Operator Interface / Optimization

Smart measurement and control in mineral processing

Insight to the measurements used to help the operator open up control and, ultimately, discover optimization opportunities.

By Greg McMillan and Stan Weiner

Stan: Almost everything we touch and use depends in some way on mineral processing. Even renewable energy depends upon mineral processing. Steel is extensively used for wind power and silica for solar power.

Greg: We found in last month's column many of the difficulties and unique problems in mineral processing. Here we continue our discussion with Michael Schaffer, gaining insight as to the measurements used to help the operator and to open up control and, ultimately, optimization opportunities. The measurements are not just smart in the usual sense in terms of compensating for ambient conditions, but have intelligence built in on a custom basis to deal with unusual and diverse moment-by-moment situations faced in mineral processing, where good operation is more of an art than a science. Similarly, controls are made smart by adaptation to rapidly changing operating conditions.

The problems start with the variability of the raw material. How do you know what you are getting from the stockpile?

Michael: We put cameras on each feeder. The cameras are modified to survive the harsh environment. Just adding a camera and traditional vision software still leaves a lot up to the operator. We add intelligence to help establish a better blend to the semi-autogenous (SAG) mill.

Stan: How do you know what the SAG mill is doing?

Michael: You can walk the floor and hear the grinding. Old-time operators could tell from the sound what was going on inside the mill. Some companies put microphones on the mill, but correction was still very much dependent upon operator attention and expertise. We measure vibration directly, putting accelerometers on the mill surface, and then add intelligence to detect the nature and degree of a problem. While mills have bearing pressure measurements that will show a problem, vibration will indicate a problem 35 to 45 seconds ahead of bearing pressure, and provides more information. Since disturbances are so fast, frequent and diverse, elimination of delay and creation of knowledge are critical.

We use weigh meters on conveyors (gravimetric feeders) to infer the pebble load. If these meters are absent, we measure the power draw of conveyors to get an indication of load.

Greg: What about the flotation systems?

Michael: Old-time operators would scoop up and touch the froth and make qualitative decisions on the performance of the system. We create quantitative measurements by adding cameras to look at the surface. We leverage a hybrid of special sensor design and software for proper analysis. The key here is the addition of intelligence to create controlled variables (CVs) based on bubble size and froth color. The flotation cell level and air flow rate that were CVs when the operator was doing manual control are now manipulated variables (MVs) for closed-loop control. Note that we are talking about the level of the interface between the liquid and froth—the froth level is set by overflow from the concentrator. Most common is a float device to measure the interface level. An opportunity is being explored for using guided radar to provide a more sensitive and maintainable measurement.

Relatively small amounts of reagents are added that act as collectors to help minerals attach to bubbles and as an agent to make the surface frothier by increasing surface tension to prevent bubbles from popping. You don't want boiling action; you want consistency. Operators normally would adjust a valve with a volumetric reagent addition of milliliters per ton in mind based on what they would see looking at the surface of the froth. The reagent addition needs to be on a mass basis, more precise and intelligent. Here, the installation of Coriolis mass flowmeters enables us to achieve this goal. The operator adjusts a mass flow setpoint, opening up possibilities for this setpoint to be a MV rather than a CV as smart control strategies are added.

Cork valves (e.g., DART valves), where a rubber cork moves up and down, changing the annular area, and pinch valves, where a roller pinches a rubber sleeve, are used for throttling. The installed characteristic and resolution of these control valves make it difficult for an operator to manually adjust a valve position to set a precise and repeatable flow. The availability of straight-tube, thick-wall Coriolis mass flowmeters creates an incredibly accurate CV, making the valves a MV. Coriolis meters offer the potential of percent solids and, in some models, percent air measurement. However, the meters are not available for extremely large lines and cannot deal with partially filled lines.

Traditionally the process stream density has been measured manually in the field, and the water valve manually set to maintain a more uniform feed concentration. Here, magnetic flowmeters and nuclear density meters are added to the process flows. Magnetic flowmeters also are used on the water lines.

Stan: What about the hydocyclones so critical to flotation system performance?

Michael: Density pushes the coarser material down to the apex at the bottom. Plugging or roping, where coarser material is spitting from the top, can occur. The increase in vibration is tremendous when roping occurs. It is almost like an earthquake. If there were a flag on the top, it would be waving. We always use secondary measurements where possible to increase detection reliability, in this case, vibration measurements at the overflow and underflow. If both indicate roping, a different hydocyclone is switched into service. These vibration measurements can detect roping within 5 seconds compared to 10 to 15 minutes by other means.

Unfortunately, flowmeters are not used in the recirculation lines because of three-phase flow and partially filled lines. Valve positions are manually set. An inferential flow measurement based on installed flow characteristic can make this setting more intelligent.

Greg: Given the smart measurements, how do you determine what to control?

Michael: We determine the degrees of freedom (DOF) established by the measurements. We can set the use of outside versus inside feeders from the stockpile based on what the cameras are telling us. We can measure and change water flow at the inlet to the mill, which provides a faster correction than water addition to the outlet. We can measure and change air flow into the flotation cell. We can use gravimetric feeders or conveyor power draw to change the circulating pebble load. We can measure reagent, water and air flow, and use ratio control on these flows to achieve desired concentrations and, subsequently, more optimum operating conditions. We can measure the flotation level and froth characteristics to better set these ratios and determine how much fresh feed can be added.

Stan: What are some examples of smart controls?

Michael: We have developed a portage flotation stabilization (PFS) system, where we use fuzzy model predictive control (FMPC). We adapt the model dead time, time constant and process gain. The adaptation is based on fuzzy logic to deal with non-Newtonian fluids and a froth that can go from like what you see on a lager versus Guinness stout. We have a pump box controller (PBC) to maximize the absorption of variability by the pump box. Previously, the pump box level control was very tight, changing the feed flow to the cyclone all over the place. The cyclone is particularly sensitive to flow, besides pressure and density. As a result, we achieve a 60 to 80% reduction in variability off the top of the cyclone.

Greg: We have seen how smart measurements and smart controls move what was a CV for manual control to a MV for closed-loop control, elevating the role of the control system and operator to be more one of supervision and optimization. Next month, we'll see how we make sure the maximum benefits are achievable and sustainable. Meanwhile, for some comic relief, see the "Top 10 word plays for engineers in mineral processing" below.

Top ten word plays for engineers in mineral processing

(10) I told a chemical engineering joke. There was no reaction.
(9) Why do you keep staring at the pH! I got my ion it.
(8) Do you have any sodium hypobromite? NaBrO.
(7) Is it full of beryllium gold and titanium? Be-Au-Ti-ful.
(6) That equipment just spit sodium chloride at me. That's a salt.
(5) The equipment in the model blew up. Oxidants happen.
(4) Get Real. Sure, I will have a piece of Pi.
(3) Be rational. Sure, I will use imaginary numbers.
(2) Why don't you trust atoms? They make up everything.
(1) What's the matter with you meeting your complete opposite? He could be anti-matter.