Evaluation of human error using human error assessment and reduction technique (HEART) based on fuzzy logic (Case study: gas power plant

Introduction Human error is defined by Reason (1990) as the behavior of a person who did not intend to do so. It is not favorable by regulation or by observers or that the behavior causes a task or system to become beyond the accepted limits. In complex systems such as air transport, rail transport, medical services and road traffic accidents, human error has been identified as an effective factor in most accidents. Studies show that more than 70% of airline accidents and about 70% of rail accidents were due to human error. There are many techniques to evaluate human error. In 2009, Health and Safety Authority of the United Kingdom introduced 72 human reliability assessment techniques. It has identified 35 techniques as major hazard identification techniques related to safety, health and environment. In this study, the HEART technique is one of the 35 techniques mentioned and introduced as the first generation techniques with public access. The HEART is an accurate technique that quantitatively demonstrates human error, and it requires little training and is also a medium to run. The availability of publicly available technical resources and details has led to the use of the HEART technique as one of the valid techniques among the various techniques available to identify human errors in power plant utilization. In another study, Defense Sciences Institute examined techniques related to human factors. The results of this study identified 11 valid techniques for human error detection. Human error detection techniques are used to predict or analyze the potential for errors caused by a collision with the system or device under investigation. Kirwan described nine techniques for assessing human reliability in 1997. He reported that between all these 9 techniques, HEART, THERP and JHEDI techniques work well. In the next study, HEART, THERP and JHEDI techniques were studied. The accuracy rate of the HEART technique among the 30 human error assessors who used the HEART technique was 76.67%. According to a study by Kirwan and colleagues (1995), the trend of the last few years has introduced the transition from quantitative human error probability techniques such as THERP and SLIM techniques to more effective and efficient HEART resources. In the studies of Castiglia and colleagues (2012 and 2014), the modified HEART technique based on the fuzzy set concept was used to investigate the probability of wrong actions. Also Li Peng-cheng and colleagues (2010) used a fuzzy logic approach to study the importance of human error risk. This paper deals with the HEART technique combined with fuzzy logic to assess human error in a plant operation unit with a predictive approach of human error occurrence and sensitivity of control rooms in power plants.   Methodology Step One: Hierarchical task analysis (HTA): The HTA technique was developed in 1971 by Annette and colleagues to analyze complex tasks such as tasks in chemical processes and power plants. In this technique, tasks are broken into a hierarchical set of tasks and sub-tasks. The HTA technique was first introduced for process industries and power plants. To analyze the activities, all operating documentation including procedures, instructions, forms and checklists were studied. And, according to the technical experts' documentation, the activities were broken down to the last necessary stage. Step Two: Human Error Assessment and Reduction Technique (HEART): The HEART technique was developed by Williams in 1986. It is one of the human error detection techniques that attempts to predict and quantify the probability of human error. The HEART technique has been developed for use in nuclear power plants and chemical process industries. This technique is a reliable method for comparing human error probability (HEP). In this way, the consequences of each activity are also evaluated. One of the aspects of the HEART technique to reduce resource usage is that it only deals with errors that have a significant impact on the system under investigation. In the United Kingdom, the HEART technique was used to assess the risk of the Sizewell B nuclear power plant as well as the risk assessment of the Magnox reactor stations of England and advanced Gas Cold.   The steps to implement this technique are as follows: Steps One - Identify the task or scenario under analysis. Step Two - Perform the HTA for the task or scenario under analysis. Step Three - Perform the HEART Screening Process. Step Four - Classification of Inaccuracy of Task: Uncertainty Determination Using HEART General Classifications Based on Table 1. Step Five - Identify Error Producing Conditions (EPCs): Negative factors that increase the potential for human error occurrence are called error producing conditions. At this stage, task-related EPCs are identified. The error producing conditions are presented in Table 2 (4 and 20). Step Six - Assess the Impact Ratio: Determine the estimated impact ratio of each EPC by assigning a score between zero and one for each EPC. Step Seven - Quantifying Potential Human Error. At this stage, the potential human error is calculated for each identified task. The relevant formula for calculating potential human error is as follows.   