Yochan Kim

A statistical approach to estimating effects of performance shaping factors on human error probabilities of soft controls

Despite recent efforts toward data collection for supporting human reliability analysis, there remains a lack of empirical basis in determining the effects of performance shaping factors (PSFs) on human error probabilities (HEPs). To enhance the empirical basis regarding the effects of the PSFs, a statistical methodology using a logistic regression and stepwise variable selection was proposed, and the effects of the PSF on HEPs related with the soft controls were estimated through the methodology. For this estimation, more than 600 human error opportunities related to soft controls in a computerized control room were obtained through laboratory experiments. From the eight PSF surrogates and combinations of these variables, the procedure quality, practice level, and the operation type were identified as significant factors for screen switch and mode conversion errors. The contributions of these significant factors to HEPs were also estimated in terms of a multiplicative form. The usefulness and limitation of the experimental data and the techniques employed are discussed herein, and we believe that the logistic regression and stepwise variable selection methods will provide a way to estimate the effects of PSFs on HEPs in an objective manner.

A classification scheme of erroneous behaviors for human error probability estimations based on simulator data

Because it has been indicated that empirical data supporting the estimates used in human reliability analysis (HRA) is insufficient, several databases have been constructed recently. To generate quantitative estimates from human reliability data, it is important to appropriately sort the erroneous behaviors found in the reliability data. Therefore, this paper proposes a scheme to classify the erroneous behaviors identified by the HuREX (Human Reliability data Extraction) framework through a review of the relevant literature. A case study of the human error probability (HEP) calculations is conducted to verify that the proposed scheme can be successfully implemented for the categorization of the erroneous behaviors and to assess whether the scheme is useful for the HEP quantification purposes. Although continuously accumulating and analyzing simulator data is desirable to secure more reliable HEPs, the resulting HEPs were insightful in several important ways with regard to human reliability in off-normal conditions. From the findings of the literature review and the case study, the potential and limitations of the proposed method are discussed.

Empirical investigation of communication characteristics under a computer-based procedure in an advanced control room

With the development of computer-based control rooms including a computer-based procedure (CBP), shift supervisors (SSs) can directly access plant information through a CBP or personal displays instead of depending on other board operators (BOs) to obtain plant information. In relation to this change, we examined the characteristics of SS inquiry patterns in the computer-based control rooms of nuclear power plants during emergency situations. Operator behaviors and speech patterns were observed and analyzed through several experiments on simulated accident scenarios in a full-scope simulator for an advanced computer-based main control room. We found that the SS inquiry in the advanced control room had less dependency on the BOs, and that the inquiry patterns varied according to the operators and operating dates. From these findings, the necessity to establish communication standards under computer-based control rooms was discussed with some recommendations. Another requirement to reduce the cognitive workload of SSs was also discussed.

Empirical investigation of workloads of operators in advanced control rooms

This paper compares the workloads of operators in a computer-based control room of an advanced power reactor (APR 1400) nuclear power plant to investigate the effects from the changes in the interfaces in the control room. The cognitive–communicative–operative activity framework was employed to evaluate the workloads of the operators roles during emergency operations. The related data were obtained by analyzing the tasks written in the procedures and observing the speech and behaviors of the reserved operators in a full-scope dynamic simulator for an APR 1400. The data were analyzed using an F-test and a Duncan test. It was found that the workloads of the shift supervisors (SSs) were larger than other operators and the operative activities of the SSs increased owing to the computer-based procedure. From these findings, methods to reduce the workloads of the SSs that arise from the computer-based procedure are discussed.

A quantitative measure of fitness for duty and work processes for human reliability analysis

Fitness for duty and work processes have been recognized as important performance shaping factors (PSFs) for human reliability analysis (HRA). However, current HRA methods offer no or limited guidance for determining PSF levels, so analysts have relied on their expert judgment during the selection of such levels. In this study, we propose a practical framework to quantitatively measure the levels of socio-psychological PSFs using human error data, based on plant experiences. This methodology calculates the error occurrence intervals and their moving average for a certain error cause reported in inspection reports. The proposed framework is applied to the HuRAM+ (Human related event Root cause Analysis Method plus) database as a case study. The usefulness and requirements of the proposed framework are then discussed.

