اثر خستگی انسانی در خطر ابتلا به groundings دریایی - روش مدل سازی بیزی شبکه
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|29290||2014||14 صفحه PDF||سفارش دهید||محاسبه نشده|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Safety Science, Volume 62, February 2014, Pages 427–440
The article introduces a general method for developing a Bayesian Network (BN) for modeling the risk of maritime ship accidents. A BN of human fatigue in the bridge management team and the risk of ship grounding is proposed. The qualitative part of the BN has been structured based on modifying the Human Factor Analysis and Classification System (HFACS). The quantitative part is based upon correlation analysis of fatigue-related factors identified from 93 accident investigation reports. The BN model shows that fatigue has a significant effect on the probability of grounding. A fatigued operator raises the probability of grounding of a large ship in long transit with 23%. Compared to the two watch system (6–6 and 12–12), the 8–4–4–8 watch system seems to generate the least fatigue. However, when manning level, which is influenced by the various watch schemes, is taken into account, the two watch system is preferable, leading to less fatigue and fewer groundings. The strongest fatigue-related factors related to top management are vessel certifications, manning resources, and quality control.
High safety performance has become increasingly important in many high-risk industries. Nuclear power, the chemical industry, offshore oil and gas production and air traffic control are some example of such industries, but almost no research has focused on shipping (Håvold and Nesset, 2009). Maritime transportation has a history of accidents. Although today’s ships are highly equipped with navigation technology, information from the International Union of Marine Insurance (IUMI) indicate that the number of shipping accidents is increasing, and the reasons are attributed to the humans on board (Nilsson et al., 2009). Work conditions and organization are elements in a system that are presumed to contribute to accidents (García-Herrero et al., 2012). In general, seafarers are reported to experience more accidents than the onshore population (Roberts and Hansen, 2002). There is no consensus on the statistical distribution of the causes to shipping accidents due to the different viewpoints of accident analysis and investigation approaches. However, human errors, technical and mechanical failures are typically underlined as the main group of causes (Celik et al., 2009). An important reason for human error is considered to be human fatigue (Gould and Koefoed, 2007, Lützhöft et al., 2007, Xhelilaj and Lapa, 2010, Dorrian et al., 2011 and Akhtar and Utne, submitted for publication). In 2006, Norway experienced 88 ship groundings. In 8 of them the watch keepers had fallen asleep (Gould and Koefoed, 2007). Understanding and prevention of shipping accidents is still a focal matter of maritime interest and importance. The true extent of human fatigue, its causes and mechanisms in transportation, are unknown. The scholars disagree because human fatigue is a multi-dimensional construct and its effects on cognitive performance are therefore also complex. In general, they agree that statistics underestimate the true magnitude of the problem because of underreporting (Williamson et al., 2009). The poorly detailed and non- uniform accident databases scattered around the world also hinder a pure statistical approach (Li and Wonham, 2001 and Hassel et al., 2011). Yet, even though it has not been proven, studies do point to a strong connection between fatigue and the risk of accidents (Rothblum et al., 2002, Jensen et al., 2004 and Xhelilaj and Lapa, 2010). Fatigue can be classified into physical and cognitive (mental) categories. Mental fatigue is believed to be psychological in nature, whereas physical fatigue is considered synonymous with muscle fatigue (Grandjean, 1979 and Lal and Craig, 2001). Both physical and mental fatigue causes decline in alertness, mental concentration, and motivation. Fatigue decreases the speed of cognitive processing, and thus the major symptom of mental fatigue is a general sensation of weariness, increase in reaction time, lower vigilance and disinclination for any kind of activity (Grandjean, 1979 and Sneddon et al. 2012). Psychological distress is shown to be most aggravated in workers who face high demands in their jobs with, for instance, excessive work load, confined spaces and poor thermal conditions (García-Herrero et al., 2012). Human fatigue lacks a clearly defined and agreed upon definition, even though it has long been a topic of research. A definition of maritime human fatigue is “a biological drive for recuperative rest” ( Desmond and Hancock, 2001, Noy et al., 2009 and Williamson et al., 2009). However, a broader definition of fatigue is “subjective experience