Theoretical Issues in Ergonomics Science, 2015 http://dx.doi.org/10.1080/1463922X.2015.1028507 Managing risk dormancy in multi-team work: application of time-dependent success-and-safety assurance methodology Emad Farag *†, Dov Ingman and Ephraim Suhir Downloaded by [University of Haifa Library] at 02:29 26 April 2015 Faculty of Industrial Engineering and Management, Technion Israel Institute of Technology, Haifa 32000, Israel (Received 13 October 2014; accepted 9 March 2015) The success and safety of many of today’s industrial activities, such as constructing power plants, transmission lines and civil engineering objects, is often influenced by situations, when successful and safe work completion is associated with the implementation of various more or less complex multi-team (MT) layouts. Multi-team effort is characterised by the presence and interaction of numerous, not necessarily concurrent time- and space-constrained interfering activities, as well as by dormant risks. Such risks might threaten the fulfilment of the general task carried out by the main team and/or by other teams on the construction site. This analysis addresses the role of the dormant risks during the fulfilment of a non-simultaneous MT work. The objective of the analysis is to suggest an effective and physically meaningful probabilistic predictive model. The model is aimed at the understanding, quantification and effectively managing the dynamics of the system of interest. The emphasis is on the role of possible dormancies. The study is an extension of the authors’ earlier research on spatial and time dimensions in the addressed problem. The study extends the risk management approach to a holistic level. Relevance to human factors/ergonomics theory This article characterises the risk dormancy phenomenon to be considered as a proper way of taking into account multi-team (MT) aspects in the occupational safety research. Furthermore, a holistic perception of MT functional complexity allows for a generalised view of MT mutual interaction instead of focusing in the single team behaviour. Keywords: risk dormancy; team work; multi team; time-dependent probability; expected RD time 1. Introduction Modern infrastructure and industrial activities are typically executed by several different on-site teams: ‘most human work is performed by teams rather than by individuals’ (Sasoua and Reason 1999). Each team performs distinct and designated activities over time. At the same time, today’s technologies require better understanding and increasingly higher quality than in the past. Sophisticated work methods, tools and equipment have been developed for, and became available to, the workers in multi-team (MT) systems. Such methods are crucial to achieve the optimum and safe outcome of a particular project. Tight schedules are implemented today to meet customers’ requirements, and to *Corresponding author. Email: [email protected] †Present address: Department of Electrical Engineering, University of California, Santa Cruz, CA, USA. Ó 2015 Taylor & Francis Downloaded by [University of Haifa Library] at 02:29 26 April 2015 2 E. Farag et al. ensure the demands for increased productivity. This imposes additional pressure on workers. Handling of hazardous materials and energy sources, such as electricity or radiation, requires the use of highly qualified labour, as well as customised safety procedures and equipment. Dynamic physical conditions, such as noise, vibrations and working at heights, contribute also to the likelihood of hazardous situations. Teams often operate independently, with no explicit functional linkage between them. In other cases, more or less close professional collaboration is required to perform a particular task. For example, repair work in an electrical utility typically requires involvement of several teams. One team of electricians disconnects the power, another team repairs the damage and a third team is deployed outside the secluded site, often on a standby basis, to assist, if necessary, the workers inside the worksite. In this example, three different teams perform various aspects of the work aimed at a particular mutual goal. A mistake, error or a failure in one team’s actions affects other teams involved in sequential large-scale activity, and has a potential to create a hazard. The hazard might remain dormant for some time, but eventually can become or generate a risk. This risk might have an immediate impact on the team member who caused it, and/or threaten another team continuing the job at the given site. This scenario can be identified as risk dormancy (RD). When occurring in a MT situation, it becomes a MT risk dormancy (MTRD). It is this type of RD that is the main concern and the main subject of this analysis. Several researchers have addressed the MTRD lately. Mitropoulos, Howell, and Abdelhamid (2005) stated that ‘errors by one crew may create unpredictable conditions for a following crew’ and suggested that ‘future research should focus on better understanding the effect of task unpredictability and on developing error management strategies’. Reason (2000) wrote ‘Different actors’ decisions and actions can produce latent conditions or pathogens in a system. These might lie dormant for a time until they combine with local circumstances and active failure and penetrate the system’s many layers of defences, and an accident occurs’. Despite the recognised importance of the MTRD, there are no studies that suggest effective solutions to the MTRD problems. The existing studies suggesting various safety models employ quite a few of diverse approaches. Some models focus on actions or processes and examine the time and space of the occurred accidents that led to personal injury or damage to some assets. These are scilicet (s.c.) active failures. Other models are system oriented, e.g. organisation models related to the management policy, actions and decision-making (Rasmussen, Pejtersen, and Goodstein 1994, 149). Interdisciplinary models ‘focus on cause-effect relationships close in time and space to the accident sequences’ (Reason 1997; La Coze 2005). Akinci et al. (2002) examined the time–space management aspect