Zijian Yin, Hongchao Zhang, Peng Cheng, Peiyuan Wang, Zhizhong Li (2024). Nuclear Power Plant Maintenance and Commissioning Activities from the Perspective of HRA: A Comparative Study. In: Human Error, Reliability, Resilience, and Performance. AHFE Open Access, vol. 132, pp. 34–43. https://doi.org/10.54941/ahfe1004878
1 Department of Industrial Engineering, Tsinghua University, Beijing, 100084 P. R. China; National Key Laboratory of Human Factors Engineering, Beijing, 100094, P. R. China ↗
2 Sanmen Nuclear Power Co., Ltd., Taizhou, Zhejiang, 317112, P. R. China ↗
The recent Human Reliability Analysis (HRA) method developed by U.S. Nuclear Regulatory Commission, Integrated Human Event Analysis System (IDHEAS), provides a well-grounded model to analyze human errors as well as a promising interface for generalizing and integrating human error data from various sources. Nevertheless, the task context and human performance in different domains, even in different applications within one domain, manifest quite different features which make it elusive to draw an inter-sector or inter-application comparison. In light of this contradiction, one question arises: Can HRA in distinct applications be cross-referenced to one another? Motivated by this, an interview study was conducted as a preliminary attempt to figure out the similarities and differences between two typical activities in the nuclear domain, i.e., maintenance and commissioning, with respect to two vital HRA elements, i.e., error modes and performance shaping factors (PSFs). A total of 21 engineers in a nuclear power plant were recruited to participate in the interview about salient error modes and PSFs in maintenance activities as well as a comparison with commissioning activities. Results show that the two activities share analogous error modes and PSFs on the whole, but vary in the patterns and distributions of each individual PSFs. Hence, results of the present study indicate that human errors in these two comparable activities could be characterized by a unified taxonomy, affording positive evidence for the thrown question. However, the discrepancy in the PSF patterns should never be neglected, which could lead to the divergency of nominal PSF levels and nominal human error probabilities.
The cognitive-based Human Reliability Analysis (HRA), often classified as the second generation HRA, seems to provide a promising way to analyze human errors in different contexts and sectors. One example of such HRA methods is Cognitive Reliability and Error Analysis Method
Nuclear Power Plant Maintenance and Commissioning Activities 35
(CREAM; Hollnagel, 1998), which has been applied in various sectors including nuclear industry (Marseguerra, Zio and Librizzi, 2006), oil and gas industry (Zhang and Tan, 2018), etc. Another method recently developed by U.S. Nuclear Regulatory Commission (NRC), Integrated Human Event Analysis System (IDHEAS), was also expected to be “applicable, but not lim- ited, to all nuclear risk-informed applications” (Xing, Chang and DeJesus Segarra, 2022). The underlying cognitive model of IDHEAS, the macro- cognition model (Whaley et al., 2016), uses five macro-cognitive functions (i.e., detecting, understanding, decision-making, action, and teamwork) to characterize human activities. Each macro-cognitive function includes sev- eral cognitive processors which are driven by various cognitive mechanisms. Such general and technology-neutral model can enhance our understanding of human errors and provide an interface for generalizing and integrating human error data from various sources (Xing, Chang and DeJesus Segarra, 2020). Although IDHEAS is expected to suit most nuclear applications, the diver- gent features of different nuclear applications are noteworthy, not to mention the divergence across different sectors. For example, Yin, Liu and Li (2021) highlighted the difference in contextual factors between control room and commissioning work. Obviously, the control room activities and ex-control room activities are quite distinct in that control room operators monitor the indicators and alarms of the plant and manipulate the plant through soft con- trols, while local operators outside the control room interact directly with the local equipment. In addition, different ex-control room activities also possess different characteristics. Two similar but distinct ex-control room activities are commissioning and maintenance (Yin et al., 2023). Note that such “ex-control room” is not that absolute in that some actions may also require operations in the control room. Commissioning and maintenance are two typical ex-control room activities in different stages of a Nuclear Power Plant (NPP). Commissioning is the last stage before the commercial operation of an NPP, serving as an important transition from engineering to operation. The main objective of commissioning is to validate that all com- ponents and systems can function normally and meet relevant criteria, and this is achieved through various functional tests. Maintenance happens dur- ing the whole operation stage of an NPP aiming at ensuring all components and systems to perform design functions, and mainly includes preventive and corrective maintenance. Although both belonging to ex-control room activ- ities, commissioning and maintenance have salient differences with regard to several aspects such as objectives, working contents, risks, and contexts (Yin et al., 2023).
