Board Inspection Frequency: Can today's board be left unattended?

NEWSLETTER ARTICLE

In the "old days," it was common for an operating position to have both "inside" and "outside" responsibilities. As such, one could often find control rooms empty, with no one watching the board. All of the operators might be out on the unit performing other tasks. It is common with today's distributed control systems to hear operators and/or supervisors argue that the board can't be out of the board operator's sight. Given past operating practices, this begs the questions, if not required by the process (i.e., an alarm or an information request from plant personnel), how long would an operator take before inspecting their board? What is the minimum inspection frequency for an operator to sample the process under their control?

A rare opportunity presented itself recently when Beville was observing an operator who (1) had a very stable process (therefore, no alarms forcing an inspection of the DCS), (2) little activity in the plant (therefore, few information exchanges with the field), and (3) an operating style that made it easy to determine when the board was being examined (he otherwise worked with his back to the console). The operator was observed over a four-hour period. DCS inspections due to alarms and communications were removed. The result is a series of times the board went unmonitored. A histogram of those inspections is shown below.

As would be expected, the operator typically checked the board every five minutes, never going longer than 20 minutes between inspections. It is interesting to note that the time since the last inspection exhibits an exponential decay. This indicates that the 20-minute limit in time away from the monitor is not a random occurrence, but part of the operator's internal model for the need to check on the process.

So what does this mean? The first and most obvious conclusion is that the operator observed had an uneventful (boring) sample period. Second, DCS monitoring of a process can require very low levels of workload; monitoring a process is not inherently a high workload task. Third, it is possible to quantify aspects of operator behavior, which could then be compared to other jobs that require sustained monitoring (e.g., air traffic controllers). Fourth and finally, a stable process need not have the operator "glued" to the screen to maintain control of the process.

Copyright 2000 Beville Operator Performance Specialists, Inc., All Rights Reserved