of accident prevention. He indicated that ‘lack of management of activity space requirements during planning and scheduling results in timespace conflicts in which an activity’s space requirements interfere with another activity’s space requirements or work-in-place’ (Rosenfeld et al. 2006). Rozenfeld, Sacks, and Rosenfeld (2010) addressed this situation by expanding the safety model to MT problems that involve mutual risk exposure of two or more teams sharing the same timespace domain. In the analysis that follows a rather general holistic approach is used to address dormant risks encountered during consecutive MT work activities. This approach treats the problems of interest from a rather general point of view, regardless of a specific nature of a particular risk or a possible outcome. Our approach provides a probabilistic assessment of the management’s dynamic response at the organisational level. Here are several typical examples. Example 1. A scaffold is required for certain tasks performed on a public utility (PU) construction site by teams working at heights, such as, say, plasterers and painters Figure 1. Theoretical Issues in Ergonomics Science Bricklayer team, Plasterer team 3 Painter team Scaffold builder team Plumber Project Schedule (a) Downloaded by [University of Haifa Library] at 02:29 26 April 2015 Scaffold builder team Bricklayer team, Plasterer team Plumber Painter team Project Schedule (b) Figure 1. Illustration of the scaffold example (a) Prior of lean scheduling. (b) 1b lean scheduling timespace dependent model (Rozenfeld, Sacks, and Rosenfeld 2010). A scaffold building team erects the scaffold, while a plumbing team is scheduled to subsequently lay sewer pipes. In the meantime, the scaffold is being used by other teams. In such a situation, the timespace sharing is implemented as illustrated in Figure 1(a). In the safety assessment of the project in question risk exposure in team activities performed with timespace overlapping is not considered. All the teams the plasterers, the painters and the plumbers worked on the site simultaneously. The Lean Scheduling Time and Space Dependent Model (Rozenfeld, Sacks, and Rosenfeld 2010) illustrates the time segregation. This means that the plumbing team’s work is rescheduled to avoid the risk of dealing with falling objects or tools from the plastering or bricklaying teams. In space segregation, on the other hand, the plumbers would be assigned to work at the other side of the site, where no teams would be working above them. The risk posed by the scaffolding team is eliminated, as illustrated in Figure 1(b). The planner’s instructions indicate, however, that the plumbing team must lay the pipes in trenches at the foot of the building, where the scaffold is assembled. Although the plastering team could be temporarily repositioned and the excavation could be rescheduled, this still might destabilise the scaffold and increase the probability of its collapse at a later time (Figure 2). Rozenfeld’s timespace-dependent model does not address this aspect of scaffold destabilisation risk. The risk leads to an additional risk, namely to the RD. In this example, other scaffold users, such as the painting team, are exposed to an underestimated risk, which, however, has been identified beforehand. This example illustrates the RD phenomenon, i.e. a risk that is dormant, while awaiting for other teams that might use the hazardous scaffold. Figure 2 emphasises the progress that could be achieved by developing tools and methods to deal with such underestimated or misidentified risks that stem from the dynamic changing of the site (in this case, the excavation). Example 2. Power grid works are inherently created serially. A utility lineman team replaces an insulator on a high-voltage power line after obtaining permission from the responsible electrician. The electrician operates according to a written checklist issued by 4 E. Farag et al. Underestimation of the identified RD in MT work or misidentification of RD Scaffold building Excavation for sewer Plastering team Painting Project schedule Downloaded by [University of Haifa Library] at 02:29 26 April 2015 Figure 2. RD exposure in MT example risk of scaffold collapse, threat to painter team. the engineering department. The professional teams work in a serial manner with respect to both time and space: first, the engineering department writes the instruction checklist; then the responsible electrician shuts down the power at a circuit box mounted near the work site (along the power line) and linemen replaces the insulator further down the line (not even necessarily at the same site). In such situations, several teams are active at the site, and communication and coordination of their interactions might be quite complex. Any error in these activities could create RD, thereby increasing the likelihood of an electrical shock accident. Examples of this type of error are failure to check for current, misidentification of the appropriate line connector and/or the specification of the wrong transformer or pole number. The above examples (scaffold collapse or electrician’s error) represent dormant risks caused by MT activities that have no time or space overlapping. This means that accidents caused by the MTRD activities regardless of their simultaneity and space have not been addressed. MT activities create dynamic work sites (environments) with constant changes depending on the needs for the complex system, in/for which the work is being performed. Such complex systems require a tight control, i.e. appropriate risk management, which should be the main component of a safety management system (SMS). The term ‘risk management’ includes the notion of mitigating risks to an adequate and achievable level acceptable by the organisation. This can be done by the application of two major methods: (1) appropriate and effective risk control and (2) the use of the most suitable accident causation models. The obvious challenge in the assurance of the occupational safety is the development of the ability to effectively control problematic situations, identify hazards and assess and control risks. The actual down-to-earth