Most system designers would argue that the more equipment a person has, the higher the person's workload should be. Intuitively, this line of logic would seem to be axiomatic; how could it possibly be wrong? Control system designers have long used the number of control loops as a metric of system complexity and operator workload, and, for the most part, the strategy has been seemingly correct. Or, up until this point, there hasn't been data to contradict the assertion. Beville Engineering used data gathered during several recent pipe line studies to test whether or not the conventional wisdom that increasing amounts of equipment cause increasing operator workload is, in fact, correct.
One simple way to test the equipment amount/workload theory is through simple correlations. If the amount of equipment actually changes operator behavior and increases operator workload, then the effect should show up when correlations are run between the two. Increasing the amount of equipment should cause similar increases in operator behavior and workload. Beville Engineering recently performed such a study, comparing the amount of equipment under the operator's span of control versus observed measures of workload. Data for the study were taken from several different pipe line companies. Measures of operator workload collected through job sampling, such as control moves/hour, alarms actuated/hour, instrument monitoring time, communications, etc., were compared to corresponding amounts of equipment (miles of pipe line, number of MOV's, number of pumps, number of tanks, etc.) under the operator's span of control using single regression analysis. This analysis was conducted on several aspects of the data (the entire set of samples, the AM samples, and the PM samples) to determine the best possible linear relationship. The correlation coefficient (r) calculated with this analysis described the slope of the line in positive and negative terms and the strength of the linear relationship.
Single regression analysis conducted on the entire set of samples (AM and PM samples combined) revealed only 3 slightly significant correlations(-0.80 > r > +0.80) out of the 120 possible correlations. When the same analysis was conducted on the AM and PM samples separately, a total of 47 significant correlations (-.086 > r > +0.86) were found out of 240 possible correlations. The reason for the increase in significant correlations is the difference in job demands (AM samples had far greater job demands than PM samples) between the morning and evening samples.
The difference in the activity level of morning and evening samples was revealed when the difference of the AM and PM correlations was calculated. Since the workload measures were being correlated against a constant (equipment), a difference greater than one would reveal a completely opposite linear relationship. Over 34% of the data sets had a difference greater than one (Dr > 1), particularly in the number of communications, percent instrument monitoring, percent direct time, number of display changes, and Mean Time Between Tasks. This shows that the activities performed in the morning are drastically different than in the evening.
Several of the resulting correlations appear to be valid. The number of communications increases with the miles of pipeline (PM), as did the number of alarms (PM). Also, the number of control actions increases with the number of tanks (PM). However, three of the correlations are contrary to conventional wisdom or contradict each other. For example, the number of alarms actuated decreases with more transmitters (PM) and MOV's (AM), and the number of control actions decreases with an increase in the number of pumping stations (PM). The percent instrument monitoring time actually has a positive correlation with booster pumps (PM) and a negative correlation with main pumps (AM). These peculiar correlations are most likely false positives; since so many correlations were run, a certain number will be false positives based on the criterion cutoff used.
Of course, correlations do not necessarily mean causality; rather, high correlations simply indicate two variables that co-vary linearly. Some high correlations are meaningful and suggest causality does exist. Other correlations are chance and have no real significance. For example, there is a well known high correlation between ice cream sales and the murder rate in New York City, yet common sense would reason that ice cream in no way causes people to murder other people. Actual experimentation is the only method to conclusively determine causality.
In summary, if relationships do exist between amounts of equipment and measures of operator workload, the relationships are weak and not altogether well defined. A few of the strong correlations warrant further investigation, as the number of tanks and control actions is one of the few that could be argued to be valid. Most others, such as the negative correlation between MOV's and alarm actuations seem to be merely chance correlations. Control system designers should be aware that the actual amount of equipment alone is a weak indicator of system demands and operator workload. When making workload estimations based on the amount of equipment, one should also take into account the sophistication and quality of the control systems for a more accurate estimation.
Copyright © 1995, Beville Engineering, Inc. , All Rights Reserved
RELATED EXTERNAL MEDIA
|Console Operator Staffing Workload And Automation - ISA Automation Week 2012||ISA.org|
|How Good is Your Operator's Mental Model?||Mynah|
|Modularizing Emergency Procedures for Increased Ease of Use and Updating||OperatorPerformance.org|
|Operator Performance Takes Center Stage||Chemical Processing|
|Operator Training Gains Ground||Chemical Processing|
|Operators Get More Help||Chemical Processing|
|Simple, Strong and Easy-to-Use||Control Global|
|Teach Operators to Make the Right Decisions Under Fire||SustainablePlant.com|
|The Keys to Operator Performance||ControlDesign.com|
This year's Fall meeting for the Center for Operator Performance will be October 7-10 in Houston, TX. For more information, please contact Lisa Via. Guests are always welcome!
Our summer newsletter is now available. Click here!
Take our short survey on operator span of control. Click here (new window)
David Strobhar's book, "Human Factors in Process Plant Operation," is now available in both hardcover and Kindle e-book.
Copyright © 1996-2019 Beville Engineering, Inc. All rights reserved. (937)434-1093. Beville@Beville.com