The use of the SACADA taxonomy to analyze simulation records: Insights and suggestions

It is evident that diverse human reliability analysis (HRA) methods are effective for enhancing the safety of socio-technical systems through identifying the most vulnerable tasks to human errors with the associated human error probabilities. This means that reliable human performance data is an important factor affecting HRA quality. Therefore, many researchers have developed technical underpinnings (such as guidelines and taxonomies) that specify what and how HRA data can be collected from simulator experiments. Here, SACADA (Scenario Authoring, Characterization, and Debriefing Application) taxonomy recently developed by US NRC (Nuclear Regulatory Commission) is worth emphasizing, because it is constructed on the basis of a cognitive model (i.e., a top-down approach) while most of the technical underpinnings are developed by a bottom-up approach (i.e., the intensive review of existing literature). For this reason, in this study, the SACADA taxonomy is used to analyze several audio-visual records collected from the full scope simulators of nuclear power plants in the Republic of Korea. The results indicate that the SACADA taxonomy is useful to collect operator performance data in simulator training for HRA. Certain human performance information that can be provided by SACADA data provided are difficult to be covered by the bottom-up approach.

Application of a process mining technique to identifying information navigation characteristics of human operators working in a digital main control room – feasibility study

Abstract It is well known that one of the major contributors to the significant events of socio-technical systems is a human error. Accordingly, various kinds of human reliability analysis techniques have been proposed for many decades, which allow us to properly estimate human error probabilities. However, one of the urgent challenges is to quantify human error probabilities when required tasks are conducted in a control room equipped with up-to-date digital equipment (e.g., digital main control room) because existing techniques are mainly focusing on an analog control room. In this regard, it is necessary to carefully understand how professional operators in the digital control room carry out the required tasks. Therefore, in this study, information navigation characteristics (i.e., information gathering behaviors observable from professional operators who need to conduct the required tasks by using given digital equipment) were analyzed based on process mining techniques. To this end, event logs were collected from the professional operators of domestic nuclear power plants who were exposed to simulated off-normal conditions in a digital main control room. As a result, it seems that process mining techniques are useful for extracting crucial information needed for the human reliability analysis in the digital main control room.

Use of a Big Data Mining Technique to Extract Relative Importance of Performance Shaping Factors from Event Investigation Reports

In this study, the relative importance of significant performance shaping factors (PSFs), which is critical for estimating the human error probability (HEP) of a given task environment is extracted from event investigation reports of domestic nuclear power plants (NPPs). Each event was caused by one or more human performance related problems (i.e., human errors), and its investigation report includes detailed information describing why the corresponding event has occurred. Based on 10 event reports, 47,220 data records were identified, which represent the task environment of 11 human errors in terms of significant PSFs. After that, the relative importance of the associated PSFs was analyzed by using a CART (Classification and Regression Tree) method that is one of the representative techniques to scrutinize the characteristics of big data.

A novel speech-act coding scheme to visualize the intention of crew communications to cope with simulated off-normal conditions of nuclear power plants

Abstract Many researchers have commonly pointed out that the characteristics of crew communications is one of the most important factors affecting the operation safety of complicated process control systems. From this concern, a couple of speech-act coding schemes were developed from the point of view of ‘what was done by crew members?’ In this study, a novel speech-act coding scheme was developed, which allows us to see the contents crew communication from a different angle – ‘what was the communication intention of crew members?’ To this end, the communication contents uttered by MCR operators who were faced with various kinds of simulated off-normal conditions were collected from the full-scope simulator of domestic NPPs. Then, the novel speech-act coding scheme was developed by involving additional yardsticks (such as Means, Acceptance criteria, and Constraints), which are useful for elucidating the nature of communications from a task description perspective. As a result, the novel speech-act coding scheme was proposed, which consists of 45 behavioral task categories and the associated definitions. Although the novel speech-act coding scheme proposed in this study is still a preliminary version, this would be a good starting point to enhance the quality of crew communications through visualizing their communication intentions.

Design and Implementation of HuREX Analysis Supporting Interface for HRA Data Extraction

Abstract It is well-known that one of the main causes of problems affecting social-technical systems, including nuclear power plants (NPPs), is human error. For this reason, reducing human error through human reliability analysis (HRA) is important. Furthermore, sufficient and reliable human performance data collection is a prerequisite for ensuring the safety of NPPs. The Korea Atomic Energy Research Institute developed the Human Reliability data EXtraction (HuREX) framework to provide a standard guideline for the collection and analysis of human performance data from operators in main control rooms based on simulator training records of NPPs. To do this, the development of a computerized software interface is required to collect simulator-based human performance data systematically and then to enter/analyze/quantify the various forms of data obtained from the simulator. In addition, a HRA database is needed for the effective management of the data generated during this process. In this research, we develop an interface that supports HuREX analysis so that HRA practitioners can conduct more effective HRA data analyses by integrating various types of raw data (e.g., audiovisual records, plant parameters, and operator action logs) collected from simulators. In addition, we expand the OPERA database to store a standardized data structure for more effective analyses of unsafe acts via the HuREX data analyzer and the HuREX video analyzer.