and practical on-site work is more complicated, however, than the models that attempt to predict and simulate the effort. Proactive approaches, such as risk assessment and control, can never be entirely accurate, complete and successful when applied alone. ‘Effective risk management depends crucially on establishing a reporting culture. Without a detailed analysis of mishaps, incidents, near misses and “free lessons”, we have no way of uncovering recurrent error traps or of knowing where the edge is until we fall over it’ (Reason 2000). Risk management should be therefore implemented both proactively and reactively, and its complete success Theoretical Issues in Ergonomics Science 5 Downloaded by [University of Haifa Library] at 02:29 26 April 2015 necessitates reactive approaches such as accident investigation/causation. A description of the process of investigating accidents in the context of the concept presented in this paper, namely, the underlying cause RD, is set forth below. The general strategy of pursuing risk management approach in the problem in question includes the following major items. 1.1. Hazard identification Before quantifying the probability of failure, hazards related to the system operation must be identified. Several identification techniques are available (Carter and Smith 2006), some of which are based on brainstorming among people familiar with the installations at a work site, while other techniques are of a more systematic nature. According to Cuny and Lejeune (2003), recognising hazards can be a rather complicated task. Many different kinds of uncertainty factors contribute to the challenge of recognising hazards and a clear-cut determination of the source and level of the risk for each hazard is next to impossible. Furthermore, many observations are needed to accurately estimate the likelihood of an accident, particularly, if it is of rare occurrence. One of the paradigms in the Accident Root Causes Tracing Model (Abdelhamid and Everett 2000) reveals that workers, more often than not, fail to identify hazardous situations. This leads to the conclusion that the hazard identification process only partially covers the hazards that should be identified on site. Multi-team hazards are even more difficult to identify. 1.2. Risk assessment Hazards become a problem only when they could possibly result in an accident whose occurrence is preceded by a sequence of events that may cause a hazardous situation. After a hazard is identified, all possible sequences of events that can be triggered by that hazard must be studied and checked to determine whether or not they might lead to an accident. With the relevant scenarios in hand, it is possible to calculate the two elements of a risk: the probability of the events occurring, and their consequences. Reason (1997) presents three models for safety management: the Pearson model, the engineering model and the organisational model. Each of these models has a different perspective on human error. ‘Workplaces and organisations are easier to manage than the minds of individual workers. You cannot change the human condition, but you can change the conditions under which people work. In short, the solutions to most human performance problems are technical rather than psychological’. This concept can be better understood by considering the work of Papadopoulos et al. (2009), who concluded that ‘risk assessment must be conducted for each task and for each worker. This risk assessment must consider all hazards and their interactions and must be revised when changes occur’. In addition they wrote ‘However, frequent changes regarding workforce, working hours and working conditions, as well as time pressure, result in insufficient time for conducting a complete and effective risk assessment, determining training needs, setting up, applying and monitoring the corresponding OSH measures. Furthermore, the methodological tools used in risk assessment up to now are not sufficient for this complex situation’. In an earlier paper on this subject, Drivas and Papadopoulos (2004) pointed out that risk assessment needs to consider all hazards and their collaborations and must be reviewed when changes occur. Risk assessment will be even more difficult to identify for risks arising from MT work, which is characterised by difficulty in identifying or the lack of the researcher capacity to evaluate risks. 6 E. Farag et al. 1.3. Risk control Downloaded by [University of Haifa Library] at 02:29 26 April 2015 Risk management is basically a control problem (Rasmussen and Svedung 2000). The review focuses on the most suitable methods of risk control. (a) Control hierarchy is achieved by applying four main levels of action: (1) the hazard is eliminated; (2) a physical barrier is erected between the hazard and the performer (worker); (3) personal protective equipment is used; and (4) workers comply with written safety instructions. (b) Root cause analysis addresses accident causation according to four basic categories: management factors (safety and risk control), intermediate factors (procedures, work design, training), performance (behavioural and technical) factors and external (environmental) factors. (c) Comparative analysis consists of measurements for risk prevention using the following four dimensions: effectiveness, applicability, efficiency and influence (Griffel 1999). Moreover, since most serious accidents are apparently caused by the operation of hazardous systems outside the design envelope, the basic challenge in the development of improved risk management strategies is essentially to ensure improved interaction between the decision-making and planning strategies at the various levels of the organisation (Rasmussen and Svedung 2000). Thus, despite the currently implemented risk management strategy, the control of MTRD appears to be unsolved yet. 1.4. Accident causation Accident investigation/causation analyses enable one to better understand the factors and processes leading to accidents. Our analysis that follows explores further accident causation or aggravating factors, i.e. RD in MT work. The primary objective of this paper is to develop a holistic