Results concerning operators’ performance

Some dimensions of operator performance were measured providing quantitative information of the use of LSDs. The following measures were used: duration of time to event detection for each scenario, source of the first deviation detected and percentage gazing time to different information sources and number.

5.4.2.1 Duration of time to event detection for each scenario

This measure provides information of how long it takes for operators to detect the failure. It was measured by recording the time from the beginning of the deviation to the event detection. Detection of failures was based either on information on LSD screens or information on desktop screens. Overall, failures could be detected fast for all scenarios for the two crews that did not know the purpose of the test beforehand. Figure 23 shows the detection times for Scenarios 1 and 2 for the two crews. As can be seen, detection time was much longer for Scenario 1 than for Scenario 2 for both crews. There were, however, no big differences between the IRD- and Loviisa-display conditions, and failures could be detected nearly equally fast in the two conditions.

As said above, three crews of operators participated in the simulator tests.

However, in the table bellow is only compared two of the crews that had not participated in the development of the Fortum IRD pilot or in the development of the test scenarios and that were novice to use IRD displays. It could have been

supposed that the third crew that were familiar with the Fortum pilot display and scenarios would have perform more effective manner than the other two crews.

Interestingly, the differences in performance between crews were moderate, and the performance of the novice operators was in general not worse than those who participated in the display design.

5.4.2.2 Source of the first deviation detection

All the crews identified the failure in each simulator run and obtained a correct diagnosis of its cause. Operators, however, could not necessarily detect all the indicators of the failure from the IRD displays. Typically, the operators only detect some of these changes, and they do not necessarily notice the first sign of the failure but only some of its consequences. Sometimes, they had to actively search for a particular component especially if it was placed in an unexpected location.

Sometimes other information than information obtained from the IRD displays was used in the identification of the problem, and it was quite typical that in this case the first sign was received from the event list. Appendix J shows from which source the problem was first detected for each scenario and for each crew.

As can be seen, except for two simulation runs, the failure was first detected from the large screen; for one simulation run it was detected simultaneously from the LSD and from the workstation display. For the six simulation runs in which Loviisa overview displays were used, the failure was detected from the large screen in five cases; in one of the cases it was first noticed from the event list that was displayed on one of the workstation display screens. For the IRD condition, the failure was detected first from the event list in only one of the 18 simulation runs.

Since the operators knew that the usage of large screens was the target of the test, they apparently looked more tightly at the LSDs than would have been the case in the normal CR conditions and, thus, also identified the problem more often from the LSD than from the workstation display screens. Nevertheless, our results suggest that the IRD-displays provide useful information that operators can utilize in the detection of failure states in the nuclear CR.

Different signs were used in the detection of the failure in different scenarios in the IRD condition. In Scenarios 1, 2 and 6, it was typical that the change of the colour of a pump or valve symbol frame was first detected on the IRD display. For Scenario 4, the operators first detected the descent of the level of the feed-water tank on the IRD display; for Scenario 5, the failure was also detected by looking at the graphical information showing the level of the pressurizer, and for Scenario 6, the operators noticed that the RA- and RV-pumps were not working properly.

a)

b)

Figure 24. Examples of time-event plots showing gaze times to different information sources for (a) the IRD and for (b) the Loviisa LSD conditions.

As said, it was typical that the operators did not detected and identified all the indicators of the failure from the IRD displays. There seems to be many reasons for these problems. Sometimes the colour change (from one grey level to another or from grey to green) was too small; sometimes the operators had problems to interpret the changes that had occurred in graphs. The operators had also problems in identifying some of the components that were displayed because of the lack of identifiers and because of the rearrangement of displayed information.

It was also found out that some of the critical information (valve S004 that was needed in Scenario1) was totally lacking from the IRD displays.

Since the IRD displays are developed for the early detection of failures, one interesting question is whether an operator could detect a failure from the first signs (e.g. from the change of the slope of a trend curve) before the alarm was triggered. Unfortunately, since the events in most of the runs were rapidly evolving, the ‘first signs’ can be seen nearly at the same time as the alarm information is displayed. Therefore, in most of the runs, the operators detected the failure from the displayed alarm information (i.e. from the changes of the

symbol colour or from the sudden appearance of a surrounding frame). The result might have been different if the failures have been more slowly evolving.

0 10 20 30 40 50 60 70

LSD PMS Operator Other

% IRD

TRAD

Figure 25. Percentage of time operators gazed to different information sources (LSD = large screen display, PMS = process monitoring system).

5.4.2.3 Percentage gazing time to different information sources

The percentage gazing time to different information sources (ie., LSD, desktop screen, other operator) provides information of how long LSDs are gazed in relation to other information sources. Figure 24 presents two examples of graphical plot of occurrence and length of selected behavioural elements over time (time-event plots). As can be seen, the operators typically looked a longer period of time on the LSD than on the other information sources in the beginning of the simulation run. There were some differences in gazing times between IRD and Loviisa display conditions for the operators that were naïve to the purpose of the test (see Figure 25). For example, for Scenario 2 the operators gazed a little bit longer for IRD displays than for Loviisa displays. This finding suggests that Information Rich -displays may provide at least as useful information as more traditional large screens.

5.4.3 Results concerning operators’ thoughts on IRD key features