Вы находитесь на странице: 1из 6

Controller Workload: What are they doing and why are they doing so much of it?

By David A. Strobhar Beville Engineering, Inc. 201 West Franklin St., Suite D Dayton, Ohio 45459 Abstract The ability to measure a phenomena is critical to being able to control and/or modify it. The workload imposed on pipeline controllers has until recently been measured only by subjective assessments from pipeline personnel. However, Beville Engineering has undertaken recent studies to quantify controller workload. The studies relied on techniques developed for measuring the workload of oil refinery operators. Use of job sampling for steady state workload analysis will be explained. The job sampling methodology will be described along with the type of data obtained from the pipeline studies. The relationship between time spent on tasks and the number of tasks that need to be performed will be discussed. The impact of type of pipeline and SCADA configuration on steady state workload will be highlighted. Methods to lower controller workload will be discussed relative to the objective data. Changing anything requires some understanding of that which you are attempting to change. This holds true for pipeline controllers, where an understanding of the position is needed in order to change the jobs to make them more efficient, higher performing, and/or different in nature (e.g., commodity based versus geography based). Until recently, most of the understanding came from people who had worked the job previously. However, unless one person had worked all the positions, comparisons across positions became difficult. Subjective assessments of the job were utilized, but they usually lacked a common frame of reference for what was difficult or easy. Objective measures of job characteristics are needed in order to assess how a job can be altered and/or how it should be changed in light of new technology. Beville Engineering has been developing such measures and applying them to oil refinery personnel for the past ten years. Recently, the same techniques have been applied to pipeline controllers with considerable success. The technique applied is referred to as job sampling. It is a variation of traditional time and motion studies. In time-and-motion study, the objective is to determine a standard time for a repeated task. However, because of the continuous nature of process operations, repeated activities occur only over a long period. The objective of job sampling is to capture a snapshot of the controllers shift when the controller is most likely

to be heavily loaded. Taking a long sample of the controllers activity at the beginning and ending of the shifts captures those repeated activities the controller is likely to experience day-to-day. Questions typically occur on the ability to use only a few samples of a controllers activity as representative of the entire job, Isnt every day different? Certainly everyday is slightly different, but not as much as might be suspected. While the object of the activity (a specific pipeline) may vary from day-to-day, the types of activities tend to repeat or have a constancy (starting/stopping a line). Sample/Re-sample tests on the reliability of the technique show an unaccounted variance of less than 6%. Two four-hour samples were collected for each pipeline position. The sample start times were adjusted to each particular location to capture the first four hours of the controllers day shift. The second sample was taken in the late afternoon and/or early evening in order to catch a quieter, yet not dead, time period. The two samples provide indication of peak workload, the difference from peak to normal, and an average workload. All job related activities are recorded during the sample period. The start and stop time for each task is logged, and notations made as to the task itself. The task data is later categorized. In the case of the pipeline controllers, the data collected were categorized into one of three areas: (1) communication interactions, including with whom they interacted, (2) SCADA interactions, including instrument inspections, control moves, alarms, CRT and display usage, and (3) administrative activities, including logging and line balances. One measure obtained from job samples is the actual amount of time the controller spends on job related tasks, which Beville refers to as direct time. A higher direct time indicates a greater portion of time being spent on job related activities. Direct time is represented as a percentage, calculated by dividing recorded task times (in minutes) by the sample duration (in minutes). Figure 1 shows the average breakdown of percentage direct time for both controllers and refinery board operators. The breakdown of the time is different due to the different nature of the two jobs.
Controller Task Communication SCADA Interaction Admin/Logging % Time 16.8 21.4 23.2 Refinery Board Operator Task % Time Administrative 17.4 Operational 21.3 Maintenance 1.9 Inspections 9.9 Laboratory 0.8 51.3 Total

Total

61.4

Figure 1. Controller & Refinery Board Operator Direct Time Averages

Despite the difference in the task categories, some general comparisons can be made between pipeline controllers and refinery board operators. The controllers are typically spending more time, about 10%, on job related tasks that are board operators. This difference in utilization may stem from board operator jobs matching processing units, and therefore not easily dividable for balancing workload. Pipeline controllers also spend more time on logging, which is one sub-task of the board operators administrative task. While direct time is important, it captures only half the story. In addition to how much time is spent on job related tasks, the number of tasks that must be performed influences workload. Spending 40 minutes out of an hour on four tasks will be perceived to be far less busy than spending 40 minutes out of an hour on ten tasks. Combining the direct time loading with the number of tasks provides an indication of busyness. Busyness is simply the combination of two measures of steady state loading, (1) direct time and (2) the number of tasks, to form a third derived measure of workload, mean time between task (MTBT). Mean time between task is calculated as follows

[ SampleLength ( Minutes ) DirectTime ( Minutes ) ]


