Elsevier

Fuzzy Sets and Systems

Volume 293, 15 June 2016, Pages 127-143
Fuzzy Sets and Systems

Methodology for analyzing the dependencies between human operators in digital control systems

https://doi.org/10.1016/j.fss.2015.04.002Get rights and content

Abstract

In a digital control system, the dependency model between the actions of operators differs from that in a conventional control room because information sharing and the main control room (MCR) operations are team operations. Dependencies between the actions of operators are more common in a digital control system compared with a conventional control room because operators share the same information and MCR operations are directed by team decisions. Therefore, assessing the dependencies between operators is an important aspect of human reliability analysis. In this study, we use a fuzzy logic-based approach to evaluate the dependencies among the actions of operators in the present study. First, the factors that influence the dependency levels among the actions of operators are identified by analyzing the characteristic human factors in a digital control system and an analytical model of the dependencies is then constructed. Second, a method for analyzing the dependencies between the actions of operators is established based on a fuzzy logic approach. This method can simulate vague and uncertain knowledge, but it also provides a clear explanation of the origins of results and their reasoning process by tracing the steps in reasoning. Therefore, traceability and repeatability are characteristics of the proposed method. Third, we present a case study to demonstrate the proposed approach. Finally, we demonstrate that the results obtained are reasonable and that the established model is stable based on validations that involve data comparisons and a sensitivity analysis of the model.

Introduction

Probabilistic safety assessment (PSA) is an analytical tool used to identify potential system failures and to determine the likelihood and consequences of their occurrence [1]. PSA is used to assess the relative effects of contributory events on system reliability and plant risk. Consequently, resources are allocated to the most risk-significant segments of the system and unnecessary expenditure is reduced. Human reliability analysis (HRA) entails the assessment of human error potential in a system, which is usually performed within a quantitative risk assessment framework. HRA has three basic functions, i.e., the identification of human errors, the prediction of their likelihood, and reducing their likelihood, if necessary [2]. PSA is important in high-risk systems such as nuclear power plants (NPPs) and HRA is an integral part of PSA. An important activity within HRA is the assessment of dependencies [3]. Dependencies may occur between different tasks performed by the same operator while there can also be dependencies between operators in HRA [4]. In HRA, the dependency level may greatly impact the HRA results as well as affecting the PSA outcomes. Different dependency analyses require different methods, but we only consider the dependencies between operators, i.e., one operator detects and recovers from another operator's error.

With the rapid development of computer, information, and control technology, instrumentation and control (I&C) systems have transformed or built on analog systems to obtain semi-digital and fully digital control in NPPs. Thus, the contextual factors that affect human reliability have changed, including information displays (e.g., video display units, VDUs), process control (e.g., mouse and touch screen), the characteristics of human–system interfaces (HSIs) (e.g., the display and control layouts), procedures (e.g., computer-based procedures), tasks (e.g., interface management tasks), alarm systems, decision support systems, team structure, communication, and the workplace environment, thereby modifying the roles and functions of operators, human–human interactions, and human–system interactions. Therefore, the introduction and development of digital technical-based HSI has changed the context in which operators work. Obviously, these changes have also affected the dependencies between operators and the criteria employed to assess the state levels of influential factors, and thus it is necessary to establish a new dependency model or method to satisfy these new requirements.

In particular, the operators obtain information from the control panels by walking back and forth in traditional analog control rooms, which makes it difficult to share information among operators without delays. However, the use of a computerized monitoring system in advanced digital main control rooms (MCRs) allows more detailed plant information to be displayed, so the operators share almost the same information and computerized procedures. The primary loop (reactor operator, RO) and secondary loop (turbine operator, TO) operators are responsible for system operation, while the coordinator/shift supervisor (senior reactor operator, SRO) and safety engineer (shift technical advisor, STA) are responsible for monitoring and error recovery. Therefore, compared with conventional analog control systems, the dependencies between operators may be stronger in digital control systems due to information sharing. For example, in an advanced MCR, it is possible for an operator to detect and recover another operator's error when they are performing the same operation in a timely manner [5].