safety model, in which both management and labour respond to various dormant MT risk situations. Management is responsible for making decisions concerning the handling of such risks and accident causation in accordance with a SMS. Disorder might result in a potential for unsafe actions. Such actions are characterised by the following major attributes. Deviation of a process from its planned time schedule, requiring corrective action to remove the source of non-conformity in order to prevent recurrence. The corrective action is aimed at ensuring that the existing potentially hazardous situations do not lead to accidents. Time lags in the sequenced work of the various teams require communication and mutual reporting. Continuing random changes to the physical sites, thereby necessitating continuous risk assessment. Uncontrolled multiple degrees of freedom, instead of a tight and narrow path to a successful outcome. The lack of quality methodology principles being applied to monitor safe performance, so as to reduce or even prevent accidents, especially those with casualties. These attributes can create hazards that might be difficult to identify properly. Theoretical Issues in Ergonomics Science 7 Table 1. Number of employees by division. Division Number of employees North district South district Logistic Generation Construction 800 1300 700 2100 1700 Downloaded by [University of Haifa Library] at 02:29 26 April 2015 Two major problems with MT risk management have been identified. (1) RD identified in MT work is often underestimated. For example, after excavating trenches intended for power or communication lines or for drainage pits, the installation of the cables of pipes is often delayed and the trenches and pits are marked using yellow caution tape only. Despite the identification of an open trench or a pit as a hazard, such trenches and pits are sometimes left open for days and even weeks, constituting a threat to other teams working at the site. (2) The potential threat of risk situations in an MT work is often misidentified. This leads to a dynamic risk situation. For example, a maintenance technician places a rag on the floor to absorb the condensation from a faulty office air conditioner and leaves to get his tools. A secretary inadvertently steps on the wet rag, slips and falls and injures her ankle. The analysis that follows addresses the occupational accident data of the integrated electrical PU in the State of Israel between 2004 and 2011. As of 2011, the PU company employed 12,687 workers and maintained and operated several power station sites with an aggregate installed generating capacity of 13,133 MW, supplied to customers via a national grid transmission and distribution (T&D) system. The following segmentation of the employee roster into five operational divisions reflects the company’s main areas of activity relevant to this study. The works and expertise of the first and the second divisions, the north and south District ones, are the T&D of electrical power. The third division is engaged in building power plants and substations. The fourth division is in charge of logistics, and provides transportation services, cranes, heavy vehicles and workshop works to the other divisions. The fifth division is the Generation Division, which operates the power stations. The number of the employees in these five divisions is shown in Table 1. The reported accidents of five PU’s divisions during 20042011 were compiled by its safety department. The data presented in this paper were collected in accordance with the Israel Institute for Occupational Safety and Hygiene classification system. Each of the accidents was examined to determine its causation in the context of the RD in MT work. The results were subdivided into categories using two main factors: RD and non-risk dormancy. The criteria described in Section 1.3 were applied. The results, as they relate to the above five PU divisions, are shown in Table 2. These results reveal as much as 9.85% RD-related accident rate for all the accidents. 2. Analyses 2.1. Risk dormancy analysis 2.1.1. Multi-team risk dormancy RD is the time delay between the occurrence of a failure (hazard event) in the action of one team (Team A) that affects another team (Team B) involved in the process (Figure 3). 8 E. Farag et al. Table 2. RD and non-RD accidents in MT work. Division N.MTRD.A MTRD.A. TOTAL % MTRD.A North district South district Logistic Generation Construction 783 45 828 5.4% 860 81 941 8.6% 474 64 538 11.9% 1391 134 1525 8.78% 715 133 848 9.85% Downloaded by [University of Haifa Library] at 02:29 26 April 2015 N.MTRD.A non-MTRD accidents. MTRD.A MTRD accidents. Such a failure has the potential to produce a hazard that is underestimated or undetected by Team B and eventually becomes or generates a risk that lies dormant for a period of time (risk dormancy). This risk has no immediate effect on the Team A member who caused it, but might be a threat to another team (referred to as Team B) that will later continue the same job or will be engaged in a different job at the site. This situation is referred to as RD in MT work. An analysis of the RD time path reveals the following stages leading to an accident (Figure 3): T0 beginning of Team A activity that could possibly generate a hazard event that becomes a risk and could threaten the Team B work. TR hazard event time. Td RD time, which is equal to the time interval from the hazard event caused by the Team A activity until the accident occurs, injuring the next team. The time is estimated by the professional safety officer teams, based on their experience and knowledge. Tacc the time the accident occurred. This paper addresses the time aspects of RD in MT work and proposes a model for risk evaluation and management based on time-dependent probability (TDP) methodology. In this context, classification criteria for RD are related to risks generated by MT activities. Figure 3. MT risk dormancy pathway. Theoretical Issues in Ergonomics Science 9 Downloaded by [University of Haifa Library] at 02:29 26 April 2015 2.1.2. Probabilistic analysis of risk dormancy time The analysis of RD in MT work requires collecting