As a starting point, the present study tries to answer the question that whether HRA in commissioning and maintenance can be cross-referenced to one another. If so, the two activities might be assessed using a common HRA method, and human error data from the two sources could be better general- ized and integrated. Remember that HRA aims at assessing the “reliability” of a task performed by humans in a certain context. Aligning with that, the feasibility of cross-reference would depend on meeting of several criteria:
(a) Tasks in the two activities should be comparable, so that the assessed objects of HRA are the same. This criterion is the most basic of cross- reference, as it makes limited sense to juxtapose one activity with another that exerts very different requirements on human information processing. (b) Error modes in the two activities should be comparable. Error modes depend on the task type, and are the manifestation of how humans might fail to perform the task. (c) Contexts in the two activities should be comparable. There should exist some shared contextual factors, so that the effects of such factors could be modeled and analyzed across activities. (d) The effects of contextual factors on human errors should be homoge- neous across activities. HRA models such effects as the quantitative relationship between contextual factors and Human Error Probabilities (HEPs).
In the present study, criteria (b) and (c) are addressed while the others are left for future work. Hence, two vital elements in HRA are specifically considered here: Error Modes (EMs) and contextual factors (i.e., Perfor- mance Shaping Factors, PSFs). These two variables are determinants of HEPs, and should therefore be treated carefully. Although existing studies (e.g., Kim et al., 2009; Yin, Liu and Li, 2021) have investigated relevant topics in commissioning and maintenance, there lacks an inter-activity compari- son of EMs and PSFs. For this purpose, an interview study was designed to collect maintenance engineers’ intuitive perception of EMs and PSFs in their daily work. In addition, most interviewees have also been involved in commissioning activities, making it possible to expose some compara- tive features of the two activities. The interview results reported by these interviewees are derived from their personal working experience in both maintenance and commissioning activities, and therefore provide direct and reliable information related with criteria (b) and (c). Based on the findings in this study and previous work, the question thrown before could be answered partly.
Interview of Maintenance Engineers A total of 21 maintenance engineers from a domestic NPP were recruited to participate in face-to-face interviews. They have an average of 11.6-year (SD = 4.3) working experience in this NPP. Among the 21 participants, 19 have been involved in the commissioning and startup of this NPP, and therefore possess intuitive perception of comparative features of mainte- nance and commissioning activities. During the interview, the participant was required to answer the following three questions. Note that for the two par- ticipants who have not been involved in the commissioning work, only the first question was asked.
Common human errors and contributing factors in maintenance.
Common human errors and contributing factors in commissioning.
Nuclear Power Plant Maintenance and Commissioning Activities 37
Difference between these two activities, in terms of requirement on human, team structure, organization, target equipment or system, working envi- ronment, procedure and diagram, task, and risk.
The interview results are summarized in Tables 1–3. Tables 1 presents reported EMs of human errors in maintenance and commissioning. On the whole, the findings agree with those in previous studies. For example, EMs in test and maintenance reported in Kim et al. (2009), such as omission, inap- propriate action (too fast, too long, too much, etc.), and wrong object, are also reported in Table 1. According to the findings in Table 1, those items shared by maintenance and commissioning reflect the commonality of the two activities, and those unshared items should be paid more attention to. In Table 1, EMs No. 1, 4, 7, 9, 11 & 13 are present only in maintenance. However, these EMs are judged not unique to maintenance but also applica- ble to commissioning. For example, it is reasonable to speculate that EM No. 9 (errors in following procedures) is also applicable to commissioning, since there are quite a lot of procedural tasks in commissioning. In Table 1, EMs No. 15 & 16 are present only in commissioning. The absence of these EMs in maintenance could be explained by technical differences. EM No. 15 (fail to check prerequisites such as the blocking status) can be explained by the fact that blocking in maintenance is strictly controlled and will be verified by several roles including the maintenance preparation engineer, blocking man- ager, maintenance task leader, etc.; EM No. 16 (temporary alterations are disrupted due to inadequate signs) can be explained by the fact that there are a few temporary alterations in maintenance which will be well controlled. Therefore, the presence of these two EMs depends highly on the technical conditions of NPPs but not differences in human information processing.