Many methods related to HRA such as the technique for human error rate prediction (THERP) [4], accident sequence evaluation program (ASEP) [6], and standardized plant analysis risk-human reliability analysis method (SPAR-H) [7] consider the dependencies between human failure events (HFEs) in conventional MCRs. With respect to the methods used to analyze the dependencies between operators, Grobbelaar et al. [8] and Cepin [9], [10] developed a decision tree (DT)-based method for analyzing the dependencies between human actions or HFEs. Lee et al. [5] also developed an independent event tree-based method for analyzing the dependencies between operators in an advanced MCR. However, these methods do not fully consider the characteristics of digital MCRs, the factors that affect the dependencies between operators, the state levels of factors, and the importance of other factors, and thus a considerable level of expert judgment is required. Therefore, these evaluation techniques are inadequate due to a lack of traceability and repeatability as they are not based on a transparent expert elicitation process because of the fuzziness and uncertainty of expert judgment. Zio et al. [3] presented a method based on fuzzy expert systems to model the dependencies among human errors, but this method cannot be used to assess the dependencies among operators in digital control systems. To overcome these limitations, we aim to build a new fuzzy logic-based method in the present study to analyze the dependencies between operators in digital MCR. The proposed method has the following merits compared with previous approaches.

(1) The dependencies between operators are affected by many contextual factors such as the procedure quality and information quality, where the different state levels of these factors correspond to different dependency levels between operators. However, assessing the state levels of these factors is complex and uncertain due to the lack of suitable assessment criteria, the requirement to consider many features, and other issues. Moreover, some experts have difficulty determining exact values for the state levels of factors because of their limited knowledge, ability, and experience, and thus they employ descriptive language or a range of values to describe state levels, such as “good,” “about 7,” “probably 5–7,” or “(0.3, 0.5, 0.7).” Therefore, it is more realistic to use fuzzy sets to describe the fuzzy assessments of factors if the decision maker considers that a fuzzy judgment is more credible than an exact value judgment and more in line with real thinking.

(2) Fuzzy logic-based methods have traceability and repeatability, which can facilitate the implementation of sensitivity analysis, whereas traditional methods such as THERP and DT-based methods lack traceability and repeatability because they require many expert judgments.

(3) The proposed model for dependency analysis between operators considers the interrelations among factors that affect dependencies, which can avoid counting these factors twice, thereby improving the accuracy of the results.

(4) In this study, we focus mainly on studying the dependencies between operators in the digital MCRs of NPPs. We provide a fuzzy logic-based approach to analyze the dependencies, whereas other methods either focus on analog NPPs but not digital NPPs (such as THERP), or on the dependencies among HFEs/human errors but not the dependencies among operators/error recovery (such as Zio et al.'s method), or even the dependencies between operators in digital NPPs using different methods with various flaws (such as Lee et al.'s method), but without using a fuzzy logic-based method. Therefore, we consider that our proposed method has several advantages and it can provide theoretical and practical support when assessing the dependencies between operators in digital NPPs.

The remainder of this paper is organized as follows. In Section 2, we analyze the characteristics of human factors in a digital control system, where we identify the factors that affect the dependencies among operators, and we finally construct an analytical model of these dependencies. In Section 3, we build a method for analyzing the dependencies between operators using a fuzzy logic-based approach. This method can simulate vague and uncertain knowledge, as well as giving a clear explanation of the sources of the results and the reasoning process by tracing the steps during reasoning. In Section 4, we present a case to demonstrate the proposed approach. Section 5 describes empirical research using the proposed method based on comparisons with other methods, where we validate its effectiveness using sensitivity analysis. In Section 6, we give our conclusions and discus the results of this study.

Section snippets

Characteristic human factors in digital control systems

In NPPs, operators in MCRs observe and manipulate an extremely complex system. In the past, this required walking along a large control panel, taking readings from gauges, and adjusting knobs and levers. However, in a digital MCR, operators manage and control the system using a computer-based work station. This changed context affects the activities of operators. Therefore, this change may have adverse effects on the performance of operators. For example, it is likely to produce new types of

Fuzzy logic-based method for analyzing the dependencies between operators

Fuzzy logic was first introduced by Zadeh [18] as a mathematical method for representing vagueness in everyday life. In contrast to classical logic systems, fuzzy logic systems consider modes of reasoning that are approximate rather than exact. Fuzzy logic starts with the concept of a fuzzy set, which is a set without a crisp and clearly defined boundary. The fundamental difference between fuzzy logic and conventional modeling techniques is related to the definitions of sets. Traditional set