information regarding an accident and establishing the sequence of events that led to the accident. This includes identifying the team affected by the accident and the accident occurrence time, as well as the team that most probably generated the risk that ultimately materialised (risk-causing team) and the time when the risk was actually generated. The affected team and the time, Tacc, of accident occurrence are easily determined in most cases since accidents are usually investigated and documented. It is, however, often quite difficult to identify the risk-causing team and determine the time at which the risk was generated (Figure 3). Still, identifying the team that generated the risk is quite complicated and requires efforts of a team of professional analysts, such as the safety officer’s team, which is supposed to be familiar with the stages and layout of the work. The accident investigation determines not only the active causes leading to the accident, but also includes the attributes of the schedule itself, as well as tools, places, personnel, etc. involved in the risk creation. As a result, the safety officers can determine with high confidence the commencing time T0 of the Team A activity that most likely created a hazard event TR, which became a risky one and threatened the Team B. Finally, a team of safety experts analyses all actions that preceded the occurrence of the accident, since the range of the RD time uncertainties is basically the risk creation time until its materialisation. One can therefore calculate the statistics of the RD duration from its creation, TR, until accident occurrence, Tacc, regardless of the teams involved in the hazard generation. RD time is a random variable; hence a probabilistic analysis should apply. Actually, Tacc could be established rather accurately due to the time recording of an accident’s occurrence, while T0 and TR should be considered as best estimates made by the professional safety officers. RD is clearly a positive value, and its range is between the RD generating point TR on one side and the time of the accident occurrence Tacc, where RD terminates at the other. The RD time Td is a random variable extending between the beginning of Team A activity T0, which could possibly generate a hazard event TR, and the accident time Tacc. When TR is close to T0, then the time Td reaches its maximum value. Similarly, when the time TR is close to Tacc, then Td approaches zero. Thus, the entire range of RD time uncertainty can be expressed as Td D Tacc ¡ TR : (1) In accordance with the maximum entropy principle, we choose a uniform distribution for RD time (Figure 4) regardless of which particular team caused it. Figure 4. Risk dormancy time distribution. 10 E. Farag et al. The probability density function (pdf) is therefore a constant, as expressed by the equation: f ðTd Þ D f ðTacc ¡ T0 Þ : ðTacc ¡ T0 Þ (2) Here,f is the Heaviside step function: Downloaded by [University of Haifa Library] at 02:29 26 April 2015 f D 1 T0 < Td < Tacc : fD0 Tacc < Td (3) f (Td) is the normalised pdf of RD time (normalisation of pdf by Tacc to achieve an integral equal to 1) and Td is the random RD time uniformly distributed. Our observations are based not only on a single event but also on the distribution of RD time generated by a number of accident occurrences. Thus, the distribution should be averaged for all events. The average is the arithmetic mean of all the RD times: ( f D ðTd Þ D Xnk iD0 f ½Td ; Tacc i (4) n ( Here f D average of RD time of each division, i index of time to accident on each division, n RD accidents number on each division, k division index. The RD time probability of all accidents at the PU (in the five divisions, to be precise) is expressed in Equation (5): X5 njk ðf ðTd ÞÞk k D 1X f ðTd Þ D nk : (5) Here, f ðTd Þ average probability of RD time of all divisions Z Fexp ðTd Þ D Td f ðTd ÞdTd : (6) 0 Fexp experiment cumulative distribution function (CDF) of RD time of all divisions 3. Results As shown in Equation (7), the CDF fit function is specified by the five divisions of the PU. This fit function positively predicts the experimental CDF function Fexp in Equation (6) of RD time for MTRD accidents data: FðTd Þ D 1 ¡ ae ¡ b1 Td u1 ¡ ð1 ¡ aÞe ¡ b2 Td u2 (7) F (Td) CDF fit function, u1 scale parameter, short-term expected time, u2 scale parameter, long-term expected time, b1 shape parameter, short-term expected time, b2 shape parameter, long-term expected time, a partition parameter, dividing the data into short a part and long (1 a) content. Theoretical Issues in Ergonomics Science 11 Table 3. Distribution characteristic parameters. Division/parameter North district South district Logistics Generation Construction Downloaded by [University of Haifa Library] at 02:29 26 April 2015 a u1 u2 b1 b2 0.47 0.68 0.69 0.9 0.8 3.98 4 19 20 51 570 539 622 660 750 0.92 0.92 0.67 0.66 0.49 1.4 0.96 1.32 1.55 2.32 Majority of RD time appearances of short-term character compared to a smaller long-term RD appearance. The RD time distribution parameters for the five PU divisions are shown in Table 3. Each one is also subdivided by two distinctive populations. Sub-data are well described by the Weibull distribution. Accordingly, each data subset is characterised by a vector of three parameters as shown in Table 3. Table 3 distinguishes between two different groups. First, the three divisions: north, south and logistics, showing a clear distinction between short- and long-term. Expectedly, the fit function of this group shows a good fit to RD time data, see Figures 5(a)5(c). However, the second group of the two other divisions, generation and construction, has a majority of RD time appearances of short-term character, about 0.9 and 0.8, respectively, compared to a smaller long-term RD appearance. Because of that we ignore the longterm for this group, whose sub-data are well described by the Weibull distribution. Each subset data are characterised by a vector of two parameters (Table 4). Accordingly, a modified CDF fit function of the Weibull type is specified: FðTd Þ D 1 ¡ e ¡ b1 Td u1 : (8) The fit function shows a sufficient fitness to RD time data, see Figures 5(d) and 5(e). Monte Carlo simulation is used to generate random points from the domain RD time distribution data to determine the validity of the five divisions’ parameters a kind of bootstrap simulation. The simulation data show rather poor correlation between a, b and u parameters. Consequently, we are considering them as independent parameters. 