Basic steps, such as measuring the percentage of the plant running in manual mode, are a good place to start. A typical process plant runs with between 20 and 30 percent of controls in manual mode. Consider that a typical control loop costs $10,000 or more. A typical oil refinery may have 3,000 control loops, with 600 to 900 in manual at any given point, representing a lost investment of over $6 million. And that doesn't even count the annual process losses, which are at least as high.
Additional simple measures, such as the percentage of control valves at a limit, and overall variability, can give you a better idea of the opportunity for savings at your plant. Start by taking a random sample of 100 control loops. In one hour sitting with the operator, you can confirm whether your plant has a significant opportunity to improve.
The next step is to measure and monitor control performance regularly. Software tools can monitor various measures, provide reporting and notification of problems. This step is critical to recognizing where problems exist in your plant on any given day.
Automated monitoring also allows for more sophisticated measures and diagnostics. Modern software tools can find failed instruments, broken control valves, poorly tuned controllers and process problems. These analytics are needed for the next step: resolving the root cause of the problem.
If 30 percent of controllers are operating in manual mode, the solution is not as simple as putting everything into automatic mode. Chances are, the operator put these controllers in manual for a reason. Any good control performance project can use data from your process historian to help pinpoint and resolve the true root cause of the issue.
Some problems, such as valve stiction, have a unique signature pattern, as shown in Figure 1. The square wave pattern on the process measurement and sawtooth movement of the control output are a sure sign of valve stiction. The ladder pattern in the PV/CO plot shown in Figure 2 further confirms valve stiction.
Other problems, such as oscillations and interactions between various parts of the process, need more sophisticated analysis. Tools such as Fourier Transforms and the resulting power spectrum chart, shown in Figure 3, can find the cause of oscillation problems.
Process Interaction analysis, as shown in Figure 4, can also be used to pinpoint the cause of a problem that may be far upstream in the plant.
With the problems identified and prioritized, your focus can now turn to managing the required repairs, tuning changes and process adjustments. This work should be managed and organized like any project: with specific plans, assignment of responsibilities and expected completion dates. Some repairs may need to wait until planned shutdowns, but others can be completed on the fly.
Finally, it is important to track the effect of these improvements. A simple before/after trend can be very useful to show the results. Whenever possible, you should also identify the economic impact of your work, and share it with your management team.
If you have enough staff at your plant, each of these steps can be accomplished in-house. Experienced software and service suppliers can also provide all or part of these activities as a service, starting with site evaluation and progressing all the way through capture of benefits and sustaining the results.
What kind of results can you expect? That depends, of course, on your starting situation. But it is common to see reductions in variability that lead to energy savings, production increases and quality improvement. Because automation system improvements have a direct effect on process results, the return on investment is typically measured in months, not years.
This article was adapted from content provided by George Buckbee, P.E., ExperTune
Liked this article? Download the entire playbook here.