Table 1. Common human errors in maintenance and commissioning, comparable items placed in the same row.
Table available in the original PDF paper.
No. Maintenance Commissioning 1 Perform tasks following one’s own habits, memory or understanding, rather than the procedure and diagram (7*) 2 Not aware of the change and risk in working conditions (3)
Not aware of the change in working conditions (1) 3 Operate on a wrong target (9) Operate on a wrong target (2) 4 Accidental touch (4) 5 Wrong operation due to lack of skills (3) Wrong operation due to lack of skills (1) 6 Wrong operation by mistake (8) Wrong operation by mistake (2) 7 Inadequate preparation, such as tools and procedures (4) 8 Errors in using tools, such as multimeter (1) Errors in using tools (1) 9 Errors in following procedures, such as skipping and omitting steps (5) 10 Errors in preparing instructions (1) Omit some points to be validated (1) 11 Fail to supervise (2) (Continued)
Table 1. Continued No. Maintenance Commissioning
Table available in the original PDF paper.
12 Forget to recover temporary alterations (2) Forget to recover temporary alterations (2) 13 Errors in following administrative procedures, such as no-permit work (1) 14 Errors in communication and handover (3) Errors in communication (2) 15 Fail to check prerequisites such as the blocking status (2) 16 Temporary alterations are disrupted due to inadequate signs (1) The number in the bracket represents for the frequency of this item being mentioned by interviewees. Same for the other tables.
Table 2 presents reported PSFs in maintenance and commissioning. Similar with Table 1, the unshared items deserve careful consideration. In Table 2, PSFs No. 3, 7–10, 12, 17, 19, 20, 22–24 are present only in maintenance. However, these PSFs are judged not unique to maintenance, some of which have been reported in the previous study on PSFs in commissioning (Yin, Liu and Li, 2021). Such PSFs (as well as those EMs in Table 1) are not mentioned in commissioning during the interview, partly because of the limited interview sample size. Besides, it had been years since the accomplishment of commis- sioning of this NPP, so the participants might fail to list some commissioning EMs and PSFs due to memory fades. In Table 2, PSFs No. 25–32 are present only in commissioning, reflecting some differences in contexts. For example, PSF No. 28 (high personnel mobility) is closely related with the flexible team structure in commissioning, and PSF No. 30 (concurrent work in the work- place) reflects the existence of lots of concurrent work (e.g., construction activities) during the commissioning stage.
Table 2. Common PSFs in maintenance and commissioning, comparable items placed in the same row.
Table available in the original PDF paper.
No. Maintenance Commissioning
1 Lack of knowledge, experience, familiarity, and skills (9)
Lack of knowledge, experience, and skills (3) 2 Time and progress pressure (6) Time and progress pressure (3) 3 Negative personality, such as impatience (1) 4 Adverse mental and physical status, such as fatigue (8)
Fatigue (3) Negative emotion (1) 5 Distraction (4) Irrelevant talks in the workplace (1) 6 Lack of responsibility (3) Lack of responsibility (3) 7 Close to getting off work or holiday (2) 8 Cutting corners (1) 9 Empiricism (1) 10 Habit in routing tasks causing habit intrusion (1) (Continued)
Nuclear Power Plant Maintenance and Commissioning Activities 39
Table 2. Continued No. Maintenance Commissioning
Table available in the original PDF paper.