3.1. Hazard function To confirm the results of the effect of RD, one could examine the impact of these parameters by employing the hazard function for each division: hðTd Þ D ¡ dlnð1 ¡ FðTd ÞÞ : dðTd Þ (9) The hazard function has resulted in the same groups of CDF functions with respect to RD time: Group One: north, south and logistics divisions characterised by two shape parameter b1 and b2 as follows: 0.92, 0.92, 0.67 and 1.4, 0.96, 1.32, respectively (see Tables 3 and 4). The results for the b value were found to be close to 1, demonstrating Downloaded by [University of Haifa Library] at 02:29 26 April 2015 12 E. Farag et al. Figure 5. (a) CDF of RD time of north district division (experimental and fit function). (b) CDF of RD time of south district division (experimental and fit function). (c) CDF of RD time of logistic division (experimental and fit function). (d) CDF of RD time of generation division (experimental and fit function). (e) CDF of RD time of construction division (experimental and fit function) Table 4. Distribution characteristic parameters. Division/parameter Generation Construction u1 b1 26 121 0.4 0.45 Theoretical Issues in Ergonomics Science 0.04 0.04 trace 1 Hazard function Hazard function trace 1 0.03 0.02 0.01 0 0 200 13 400 600 800 0.03 0.02 0.01 0 3 1×10 0 200 400 (a) (c) 0.04 0.04 trace 1 0.03 Hazard function Hazard function 0.02 0.01 200 400 600 0.03 0.02 0.01 0 800 0 200 400 Dormancy time 600 800 3 1× 10 Dormancy time (b) (d) 0.04 trace 1 Hazard function Downloaded by [University of Haifa Library] at 02:29 26 April 2015 trace 1 0 800 Dormancy time Dormancy time 0 600 0.03 0.02 0.01 0 0 200 400 600 800 3 1× 10 Dormancy time (e) Figure 6. (a) Hazard function of north district divisions. (b) Hazard function of south district divisions. (c) Hazard function of logistic divisions. (d) Hazard function of generation divisions. Hazard function of construction division. almost constant failure rate in time, see Figure 6(a)6(c). However, we are unable to explain the volatility behaviour of the two models presented in Equation 7, which brings one to the three choices of Weibull. Similarly, one could question why these parameter values were obtained. These questions will be addressed in the future work. Group Two: generation and construction divisions characterised by single-shape parameter b1 of 0.4 and 0.45, respectively, see Table 4. The results of b were found to be less than 1 demonstrating a decreasing failure rate in time as shown in Figure 6(d) and 6(e). 4. Discussion The type of organisations considered in this study are quite complicated, as they are characterised by significant and strong interdependence between the management and MT professional labour, in addition to the effect of interaction among themselves, regardless of simultaneity. This complexity could lead to problematic aspects in internal company behaviour, sometimes causing safety problems. 14 E. Farag et al. The scope of this paper extends the risk management approach from the well-known management models, with the addition of the important aspect of the MT safety perspective beyond simultaneous situations, i.e. RD. The research provided here extends the existing approaches to a holistic level. The obtained data on RD accidents indicate that the RD time distribution is characterised by the Weibull parameters: u1, b1 short-term, u2,b2 long-term, respectively, and partition parameter a as shown in Tables 3 and 4 above, for situations, in which the very nature of the work dictates the dormancy time, as supported in the following data discussion. First, u1, b1 short-term expected RD time and a partition parameters. Downloaded by [University of Haifa Library] at 02:29 26 April 2015 (1) North and south divisions T&D districts Scale parameter u1, whose RD time is 3.9 and 4 hours, respectively. An examination of the accident investigation data shows that tasks in the T&D districts have the following characteristics. (a) Most tasks are scheduled and completed in one day, because the nature of T&D work requires power supply resumption to customers as quickly as possible. As a result, the short-term of RD time lasts a few hours. (b) Tasks are performed sequentially by a professional MT. (c) Subtasks are performed sequentially. (d) Similar field of activity, e.g. lineman-teams of differing proficiency levels are necessary for executing complementary parts of power line work due to the complexity of the work and the existence of risk factors such as electricity, and, as a consequence of that, have high safety level requirements. For example, erecting a transformer requires at least two different teams: an electricians’ team to de-energise transformer connections to the power lines and install grounds and an overhead line-work team to install the transformer. Thus, the work teams require the necessary expertise for each phase of the work. 4.1. Shape parameter b1 of 0.92 for both districts These b1 values are close to 1, indicating an almost constant accident rate regarding RD time. This happens if there is maximum entropy, characterised by exponential distribution process. Remarkably, hazards and risky situations analysed in this paper, and which are more likely to cause accidents at a constant rate, are related to electrical supply divisions (south and north divisions as shown in Figure 6(a) and 6(b). Partition parameter a divides the south and north divisions’ RD time appearances into two separate categories in which the short-term is 0.47, 0.68, respectively. Case study 1: a lines work team of PU North division performed an underground cable connection to an overhead line. According to the safety instructions, the cable to be worked on must be positively identified by tags and must be isolated from the electric supply sources. Furthermore, tests must be performed to verify