NHEP: Nominal Human Error Probability in Step 4 (extracted from Table 1) EPCi: The maximum amount of impact (weighting factor) of each EPC obtained in Step 5 (extracted from Table 2) APi: The estimated impact ratio of each of the identified EPCs obtained in Step 6 HEPF: The Ultimate Human Error Probability   Step Eight - Control measures: This step involves identifying and suggesting control measures for known errors. Although the HEART technique offers some general control measures, the analyst may need to provide more specific control measures based on the nature of the error and the system under analysis. The control measures provided by the HEART technique are general and are not system-specific.                                     Table 1. Nine general duty groups of the HEART technique and Nominal Potential Human Error   Public duty Nominal Potential of Human Error A: Completely unfamiliar, doing it quickly without realizing the potential consequences 0.55 B: Move or restore the system to a new or original state with an unsupervised action or procedure 0.26 C: The complex task requires a high level of understanding and skill 0.16 D: Doing relatively simple work quickly or with little attention 0.09 E: Routine task, fast, accomplished, highly accomplished, requiring relatively low skill level 0.02 F: Move or restore a system to a new or original state following the executable method with little checking 0.003 G : A thoroughly familiar, well-designed, highly experienced, routine task that occurs several times every hour is performed to the highest possible standards, with a highly motivated, highly trained and experienced individual, fully aware of the consequences of failure, with time to correct potential errors but without the use of significant aids 0.0004 H:  Correct response to the system command even when an automated monitoring system provides a complete interpretation of the system status. 0.00002 I:  Miscellaneous task for which no description can be found. 0.03   Table 2. HEART Technique EPCs Row Error Producing Conditions (EPC) The maximum value 1 Unfamiliarity 17 2 Lack of time 11 3 Low signal-to-noise ratio (high noise) 10 4 Ability to ignore attributes 9 5 Functional and spatial mismatch 8 6 Mismatch Model 8 7 Immutability 8 8 High load on the channel 6 9 Technique not learned 6 10 Requires specific knowledge transfer 5.5 11 Functional ambiguity 5 12 Misconception of risk 4 13 Weak and vague feedback 4 14 Incomplete / delayed feedback 4 15 Operator inexperience 3 16 Trivial information 3 17 Not checking enough 3 18 Conflict of goals 2.5 19 Lack of diversity 2 20 Educational mismatch 2 21 Dangerous motivations 2 22 Lack of mental and physical fitness 1.8 23 Unreliable instrumentation 1.6 24 Absolute judgment 1.6 25 Assign tasks indefinitely 1.6 26 There was no progress tracking 1.4   Fuzzy HEART Based fuzzy HEART technique, the estimated impact ratio (APi) of each EPC was collected based on the opinions of 4 experts. By using the fuzzy logic toolbox of MATLAB software, the estimated fuzzy impact ratio (APlvi) obtained using the following formula of fuzzy HEART calculations. The points of the membership function of each of the VL, L, M, H, and VH levels will be calculated using the following formulas.                                                                                                          The reason why the fuzzy logic is applied in the HEART technique is fuzzy logic is an efficient tool for solving problems in uncertainty, so it is also used in this research. Generally, the error producing conditions of each activity were identified and extracted based on the opinions of 4 technical experts due to their thorough familiarity with their work process and high work experience according to Table 2. To calculate the impact ratio of each of the identified EPCs using fuzzy logic collected the opinions of 4 experienced experts. By using the MATLAB software, defined fuzzy logic toolbox membership functions of input and output values and recorded the required database rules. In next step, the impact ratio of each EPC calculated by entering the collected data. Then, the probability of a final human error was calculated for each task.   Results Based on hierarchical analysis of activities obtained 1150 rows of tasks and sub-tasks that Includes all V94.2 gas plant operation activities. Screening activities and extracting the most important activities of the operation group considering the first and second steps of the HEART technique for screening tasks from level two, three and four HTA tasks (Total of 1150 tasks and sub-tasks), The main tasks were extracted according to the experts. In total, 13 tasks and 119 sub-tasks were achieved.  In table 3 is shown. Implementation of fuzzy HEART technique: According to Table 1, each of the tasks was initially assigned to one of the nine general task groups of the HEART technique. The nominal human error probability for each activity was obtained. This summary is summarized in Table 3. Table 3. Number of Member Tasks of Each Task Force in HEART Technique Number Public duty 0 A: Completely unfamiliar, doing it quickly without having a real idea of the likely consequences 10 B: Restore the system to the original state with an unsupervised action or procedure 13 C: The complex task requires a high level of understanding and skill 49 D: Doing relatively simple work quickly or with little attention 16 E: Routine, fast-performing, highly accomplished task requiring relatively low skill level 12 F: Restore the system to a new or original state following the executable method 11 G: Fully familiar, well