11 Inconsistency with behavioral norms (1) Inconsistency with behavioral norms (1) 12 Taking a chance (1) 13 Inadequate peer-check or supervision (4) Inadequate supervision (1) 14 Inadequate preparation (1) Inadequate preparation (2) 15 Technical system faults (1) Installation errors in construction (1) Manufacturing errors (1) Poor equipment quality (1) 16 Inadequate procedures and documents (2) Inadequate procedures and technical guides (1) Lack of test criteria (1) 17 Frequent upgradation of procedures (1) 18 Incorrect procedures (1) Incorrect procedures (2) 19 Mismatch of procedures and actual conditions (1) 20 Harsh working conditions, such as radiation (3) 21 Human error traps in working conditions and equipment, such as similar coding and chaos layout (7)
Human error traps in working condi- tions (1) Densely laid-out cables (1) 22 Narrow working space (1) 23 Noise (1) 24 Unclear task requirements (1) 25 Lack of incentive (1) 26 Lack of vigilance due to low operation risk (2) 27 Poor ability to work with others (1) 28 High personnel mobility (1) 29 Complicated personnel constitution in a team (1) 30 Concurrent work in the workplace (5) 31 Dynamically changed working conditions (1) 32 Culture difference in design (1)
Table 3 focuses especially on the differences in contexts of the two activ- ities, which should be considered carefully when we model the context. Findings in Table 3 highlight several meaningful concerns.
Some differences cannot be adequately identified in current PSF frame- works. For example, the level of detail of maintenance procedures is lower than that of commissioning procedures. The effect on HEP needs to be clarified in further research. Another example is the risk faced during a task: There exist more plant operation risks during maintenance, which could influence operators’ behavior. Though not modeled in PSF, the effect of such factors will contribute to HEP variability, which is less studied in HRA.
Some differences cannot be adequately modeled using current PSF defi- nition and level designs. For example, maintenance and commissioning exert different requirements on human knowledge. A simple distinction
of several knowledge levels (e.g., low, nominal, and high) seems to be incompetent to model such difference. This issue along with the afore- mentioned one constitutes the main source for within-category variability of HEP (Greco, Podofillini and Dang, 2021).
The judgement criteria of PSF status or definitions of PSF levels in differ- ent activities may vary. A “complex” task in maintenance may be judged as a simple one from commissioning engineers’ perspective, as in commis- sioning tasks there are more device and systems involved. This concern highlights the necessity of objective criteria for PSF evaluation.
The distribution of single PSF status in different activities may vary. For example, commissioning teams are often loosely organized. Hence, the probability of the team structure being loosely organized in commissioning would be higher than that in maintenance.
Table 3. Comparison between maintenance and commissioning.
Table available in the original PDF paper.
Aspects Differences Requirements on human
In commissioning, one engineer is responsible for a plant system (e.g., Turbine Bypass System), and is therefore supposed to possess necessary knowledge on the whole system. In maintenance, one engineer is in charge of one or several types of equipment (e.g., valves), and the required knowledge is narrower but more in-depth (6)
Commissioning engineers’ skills focus on the test of system functions, while maintenance engineers’ skills focus on the defect handling of equipment (3)
In commissioning, engineers are expected to deal with more complex, unplanned, and dynamic conditions and control the entire test progress. This brings a higher requirement on skills. In maintenance, the working conditions are more standardized. Yet, some online maintenance tasks under emergency working conditions are also challenging (4)
Commissioning tasks are somewhat error-tolerant since the plant has not come into power operation. Maintenance engineers are strictly required to perform actions with determinate and predictable consequences (2)
Besides technical skills, there are also fair requirements on commissioning engineer’s non-technical skills, such as coordination and project management (1) Team structure
There exist much more inter-team cooperation activities in commissioning. Many entities, such as architects, constructors, and the plant owner, are involved in. Hence, the team structure of commissioning teams is often more loosely organized and flexible, while the team structure of maintenance teams is often more compact and fixed (8) Organization
Various entities need to be coordinated in the organization of commissioning (2)
The progress of commissioning is controlled by the commissioning department, while the progress of maintenance is controlled by the operation schedule. The former is more loosely scheduled (4) (Continued)
Nuclear Power Plant Maintenance and Commissioning Activities 41
Table 3. Continued Aspects Differences
Table available in the original PDF paper.