that the cable is de-energised, and grounds of an approved type must be applied to protect workers from all the energy sources. Earlier in the morning of the same day, an authorised clearance team should be assigned to identify and de-energise the underground cable, in accordance with an authorisation provided and documented by a system operator. The clearance team is supposed to install the protective short-circuiting and grounding equipment required for Theoretical Issues in Ergonomics Science 15 the protection of the team working on cable connection. The permission to start working was given at the work site. While the cable-cutting work was in progress, an explosion occurred. The team cut an energised cable. The authorised clearance team misidentified the correct cable and gave the work authorisation for the wrong cable. Workers were injured due to electrical arc flash. In this case, the above-mentioned scale parameter characteristics apply about 2 hours of short-term dormancy time. Downloaded by [University of Haifa Library] at 02:29 26 April 2015 (2) Logistic division Scale parameter u1, whose RD time is 19 hours. The present results revealed a prominent attribute related to work course duration. The large majority of appearances of these RD accident occurrences are at the end of a working day or a shift. The following theme is the main characteristic of the RD causation: in the logistic division the main MT activities occurring at the beginning or at the end of the working day/shift were the loading and unloading of trucks. Shape parameter b1 a shape parameter value of b1 D 0.67 < 1 indicates that the accident rates decrease over time. This happens if significant hazards or risky situations are generated resulting in an accident at a decreasing rate over time. Partition parameter a, which is dividing the logistic division RD time appearances into two substantial populations’ 0.69, 0.31 of short- and long-term, respectively. Case study 2: a workshop employee was on his way to repair a metal processing machine. A stack of iron bars, delivered the previous day, was still on the workshop floor, protruding into the employee’s pathway. He stumbled as he passed the stack and injured his leg. Material deliveries are usually made in the morning and materials are unloaded from trucks at the workshop yard close to where the machines are placed, pending transfer to storerooms. These irons bars were unloaded a day before the accident. A 24-hour dormancy time was estimated by the safety officer. Case study 3: the PU owns a rather big truck fleet, used for truck-mounted work platforms and truck-mounted cranes, which are used for loading/unloading and uplifting workers to heights. In this case, one of the trucks was sent back to duty from in-house periodic maintenance service on the morning of that day, the truck driver opened the engine hood during a routine cleaning and checking procedure at the end of the shift and was injured while trying to remove a ‘piece of rubber’ that was inadvertently left there by a garage worker. An eight-hour RD time was estimated by safety department. (3) Generation and construction divisions with significant short-term expected RD time It is obvious from Table (3) that partition parameter a of the magnitude of 0.9 and 0.8, respectively, indicates the dominant short-term RD time appearances in these divisions. Therefore, we neglected/ignored the long-term appearance in those two divisions as shown in Table (4). Scale parameter u1, whose RD time is 26 and 121 hours, respectively. Task substitute time, i.e. first task completion and transition to the next task by MT at generation, construction and logistic divisions are longer than in T&D districts. Shape parameter b1 of 0.4 and 0.45, values of b1 < 1, indicates that the accidents rate decreases over time. This happens if there are significant hazards or risky situations generated early and leading to an accident in a decreasing rate over time. 16 E. Farag et al. Case study 4: To carry out a maintenance job in a turbine building of a power station, a scaffold was erected and placed on the route of an overhead bridge crane. The crane consists of parallel runways with a travelling bridge spanning the gap and equipped with hoist that travels along the bridge. These cranes are electrically operated from ground level by a control pendant. Three days later, a team from another department used the crane for lifting heavy valves as part of a job. The crane hit the scaffold, causing extensive damage. In this case, the above-mentioned generation scale parameter characteristics apply a RD time of about 72 hours. Downloaded by [University of Haifa Library] at 02:29 26 April 2015 4.2. Second, u2, b2 long-term expected RD time and partition parameters (1) North, south and logistic divisions Scale parameter u2 whose RD time is 570, 539 and 622 hours, respectively, the long-term RD accidents are likely to affect teams with no professional linkage between them. There is no significant difference observed in the results for all the divisions as seen in Tables (3) and (4). Case study 5: a working team of the Southern PU district was sent for carrying out maintenance work on electric supply line of low voltage network. An employee whose job included activities installing, constructing, adjusting, repairing, etc. climbed on a metal pole and started the repair work. An electrical current flow as a result of contact between transformer cables and the pole causing electrical shake to worker. Investigation found that a month before another working team of the same district performed different work on the same transformer, causing faulty cables connection of the transformer. As a result, loose connection of the cable made contact with the conductive metal pole and electrical flash of short circuit injured the team. In this case, different teams, namely maintenance and operation teams of the same district, performed different tasks without professional affiliation or linkage between them. The scale parameter u2 characteristic applied a RD time of about 720 hours. (2) Generation and construction divisions. (3) Negligible long-term expected RD time. 