Unit Type vs. Board Workload – Is there an association?

NEWSLETTER ARTICLE

If you were asked to rank order a crude, coker, hydrotreater/reformer, and cat cracker by difficulty of control, with the hardest unit ranked first, what would your results look like?

Over the past 20 years, Beville has amassed a considerable amount of workload data. Having conducted job samples of over 600 operating positions, we are frequently asked questions such as this. While we routinely perform comparisons for every workload project, with a database that contains detailed information about operator workload, we have the unique opportunity to make objective assessments.

While many factors contribute to overall board workload, we decided to focus on one in particular and how it relates to unit type - control changes. Our initial hypothesis was that the coker would rank highest, and the hydrotreater/reformer lowest.

Using raw data alone (the average of control moves, normalized for control loop count), we discovered our hypothesis was wrong. The rank order is as follows: (1) coker, (2) hydrotreater/reformer, (3) cat cracker, and (4) crude. However, some of these numbers are so close, a statistical analysis was necessary to determine if the differences were significant. In other words, we wanted to ensure that if the data were collected again, would we still see the same rank order?

It turned out that the only statistically significant differences were (1) between the coker and the crude unit and (2) between the hydrotreater/reformer and the crude unit. In other words, the coker has greater control demands than the crude, and the hydrotreater/reformer has greater control demands than the crude. The cat cracker is likely somewhere in between.

While it is true that this data was collected "in the field" and therefore could be the result of many factors (distributed control system, alarm system configuration, advanced control, refinery location), unit type likely contributes to these differences, due to statistical significance of the values.

What implications, if any, does this have? First, the approach of using control loop count to determine workload and board staffing is incorrect. Since our analysis involved normalizing for the number of control loops, we should expect to see no differences between the control changes. Apparently, not all control loops are equal, and hence should not be treated as such.

Secondly, this type of approach can be used to predict operator workload and determine staffing levels for new and/or expanding units. With a database containing such information as control changes, alarms, rounds, maintenance activities, administrative duties, and other control activities, we have the ability to assess both current and future workload levels.

Finally, understanding the components of workload provides the ability to better control it, instead of reacting to over or under-load situations. Knowing where an overloaded operator is spending his time provides both the ability and the opportunity to reallocate and/or redistribute duties.

Copyright 2003 Beville Operator Performance Specialists, Inc., All Rights Reserved