Development and Validation of Active Performance Indicators of Electrical Safety Using Bow Tie and Bayesian Network Techniques (Case Study: Oil and Gas Industries Construction Projects)



Karimi A¹ ; Falahati M² ; Zokaei M² ; Biabani A² ; Torshizi YF³ ; Salehi F⁴ Show Affiliations
Source: Iran Occupational Health Published:2019

Abstract

Background and aims: With the developing use of electricity in all aspects of human life, electricity accidents have also increased. Electricity is the most important type of energy used in workplace and has a direct relationship between economic development and its consumption. Every year, more than 5,000 deaths from electric shock occur in the US construction industry, accounting for about 20% of the deaths from accidents. The incidence of electrical accidents is 25 times greater than the incidence of falls. Therefore, it is essential to take the necessary measures to prevent such incidents. One of the main components of the prevention policy, is the safety performance assessment of the organization's or industries by using appropriate performance indicators with related operations. Safety performance indicators are divided into two main categories: the leading and the lagging indicators. The leading indicators show the organization's actions in predicting and preventing incidents before they occur, while the lagging indicators show the organization's performance after the occurrence of the incidents. In many activities, such as construction phase due to rapid changes in the nature of activities and subsequent changes in the level of operational risk, it is essential that safety performance indicators become more responsive and sensitive to changes in the safety level of operations. The aim of this study is to develop and validate active indicators for assessing the safety performance of electricity in the phase of construction of oil and gas refineries using Bow Tie and Network Bayesian techniques. Methods: This is a descriptive-analytical study carried out in the phase of construction of oil and gas refineries. In the first phase, the construction operations were identified by studying the WBS refinery project. Work Breakdown Structure is a graphical network or graphical structure for showing a product or service production method, including hardware, software, services, and other tasks of the organization or company. At this stage, the main activities of the construction phase were identified using the WBS project. In this study, hazardous construction phases were identified and classified by PHA. Considering that determining the causal network and probability of a failure is one of the important factors of risk assessment and determination of safety performance indicators, in the third stage, the relationship between the causes of the incident and the occurrence of the underlying causeswere researched using expert panel. The Bayesian network is an important tool for determining causal relationships and calculating the probability of accidents. The Bayesian network is a graphical probability model that shows a set of random variables and conditional dependencies between them. After determining the probability of occurrence of the root, intermediate and direct causes of the electric shock, the Bow-tie risk assessment technique used to determine the control measures related to each of the causes. By surveying the relationship between the causes of the electrical incident, the active indicators related to the field of electrical safety were identified and in the last step, validation of the indicators over a 6-month period done by comparing the results of measuring the active indicators and the frequency of incidents recorded through Regression correlation test was performed. Results: In this study, six disciplines including electrical, instrumentation, piping, tanks, communications and buildings, and 18 electrical hazardous activities identified. Based on the results of the WBS review and the preliminary risk analysis of identified activities, the causal network electric shock accounted for the expert's opinion. The probability of occurrence of electric shock was 0.053 calculated using the method of BN. The electric shock caused by a collision with the under-ground cable (A) and the contact of the scaffolding pipe with the power cable (E) the largest share of the potential for electric shock. By using technique Bow tie, control measures including a RCD, earthing system, inspection and supervision, training, permit to work at drilling, scaffolding, work to height and maintenance operations and Ricket Fire, are set to reduce the probability of occurrence of events. 11 active performance indicators include drilling performed with permission to do work, boards with RCD and Earthing, earthing generators, generators with installed safety instructions, generators with fire extinguisher, scaffolds authorized to do work, Number of electrical safety inspection, cranes with electrical insulation cabin, operators trained and earthing Conex. In order to determine the correlation between active indicators and incident frequency indicators, information on five contractor companies collected and measured during six months. Then the linear regression model used to determine the statistical relation. The R-square value of the incident frequency with the active performance indicators of A, B, C, D and E companies is 0.996, 0.495, 0.998, 0.863 and 0.797, respectively, with the highest linear regression, respectively C, E, A, B and D companies. The p-value in A, B and D companies is greater than 0.05 and in C and E companies is less than 0.05. Therefore, these results indicate that the correlation between incidents frequency as dependent variables and active performance indicators as Independent variables in A, B and D companies are not significant, but the correlation between C and E companies is significant. The R-square value of the incident frequency with the active indicators of the total project was 0.819 and P-value was less than 0.05. Therefore, these results indicate that the correlation between incident frequency as an associated variable and active indicators as independent variables in the whole project is significant. Conclusion: This study was performed to determine the safety performance active indicators of electrical of construction phase of oil and gas refineries. In previous studies, such as the Podgorski (2015) and Flahati (2017), the leading performance indicators developed according to ILO-OHS-2001 and OHSAS 18001 management system components. One of the main objectives of the development of safety performance indicators in construction projects is to rank and compare the safety performance of the contractors in the project. Therefore, it is necessary to consider all aspects of the performance management system in order to make a proper judgment of the existing situation. The performance indicators presented in the Podgorski study (2015) emphasize only the components of the safety and health management system. While the model presented in the present study categorizes the indicators according to the definition of Hinze (2013) into two types of active and passive indicator. Given the rapid changes in operating conditions in the construction phase, leading performance indicators should be able to detect rapid changes in the level of safety of activities. Abdelhamid TS et al. (2000) declared the failure to identify unsafe conditions as one of the root causes of construction projects incidents, so active performance indicators should be able to measure the safety of construction operations in short term periods. The strength of this study is to use the Bayesian Network to determine the cause of the incident. The superiority of Bayesian network in assessing the risk and determining the route of the events is that nodes can considered dependent. Therefore, it is easier to determine the relationships between the different levels of the causes of the incident and the estimation of the probability of occurrence of accidents is more accurate. One of the main constraints of previous studies was the lack of validation of leading performance indicators. In this study, were validated the developed indicators. Correlation between accident frequency as dependent variable and active performance indicators as independent variables is high in all five companies, but in firms A, B and D this correlation is not significant. One of constraints of this study is the short duration of measurement of indicators in a six-month period, so the results will be more accurate with increasing the measurement time of the indicators. This study indicates that determining the leading indicators in addition to the components of the safety management system should be based on the type of operations and identified workplace hazards. The leading performance indicators of the safety management system components are mostly passive and cannot show the changes in the level of safety of the workplace in a short time. Because construction operations are inherently high-risk and safety-related changes are taking place rapidly, it is therefore necessary to adopt active indicators as complementary passive indicators. Considering the many variables that affect the occurrence of major industrial accidents, determining the causal relationships between these variables is complex. Therefore, using high reliability methods such as Bayesian network increases the reliability of the active performance indicators derived from the network causing accidents. © 2019 Iran University of Medical Sciences. All rights reserved.