5. Conclusions A novel probabilistic risk management model has been introduced to characterise the RD phenomenon to be considered as a proper way of taking into account MT aspects in the occupational safety research. Furthermore, a holistic perception of MT functional complexity allows for a generalised view of MT mutual interaction instead of focusing in the single team behaviour. The following conclusions can be drawn from the carried out analysis. The model is innovative in two major ways: First, the identification of RD in sequential MT work, i.e. MTRD. Though, unidentified dormant risks or the underestimation of identified dormant risks are a ‘ticking bomb’, each such risk represents an unsafe/hazardous event that is certain to happen in the foreseeable future and which threatens other teams continuing the same or a different job on site. Indeed, the current timespace approach does not address or offer solutions to such risks. Therefore, we have developed an RD approach that offers a solution for predicting TDP. Second, the PU accident Theoretical Issues in Ergonomics Science 17 database enables us to evaluate and determine the above-mentioned significant risk aspect tendencies. Accordingly, the proposed model defines RD, a new facet of risks generated by MT work in modern industrial and infrastructure organisations, regardless of the time frame involved. Acknowledgements Downloaded by [University of Haifa Library] at 02:29 26 April 2015 The authors would like to thank the Israel Electric Company for allowing us to use their safety data base which made valuable contribution to the research. This study is dedicated to my friend, late Dr Majdi Latif, who inspired and encouraged me. “True friendship is the willingness to sacrifice one’s self for the other.” Disclosure statement No potential conflict of interest was reported by the authors. About the authors Emad Farag is currently a PhD student in Industrial Engineering and Management at the TechnionIsrael Institute of Technology, under the supervision of professor Dov Ingman. His research deals with how to use reliability methods (such as time dependent probability, hazard function and modelling) as tools for managing risks in the modern multi-team organisations as part of occupational safety management, with a special focus on the dormant risks and teams interaction. Dov Ingman is a staff member of Industrial and Management Engineering faculty at TechnionIsrael Institute of Technology. His research interests include element and system reliability, damage accumulation processes, physical kinetics, pattern recognition, information theory, neural nets, measurement theory and instrumentation, desalination technology, non-destructive testing and quality control. Ephraim Suhir is a fellow of the American Physical Society, the Institute of Physics, UK, Institute of Electrical and Electronics Engineers, American Society of Mechanical Engineers, Society of Optical Engineers, International Microelectronics and Packaging Society, and the Society of Plastics Engineers. He is a Fulbright Scholar in information technologies, State Department, USA, and Foreign Full Member of the National Academy of Engineering, Ukraine. Ephraim has authored about 300 technical publications (patents, books, book chapters, technical papers) and received numerous professional awards in various fields of engineering and applied science. Currently, he is a staff member of Mechanics and Materials Department, Portland State University, Portland, OR, USA. ORCID Emad Farag http://orcid.org/0000-0002-1595-5833 References Abdelhamid, S.T., and J.G. Everett. 2000. “Identifying Root Causes of Construction Accidents.” Journal of Construction Engineering and Management, ASCE 126: 5260. Akinci, B., M. Fischen, R. Levitt, and R. Carlson. 2002. “Formalization and Automation of TimeSpace Conflict Analysis.” Journal of Computing in Civil Engineering 16(2): 124134. Carter, G., and S. Smith. 2006. “Safety Hazard Identification on Construction Projects.” Journal of Construction Engineering and Management 132 (2): 197205. Cuny, X., and M. Lejeune. 2003. “Statistical Modelling and Risk Assessment.” Safety Science 41 (1): 2951. Drivas, S., and M. Papadopoulos. 2004. Occupational Risk Assessment. Chapter 2 in Guide for the Health and Safety of Workers, Athens: ELINYAE EKA. Downloaded by [University of Haifa Library] at 02:29 26 April 2015 18 E. Farag et al. Griffel, A. 1999. “Countermeasure Analyses: A Strategy for Evaluating and Ranking Safety Recommendations.” PhD diss., Haifa University. La Coze, J. 2005. “Are Organisations Too Complex to be Integrated in Technical Risk Assessment and Current Safety Auditing?” Safety Science 43 (2005): 613638. Mitropoulos, P., G.A. Howell, and S.T. Abdelhamid. 2005. “Accident Prevention Strategies: Causation Model and Research Directions.” ASCE. San Diego: Construction Research Congress. Papadopoulos, G., G. Paraskevi, P. Christos, M. Katerina. 2009. “Occupational and Public Health and Safety in a Changing Work Environment: An Integrated Approach for Risk Assessment and Prevention.” Safety Science. doi:10.1016/j.ssci.2009.11.002 Rasmussen, J., A.M. Pejtersen, and L.P. Goodstein. 1994. Cognitive System Engineering. New York, NY: Wiley. Rasmussen, J., and I. Svedung. 2000. Proactive Risk Management in a Dynamic Society. Karlstad: Risk & Environmental Department, Swedish Rescue Services Agency. Reason, J.T. 1997. Managing the Risks of Organizational Accidents. Aldershot: Ashgate Publishing Ltd. Reason, J.T. 2000. “Human Error: Models and Management.” British Medical Journal 320: 768770. Rosenfeld, Y., O. Rozenfeld, R. Sacks, and H. Baum. 2006. “Efficient and Timely Use of Safety Resources in Construction.” Proceedings of the CIB W99 International Conference on Construction, edited by Safety, D. Fang, R.M. Choudhry and J.W. Hinze, 290297. Beijing, China. Rozenfeld, O., R. Sacks, and Y. Rosenfeld. 2010. “CHASTE Construction Hazard Analysis with Spatial and Temporal Exposure.” Construction Management & Economics 48 (4): 491498. Sasoua, K., and J. Reason. 1999. “Team Errors: Definition and Taxonomy.” Reliability Engineering and System Safety 65 (1): 19.
© Copyright 2026 Paperzz