Process Control Loop Performance Guidance

Variability in a process can impact unplanned downtime frequency, product quality, energy usage and more. I received some questions about measures of control performance from a reliability specialist in a refinery. I thought that these questions and answers I received were too valuable to remain buried in email. The questions were:

  1. How many controllers or what % of the controllers to be operated in manual considered as accepted.
  2. Controller Output – What is the guideline or acceptance range of controller output (for example: controller output from 40% to 70% is considered as healthy)
  3. What is the best practice or accepted variance of PV (Like + or – 10%)

Emerson's Andrew Waite


I shared the email with members in our Control Performance Consulting Services team for their thoughts. Emerson’s Andrew Waite responded:

1. Number of controllers in manual mode

  • Generally we suggest that the number of controllers in Manual mode, (or the Incorrect mode), should be less than 10%
  • You can refine this criterion by identifying when the controller is in the incorrect mode rather than just in manual mode
    • This would then identify a loop that is in local automatic when it should be in cascade, for example
    • This would also not flag a “manual loader” as being problematic
    • If your system allows, consider defining a percent of time that the controller is in the incorrect mode. If a controller is in the incorrect mode for less than 5% of the time, for example, it would not be counted

2. Controller output in %

  • 40% to 70% output sounds reasonable. You may want to consider a slightly wider range – 30% to 80%

3. Process Variable (PV) variation

  • We generally monitor normalized process variability: 2-sigma as a percent of mean. We calculate 2 times the standard deviation or “2-sigma” and divide by the mean.

2 * (Standard Deviation) / mean * 100

  • The 2-sigma can be calculated from the SP-PV, to better accommodate cascade loops, and frequent setpoint changes

    • The process variability should be:

      • Less than 0.5 percent of mean – critical loops
      • Less than 1 percent of mean for most other loops
  • You may want to monitor temperature variability as 2-sigma in engineering units – °F or °C, not normalized – sometimes this is more applicable
  • Variability monitoring of level loops can be misleading. Many tanks are intended to absorb process swings so level deviation is not necessarily a bad thing
    • Sustained oscillation of a tank level must be avoided – this is very detrimental to the process.
    • Consider monitoring the normalized variability of the controller output. In this case, however, the acceptable level of controller output variability may be different than the acceptable level of normalized process variability.

The Emerson Educational Services team has a number of great courses that apply to tuning and variability management. These include:

You can connect and interact with other control performance experts in the Improve & Modernize group in the Emerson Exchange 365 community.

Leave a Reply