Plant systems are independent during commissioning, but are strongly coupled during maintenance (2)
Target equipment or system
The plant is in power operation during maintenance, but not during commissioning. The different plant conditions lead to different error tolerances and different requirements for blocking (5)
Some systems are unavailable before they are tested during commissioning (2)
During commissioning, the noise and temperature in the factory buildings are lower since many machines are not running (1)
There are more uncertainties during commissioning such as temporary alterations (2) Working environ- ment
There exists concurrent work, especially construction work, during commissioning. After the plant comes into power operation, the cleanliness becomes better (13)
The level of detail of maintenance procedures is lower than that of commissioning procedures (10)
The nuclear fuel has not been loaded during commissioning, so there exists no radiation (5) Procedure and diagram
The procedures and diagrams in commissioning have more errors and changings, leading to frequent upgradation. The procedures and diagrams used in maintenance have been validated many times, and are therefore more accurate (14) Task
Commissioning tasks aim at testing and validating the functions of plant systems, while the objective of maintenance tasks is the maintenance and repair of equipment. Hence, the working contents of these two activities are very different. Some maintenance tasks like equipment tests are somewhat similar with commissioning tests, but the objective and criteria are still largely distinct. In addition, many tests would not be conducted ever since they have been done during commissioning (14)
There are more equipment and systems involved in commissioning tasks than in maintenance tasks. Many commissioning tasks would test the inter-operation of various systems, while most maintenance tasks only focus on single equipment (4)
The work intensity during commissioning is somewhat comparable with that during the major overhaul and higher than that during routine maintenance (10) Risk
There exist more industrial safety risks during commissioning, due to the concurrent construction work (2)
There are damage risks for equipment during commissioning, due to potential inappropriate tests (1)
There exist more plant-related and nuclear risks during maintenance (2)
DISCUSSIONS AND CONCLUSION Tables 1–3 report the interview results with regard to the comparison of maintenance and commissioning activities. From the results, several mean- ingful insights can be summarized. Firstly, Maintenance and commissioning share very similar human error modes. It can be concluded from Table 1 that possible EMs in these two activities are quite similar. As discussed before, most EMs that are present only in maintenance are deemed as also applicable
in commissioning. Hence, we can draw a conclusion that EMs in mainte- nance and commissioning are quite comparable. This implies the possibility of compiling a unified error taxonomy to analyze human errors in these two activities. To confirm this speculation, a comparative task analysis study is necessary to compare task characteristics in maintenance, commissioning, and other ex-control room activities, since EMs are closely related to the task.
With regard to PSF, Common PSFs in maintenance could also be found in commissioning. By comparing common PSFs presented Table 2 and the com- missioning PSF list in Yin, Liu and Li (2021), it can be concluded that most PSFs in maintenance also exist in commissioning. Having said that, some unique commissioning PSFs in Table 2 are judged only applicable in com- missioning such as “lack of vigilance due to low nuclear risk” and “high personnel mobility.” This indicates the possibility of compiling a unified PSF framework, or at least different PSF frameworks with minor adjustment, to characterize task contexts in these two activities. In fact, Yin, Liu and Li (2021) have demonstrated that the PSFs in commissioning and in control room overlap heavily. A full set of PSFs that takes all control room and ex- control room activities into considertaion would benefit the generalization and integration of human error data in the nuclear industry. Nevertheless, it is noteworthy that the nominal PSF levels in maintenance and commissioning may be defined differently. Table 3 demonstrates typical differences of these two activities, which arise from different objectives and plant conditions. For example, the two activities have different requirements on human knowledge, with commissioning emphasizing knowledge on systems and maintenance emphasizing knowledge on equipment. How to define an appropriate nom- inal level for the PSF knowledge then becomes a problem. Although there exist some potential indicators such as the number of working years and training frequency, it is hard to say that maintenance and commissioning engineers with the same working years would commit errors with the same probability. Another example is task complexity. Table 3 indicates that com- missioning tasks are more complex than maintenance tasks and have higher requirements on human knowledge and skills. Hence, a commissioning task with nominal task complexity might be evaluated as high task complexity using the criteria for maintenance. Then, a divergence in PSF levels occurs. In other words, the HEP of a commissioning task with all PSFs in nominal levels may equal to the HEP of a maintenance task with the task complexity PSF in high level. This would make the results of HEP generalizing and integrat- ing (e.g., Xing, Chang and DeJesus Segarra, 2020) problematic. Besides the importance of a universal definition of PSF and PSF levels, the distribution of PSF levels could be divergent. As mentioned before, if commissioning and maintenance share a common PSF, the distribution of corresponding PSF lev- els, or the probabilities of this PSF locating at each level, could be divergent. Divergent distributions of PSF levels will lead to different expected probabili- ties of each PSF level and different expected HEPs finally. This issue should be considered in predictive HRA, especially when a PSF status is prospectively unknown.
