Wouldn’t it be great if a production process kept the same performance over time? Unfortunately, the dynamics change as the assets wear and production rates vary. Production performance consultants work to adjust the tuning and control strategies to address these changes and maintain optimum performance.
One of Emerson’s Principle Consultants, Lou Heavner, was recently asked a series of questions about process optimization in the mining industry. Here are Lou’s answers to the questions asked.
The most commonly used software solutions for process optimization are probably Expert Systems or Fuzzy Logic developed by subject matter experts. However, a complementary technology is seeing some use with a significant growth opportunity. Model-Predictive Control (MPC) is usually accompanied by an optimizer (linear programming—LP or quadratic programming—QP). Whereas expert systems can tell you the optimum operating conditions, they are steady-state solutions and don’t usually have a sense of process dynamics. MPC cascades directly to existing feedback control loops and can drive the process to the optimum conditions while honoring all measurable constraints. MPC applied as a constraint optimization solution can handle shifting constraints more effectively than a prescriptive solution like expert systems.
Presumably, they are all intended to optimize the process. The difference is in understanding steady-state optimization vs. real-time optimization. The sole purpose of MPC is to push the process to some control objective like maximizing throughput up to process constraint limits in real-time. If process operations or metallurgists can define optimization objectives like maximizing throughput, minimizing unit energy consumption, or maximizing instantaneous profit margin, the controller will drive the process to optimize the objective until it is unable to move further due to constraints limits. The constraints that are limiting may change for unknown (unmeasured) reasons.
For example, in a mining grind circuit, the ore feed rate limitation may be bearing pressure one day and mill amps on another based on small changes in ore properties. On yet another day, it may be particle size or downstream capacity, perhaps level in a thickener that limits feed rate. Manipulated Variables (MVs) such as ore feed rate, mill speed, water flow, etc., are process inputs, and constraint or control variables (CVs) such as bearing pressure, density, particle size, mill amps, etc. are process outputs (aka observations or responses). Feed-forward variables (DVs) are also process inputs that can improve control by informing the controller of measured process disturbances that will affect process outputs.
Process inputs are usually material or energy flows, and process outputs are the rest of the process measurements, including levels, temperatures, pressures, compositions, and other stream/tank properties. In theory, an MPC controller can drive the process to as many simultaneous constraints as there are MVs. One noticeable result with MPC is a reduction in overall process variability. As a multivariable controller, it tends to avoid aggressive moves in order to control more robustly. Rather than having interacting loops fight each other, it coordinates MV moves purposefully.
Model-based future prediction tends to reduce oscillatory behavior common with single loop PID control, especially on loops that have difficult dynamics like deadtime and inverse response. Another important result of MPC control is that it tends to control the process more consistently than multiple single loop controllers. This consistency leads to more uniformity between operating shifts. Furthermore, unlike operators working in a multitasking environment, the MPC controller executes continuously without distraction resulting in the process operating closer to optimum conditions than even the best operators will achieve. This continuous performance helps the operator control the overall process better and perform higher-level tasks.
MPC incorporates a dynamic model that moves the process smoothly and efficiently toward the current optimum in the face of changing constraints. It does this even if the process has deadtime as the dominant process dynamic like a carbon-in-leach (CIL) circuit or a high degree of interaction as in a grind circuit. MPC is a software product that may be independent and installed on a server, such as Aspen DMC, or built-in and embedded in a DCS control system like DeltaV PredictPro. The MPC products are usually licensed based on the size (frequently measured as MVs), and licensing fee schedules vary from vendor to vendor.
MPC is cascaded to regulatory control loops as a supervisory function. The controller configuration includes defining the MVs, CVs, and DVs. With an LP or some type of optimizer, some variables may be defined as maximizing or minimizing variables. Step testing (similar to loop tuning procedures) is used to identify the process response model between process inputs and process outputs. The controller then attempts to use that information to minimize the error between operating targets and optimum limits over a future interval with a minimum amount of well-coordinated moves.
This technology is applicable virtually anywhere that requires improved control. It has been used in smelter feed hopper level control, grinding circuits, CIL circuits, thickeners, etc. It is widely used in many process industries, but the first and highest adoption has been in the petroleum refining industry. The best candidates are continuous processes with a high degree of multivariable interactions, difficult dynamics like deadtime and inverse response, and processes that benefit from constraint optimization. It has been applied to some “fast” processes like combustion. Typically, it would deliver master control for the whole kiln or roaster and leave the basic fuel and air controls in the combustion system to faster regulatory control strategies.
As the mining industry becomes more aware of MPC technology, the adoption rates should increase. I am not aware of any market projections for the adoption of Advanced Control in general or MPC in particular for the mining industry. Some may view it as a competing technology to traditional Expert Systems, but it should be viewed as complementary.
Visit the Production Performance Consulting and DeltaV Advanced Control sections on Emerson.com for more on the technologies and services to drive production performance improvements.
The post Mining Process Optimization appeared first on the Emerson Automation Experts blog.