Automation Inflicted Differences on Operator Performance in Nuclear Power Plant Control Rooms

Today it is possible to automate almost any function in a human-machine system. Therefore it is important to find a balance between automation level and the prerequisites for the operator to maintain safe operation. Different human factors evaluation methods can be used to find differences between automatic and manual operations that have an effect on operator performance; e.g. Predictive Human Error Analysis (PHEA), NASA Task Load Index (NASA-TLX), Halden Questionnaire, and Human Error Assessment and Reduction Technique (HEART). Results from an empirical study concerning automation levels, made at Ringhals power plant, showed that factors as time pressure and criticality of the work situation influenced the operator’s performance and mental workload more than differences in level of automation. The results indicate that the operator’s attention strategies differ between the manual and automatic sequences. Independently of level of automation, it is essential that the operator retains control and situational understanding. When performing a manual task, the operator is “closer” to the process and in control with sufficient situational understanding. When the level of automation increases, the demands on information presentation increase to ensure safe plant operation. The need for control can be met by introducing “control gates” where the operator has to accept that the automatic procedures are continuing as expected. Situational understanding can be established by clear information about process status and by continuous feedback. A conclusion of the study was that a collaborative control room environment is important. Rather than allocating functions to either the operator or the system, a complementary strategy should be used. Key parameters to consider when planning the work in the control room are time constraints and task criticality and how they affect the performance of the joint cognitive system.However, the examined working situations were too different with respect to levels of automation and therefore it is not possible yet to propose general automation level guidelines. Further studies are still needed