NumberofTasks A chart showing the relationship between the direct time and number of tasks is shown in Figure 2. The lines on the chart are equivalency curves, where different levels of direct time and number of tasks have the same degree of busyness, or mean time between tasks. Movement up or right on the chart indicates increasing busyness, decreasing mean time between tasks. Three equivalency curves are shown on the chart. A MTBT <0.5 (upper right of chart) is characteristic of upset workload and indicates a higher loaded job. A MTBT >2.0 (lower left of chart) is characteristic of an under-loaded job. The controller workload samples shown in the figure represents different types of controller workload: Position A (Crude Pipeline) Large swing in workload is seen between the AM and PM samples. The controller was handling predominately crude pipelines, which are characterized by a significant increase in workload for the early morning period. The workload for the AM sample is less than .5 minutes mean time between tasks, which is more typical of an upset situation. This is in contrast to the PM sample, which was in the under-loaded range. Position B (LPG Pipeline) The busyness for this controller stayed constant over the two sample periods, with a shift in number of tasks to direct time from the AM to PM period. This type of response is typical for stable positions such as product pipeline control.

160
.5 min MTBT A AM

140 120 100


Nu mbe r o f Ta s k s

1 min MTBT D AM

2 min MTBT

80 60
C AM A AM

D PM B AM B PM

40 20 0 30%
C PM

40%

50%

60%
D i re c t T i m e (%)

70%

80%

90%

100%

Pipeline Controllers

Figure 2 - Pipeline Controllers Busyness

Position C (Olefin Pipeline) This position is consistently under loaded. While the morning got busier, it was still less than two minutes mean time between task. Position D (Crude/Product Pipeline) This position shows the value of examining busyness. While the direct time was almost equal for the two samples, the number of tasks increased significantly for the AM period. Short, multiple tasks in the morning (e.g., communication contacts) increase the operators busyness without significantly increasing the controllers direct time load. Poor alarm system management has been found in refineries to significantly increase busyness by requiring the operator to respond to a large number of alarms, in response to which they take few actions.

As part of capturing overall controller workload, the job samples captured SCADA interaction. Each time a display was accessed, information on what was done (inspecting, control change, etc.) was recorded along with which CRT was utilized. The result is that not only is information on operator characteristics available, but the SCADA system characteristics are also available. Figure 3 represents the total time each display was on a CRT. It is interesting to note how a few of the displays were up most of the time and how only a few of the available CRTs were used to access most of the displays. Eight of the 21 displays accessed were in view the vast majority (over 95%) of the time. Job sample results show that the controllers changed displays an average of 29.4 times per hour. Of the display changes, 90.5% occurred on two CRTs. This should not be

500 450 AMOUNT OF TIME ON EACH CRT (MINUTES) 400 350 300 250 200 150 100 50 0
S TA TI C O RU N S DE 14 TA S TI YS O N TE 13 M TA N S K TA S TI O N S 12 TA TI O N S 11 TA TI O N 10 S TA TI O N ST 9 AT IO N S 8 TA TI O N S 7 TA TI O N S M 6 TA ET TI ER O N S 5 DI SP LA S Y TA TI O N S 4 TA TI O N S 3 LI TA NE TI O #4 N B 2 AL AN C S E TA LI N TI E O N #3 1 LI PR NE O #1 FI LE B AL AN CE LI LI N NE E #1 #2 PR O FI LE

CRT DISPLAY CRT 1 CRT 2 CRT 3 CRT 4 CRT 5

Figure 3 - Control Display Usage

confused with how often each CRT was looked at. Since eye movements were not recorded, it is impossible to say precisely which CRTs were looked at. However, despite a large number of screens, the controllers typically only actively interacted with two. This is consistent with similar findings for refinery board operators. While the other screens may be providing value, most the control actions will be on two. The job sample data provided insights and information on ways to alter controller jobs to enhance performance. Understanding the current workload allows a determination to be made as to whether the controller can handle additional tasks or systems without any associated job changes. If the position is under loaded, additional tasks or responsibilities can be added without other changes to the system. If the job is overloaded, or nearly overloaded, then changes can be made to the job to reduce the workload, either to improve performance or enable the controller to take on additional tasks/responsibilities. The sample data can indicate opportunities to alter controller workload. If one task is taking up a disproportionate amount of the controllers time, as indicated by the direct time breakdown (such as logging), then it is a candidate for automation or possible system re-design. Busyness indicates opportunities to reduce operator workload. Reducing frequent, short duration tasks, such as responding to alarms or communication with the field, can dramatically reduce the controllers level of busyness. Large swings in busyness,

similar to that for the crude pipeline controller, are candidates for either differential staffing (provide AM help) or dynamic task reallocation (where another controller could help on some of the tasks). If a large amount of display paging is occurring, display redesign can reduce the paging demands. Workstation design should reflect the reliance on the number of CRTs used to make most of the control changes. The data should not be interpreted as saying that only two CRTs should be provided, but it does say eight CRTs for one controller is overkill. In addition, the display system structure should reflect the number of CRTs that are typically used for most operational adjustments by the controller. Job sampling to identify controller workload has been successfully used for oil refinery operators and has now been successfully applied to pipeline controllers. The data provide objective measures of current controller characteristics that provide insight into how the job can or should be altered in the future. By establishing where the controllers are now, it is easier to understand how to get them to where you want them to be in the future.

Вам также может понравиться