Case study

In this section, we present an application of the proposed method. In digital control systems, it is assumed that a steam generator single tube rupture (SGTR) occurs in a NPP and the pressure of the primary loop decreases. The SRO is required to diagnose the current plant situation according to the emergency operating procedures. However, due to the failure of sensor N16, the SRO misdiagnoses the SGTR accident as a small loss of coolant accident, and thus there is a failure to isolate the

Model validation

In general, it is very difficult to validate HRA models using empirical data [25]. When modeling human operator dependencies, the problem is complicated further because the probabilities of failure for operator dependencies are conditioned systematically and realistically within a testing procedure. Therefore, the proposed model/method was compared with other methods based on operator dependencies to validate the results obtained. The state levels of factors that affect the operator

Conclusions and discussions

Dependencies between the actions of operators are more prevalent in a digital control system because the operators share the same information and MCR operations are directed by team decisions. Therefore, in order to obtain more precise and reliable HRA data, the dependencies of the actions of operators must be considered in HRA. Therefore, the main purpose of this study was to construct a method for analyzing the dependencies of the actions of operators in digital control systems.

By analyzing

Acknowledgements

This study was supported by the National Natural Science Foundation of China (71371070, 71071051, 71301069), Research Project of Lingdong Nuclear Power Co. Ltd. (No. KR70543), the Innovation Ability Construction Projects based on the new Industry-Academy-Research Cooperation of Hunan Province (2012GK4101), and the Construct Program of the Key Discipline in Hunan Province. We would like to express our gratitude to the staff of two Chinese nuclear power plants (Qinshan Phase 3 and Dayabay) for

References (26)

  • A.D. Swain et al.

    Handbook for human reliability analysis with emphasis on nuclear power plants application, NUREG/CR-1278

    (1983)
  • S.J. Lee et al.

    Modeling of a dependence between human operators in advanced main control rooms

    J. Nucl. Sci. Technol.

    (2009)
  • A.D. Swain

    Accident sequence evaluation program: human reliability analysis procedure, NUREG/CR-4772

    (1987)
  • Cited by (22)

    • Assessing dependence in human reliability analysis using probabilistic linguistic term sets

      2022, Annals of Nuclear Energy
      Citation Excerpt :

      Many studies have been conducted to assess the dependence between operators (Gao et al., 2022; Wang et al., 2020; Hou et al., 2021). For instance, Li et al. (2016) proposed a fuzzy logic-based approach for evaluating dependencies between human operators in digital control systems, which could effectively deal with vague and uncertain information while having good interpretability and traceability. Liu et al. (2018) used the linguistic variables to represent the judgments of experts, and proposed a large group dependence assessment method based on interval 2-tuple linguistic variables and cluster analysis method with an extended Muirhead mean operator.

    • Methodology for dynamic reliability assessment of team situation awareness of digital nuclear power plants

      2022, Progress in Nuclear Energy
      Citation Excerpt :

      The digital human-system interface (HSI) has changed the way of information presentation (e.g., a large amount of information is displayed limitedly on a computer screen) and procedures (e.g., computer-based procedures). control (e.g., soft control), task (e.g., interface management task), team structure, communication and cooperation etc.(Li et al., 2016). The above-mentioned changes in the contextual characteristics not only affect the individual's SA (e.g., Keyhole effect) (Seong, 2009), but also affect the TSA.

    • Methodology for assessing dependencies between factors influencing airline pilot performance reliability: A case of taxiing tasks

      2020, Journal of Air Transport Management
      Citation Excerpt :

      PSF dependency assessment is an important and challenging task in HRA (Ambroggi and Trucco, 2011; Chang and Mosleh, 2006). However, unlike the widely recognized and studied human failure event (HFE) dependencies (Čepin, 2008; Li et al., 2016; Podofillini et al., 2010; Zheng and Deng, 2018; Zio et al., 2009), PSF dependencies are always ignored (Groth and Mosleh, 2010). Many HRA methods assume that PSFs are independent each other, such as the Technique for Human Error Rate Prediction (THERP) (Swain and Guttman, 1983), the Success Likelihood Index Method (SLIM) (Embrey, 1984) and the Human Error Assessment and Reduction Technique (HEART) (Williams, 1986).

    View all citing articles on Scopus
    View full text