried out in hydro cyclones with the product feeding either a flotation or leach circuit. The product is then treated with thickeners, filters and in some cases, dryers. At certain plants, there may be further processing through solvent extraction and electrowinning (SX-EW).

Greg: What are the potential benefits?

Michael: We have typically gained a 5 to 10% increase in throughput, 1 to 4% increase in recovery, and a 4 to 9% reduction in energy use. Energy use is huge. It is not uncommon for a milling circuit to draw more than 30 MW.

Stan: How are these benefits achieved?

Michael: You can't control what you can't measure, and you can't optimize what you can't control. The first step of adding the measurements is needed even for manual control. How can the operator correct what is not seen? Stopping here is leaving much of the opportunity on the table because the operator cannot keep up with the disturbances and is challenged by recirculation, and knowledge is diminishing with retirement. By nature, human action is late and not repeatable, even under much easier circumstances than these.

We add the measurements, controls and optimizing strategies using technology developed and intelligence gained from worldwide applications. However, the problems and solutions can be incredibly different for any given site, so a site-based best operating practice is established. Success requires an extensive and focused onsite study with communication and flexible technology. The plant's feed, equipment, special operational practices and culture must be understood, and changes made as needed to both. We go onsite and develop a fit for purpose (FFP) definition of best practices. We spend 24 hours a day, seven days a week working with operators to see how operations change, based on problems and staff. We also seek to get the knowledge and synergy from the metallurgist and controls engineer. This approach is essential because what works in North America is different from what works in South America, which may be different from what will be effective in Africa. We seek to make the best practices more frequent, repeatable and timely through measurement and control. We move what were perceived as controlled variables (e.g., flotation cell level and air flow rate) for manual operation to be manipulated variables for process control of new controlled variables (e.g. froth velocity, bubble size and color of the froth).

One thing that we are always aware of is the importance of cultural change with the introduction of new technology. It is critical that the team on the floor understand the new approach and take ownership of the solution for it to be sustainable. They must be part of the solution, not simply along for the ride.

The first objective when we are on-site is stabilization. We eliminate insufficient action, overreaction and delays. We can then push the limits, operating closer to the constraints. Optimization is the icing on the cake.

Greg: Since I found possible puns are extensive, we decided they would be funnier and less distracting as a list. See the online version for the "Top 10 puns that came to mind writing this column."

"Top 10 puns that came to mind writing this column"

(10) We are going to dig deep to get the best material.
(9) We don't want to just scratch the surface.
(8) We need to get down and dirty.
(7) We are after the inside scoop.
(6) We want rock solid information.
(5) We want to get to the top of the heap.
(4) There is a particular need for data mining of particle mining data.
(3) Figuring out what to control can be a real grind.
(2) We are going around in circles from recirculation.
(1) Let's rock ‘n roll with technology.