Nuclear Power Plant Maintenance and Commissioning Activities 43
In conclusion, the interview results show that commissioning and mainte- nance as two typical ex-control room activities share some commonalities in regard to EMs and PSFs. This implies that these two activities might be ana- lyzed using the same HRA model. Having said that, the different objectives of these two activities and different plant conditions result in considerable differences in various context aspects, which can lead to potential discrepant definitions and distributions of PSF levels. Therefore, human error data from the two sources should be treated with caution. In future work, more efforts are in need to explore a comprehensive framework for recording human error information, so that human error data from various domains can be integrated.
ACKNOWLEDGMENT
This study was supported by the National Natural Science Foundation of China (Grant No. T2192932 & T2192933).
REFERENCES
Greco, S. F. Podofillini, L. Dang, V. N. (2021) A Bayesian model to treat withincategory and crew-to-crew variability in simulator data for human reliability analysis, Reliability Engineering & System Safety Volume 206. Hollnagel, E. (1998) Cognitive Reliability and Error Analysis Method. Oxford:
Elsevier Science. Kim, J. Park, J. Jung, W. Kim, J. T. (2009) Characteristics of test and maintenance human errors leading to unplanned reactor trips in nuclear power plants, Nuclear Engineering and Design Volume 239, No. 11. Marseguerra, M. Zio, E. Librizzi, M. (2006) Quantitative developments in the Cogni-
tive Reliability and Error Analysis Method (CREAM) for the assessment of human performance, Annals of Nuclear Energy Volume 33, No. 10. Whaley, A. M. Xing, J. Boring, R. L. Hendrickson, S. M. L. Joe, J. C. Le Blanc, K. L.
Morrow, S. L. (2016) Cognitive Basis for Human Reliability Analysis. Washington, DC: U. S. Nuclear Regulatory Commission. Xing, J. Chang, Y. J. DeJesus Segarra, J. (2020) DRAFT - Integrated Human Event Analysis System for Human Reliability Data (IDHEAS-DATA). Washington, DC: U. S. Nuclear Regulatory Commission. Xing, J. Chang, Y. J. DeJesus Segarra, J. (2022) IDHEAS-ECA - An Integrated Human
Event Analysis System for Event and Condition Assessment (Presentation) Yin, Z. Li, Z. Liu, Z. Yang, D. Zhang, J. Long, L. Zhang, Y. Gong, B. (2023) Collection of IDHEAS-based human error probability data for nuclear power plant commissioning through expert elicitation, Annals of Nuclear Energy Volume 181. Yin, Z. Liu, Z. Li, Z. (2021) Identifying and clustering performance shaping factors
for nuclear power plant commissioning tasks, Human Factors and Ergonomics in Manufacturing & Service Industries Volume 31, No. 1. Zhang, R. Tan, H. (2018) An integrated human reliability based decision pool generating and decision making method for power supply system in LNG terminal, Safety Science Volume 101.
LISTED IN THE AHFE DIGITAL BOOKSTORE CATALOGUE
Explore the complete digital volume
Human Error, Reliability, Resilience, and Performance
Read this individual research paper free through the original publisher, or explore the full digital proceedings book.
ISBN 978-1-964867-08-3
Explore this ISBN in the AHFE Bookstore ↗Listed in the July 2026 bookstore catalogue; search by ISBN to confirm current availability · Digital book purchase optional