designed, highly experienced, routine task 8 H: Correct response to the system command 0 I:  Other duties or Miscellaneous tasks 119 Total   Among the activities, "Investigation of the openness of the transducer on the Incoming Panel (BBE bus)", obtained the highest probability of human error. The repetition of the error generating conditions shows in Table 4. Table 4. Statistical report of the Repetition Rate of Error Producing Conditions between Different Tasks Percent Number error producing conditions 98.3 117 Not checking enough 77.3 92 Operator inexperience 39.5 47 Low morale and conscience 34.5 41 Lack of time 29.4 35 Unfamiliarity 23.5 28 Misconception of risk 23.5 28 Weak and vague feedback 4.2 5 Lack of mental and physical fitness 3.4 4 Lack of diversity 1.7 2 Functional and Distance mismatch 0.8 1 Immutability   Discussion There were many error producing conditions for each of the tasks reported in Table 4. According to Table 4, EPCs of "insufficient checking", "operator inexperience", "low morale", "lack of time", "lack of awareness", "misconception of risk" and "poor and ambiguous feedback" were most of the issues related to human error, respectively. "Insufficient checking" in more than 98% of tasks was one of the error producing conditions. This may be due to lack of time, impatience and fatigue, lack of training and high workload. According to studies conducted by Kirwan (1996), Mortazavi (2010), Jahangiri (2004), Ghalenoei (2006), Mohammad Fam (1996) and Zarra Nezhad (2013), work guidelines, training, experience and time are the main factors affecting occurrence of human error. The reason may be the lack of proper guidelines, lack of adequate training. In the study of Noroozi and colleagues (2012), two activities of "open value, pump filling and leak test" and "drainage lines" have the highest probability of human error due to "lack of awareness" and "lack of time". Also, these two factors have been among the thematic and commonly used EPCs of this study. The proper time interval can be considered as a factor to control this error. In the study conducted by Castiglia and colleagues (2014), the EPCs “insufficient checking”, “risk misconception”, “information overload”, “knowledge transfers”, “poor and ambiguous feedback” and “functional ambiguity”, respectively, were the most frequent. Three of these EPCs are identical to the repeated EPCs of this study, and the differences are due to the conditions and nature of the work and the differences in the industrial structure, knowledge and experience of the workforce. In the 19th century, the Italian economist Wilfredo Pareto observed that 80% of the usable land belonged to 20% of the population. Pareto received the same distribution in other economic and natural processes. As a general rule, by formulating the finding "in any arbitrary set of elements that strive to achieve a goal, a small subset (with a small number) has the greatest effect". In this study, a total of 24 tasks have a high likelihood of final human error (HEPF), which is 20% of the total tasks evaluated. In 87% of these tasks, lack of awareness is considered to be one of the EPCs. In the study conducted by Ariefiani and his colleagues who investigated the human error using fuzzy HEART method in grinding operations of a construction company, accidents are reported as a result of human error with 66.67%. The highest human error rate was estimated to be HEP = 0.71324. The technique used and the results were the same as the present study. Another study in a door manufacturing industry performed fault analysis with the SHERPA and HEART methods. According to a HEART study, the total amount of HEP is 0.4986. Activities on "Checking the openness of the transducer on the BBE bus Incoming panel", Re-seal the 8-inch valve and check for no leakage of the flanges, Once the cyclone filter pressure reaches its final level, "Ensuring the correct arrangement of the fuel tank valves", "Ensure that the inlet and outlet cartridge filters are closed", "Compressor visit (Compressor outlet diffuser enclosure visit below)", were allocated highest HEPFs, respectively. 22 tasks out of the 24 tasks have a higher probability of human error, training and justification of operator or shift engineer than suggested control strategies. Based on the results, the HEART technique works well based on its structure and provides reasonable results. So that only 2 tasks have a very high probability of ultimate human error. The human error associated with these two tasks has occurred at least in one similar power plant. Of course, it is clear that the experience of the experts in the method involved in this research is effective on the outputs. Therefore, experts with higher education and experience were used in each step of the research to reduce the impact of individual judgment and increase the quality of outputs. Conclusion Human errors must be estimated and evaluated in any dangerous process operation in order to avoid accidents and consequences. The present study offers a framework to simply quantify the human errors in different operator tasks in process industries. Fuzzy logic concepts are useful in conventional HEART technique by considering abstract linguistic expressions and suitable weightage is allocated using expert elicitation method with consensus to analyze quantitative error probability. CONFLICT OF INTEREST The authors declare that there are no conflicts of interest regarding the publication of this manuscript.