of someone who is obliged to continue working beyond the point at which they feel confident of performing a task efficiently” ( Smith et al., 2001). Human fatigue is difficult to measure and even more difficult to state as a cause to an accident. Therefore accident investigation reports are often reluctant assigning any large importance to human fatigue. Therefore, by analyzing accident investigation reports (as done in this study), one has to rely on the subjective reports from people involved. Whether or not human fatigue is likely to have been present has to be assessed based on the mentioned fatigue influencing factors in the reports. The latter definition is therefore used in this study. The definition also covers both the mental and the physical fatigue. Throughout the article when fatigue is mentioned, it is referred to human fatigue. Grounding can be categorized into drift and power grounding. Drift grounding, which is defined as grounding with no engine power, seldom leads to high –energy impacts. However the wave actions may break down the hull over time. Power grounding occurs with the engine running, which often means grounding in higher speed and more damage (Kristiansen, 2001). Both types of accidents are included in our study, and the risk quantification in this study is therefore for drift and power groundings combined. Further on, contact of the ship’s hull with the seabed is deemed sufficient to classify a event as a grounding. It is not a requirement for the ship to actually get stranded on the seabed. The objective of this article is to present an approach for developing a Bayesian Network (BN) for modeling the risk of maritime accidents. More specifically, the article focuses on human fatigue in a ship’s Bridge Management Team (BMT) and its influence on the risk of maritime grounding accidents. There exists research on fatigue in BMT (Akhtar and Utne, submitted for publication), but since fatigue is so multi -dimensional and vague, it is regarded as a highly difficult task to measure the effect of human fatigue on the risk of maritime accidents, and to our knowledge no attempt has yet been made. Fatigue is a complex phenomenon (IMO, 2001, Allen et al., 2008 and Akhtar and Utne, submitted for publication), and looking into only a set of causes or factors individually will therefore not provide the whole picture (Smith et al., 2006). Thus, it is necessary to consider the interplay of factors when analyzing such accidents (Zhao et al., 2011). It has been argued for a more holistic and far reaching research on seafarers’ fatigue (Smith et al., 2006). Greenberg (2007) analyzed various models and techniques available in the field of accident modeling, and concluded that traditional models fall short in analyzing and evaluating phenomena that are exhibited by socio-technical systems. This is due to the difficulty of performing safety analysis when proceeding from simple components to systems and to socio-technical systems. The challenge of modeling human performance as part of a system is a problem that still is not fully solved. Human behavior is influenced by a combination of personal traits, social beliefs and the organizational system. Greenberg (2007) concluded that the most promising way is the use of probabilistic modeling, the most suitable technique being the BN, which is well adapted to model complex systems, and a range of different variables can be included into the system without too much difficulty. The article is structured as follows: Section 2 introduces the BN method, Section 3 introduces an eighth step approach for constructing a BN and explains how the study in the article was conducted, Section 4 presents the results, and Section 5 gives the conclusions.
نتیجه گیری انگلیسی
Numerous connections between the various fatigue-related factors can be explored with the BN. This article has highlighted some of them, focusing on how the probability of grounding is affected by human fatigue. Other influences to the probability of grounding could have been analyzed in a similar manner. However, our purpose was firstly to learn more about human fatigue and its consequences on the BMT, and secondly, to propose a detailed approach for developing a BN from accident investigation reports. The marginal probability of one or more persons in the Bridge Management Team (BMT) being fatigued at any given time in a random vessel is calculated to 0.23 (excluding ferries and smaller vessels). However, if some of the fatigue-related factors are present, the probability rises rapidly. Fatigue seems to play a significant role in maritime grounding accidents. A fatigued BMT has about 16% higher probability of grounding than a non-fatigued team. Of the 63 fatigue-related factor there were 8 which influenced the probability of fatigue the most: “LTA vessel certifications”, “LTA manning recourses”, “LTA manning”, “efficiency pressure”, “LTA monitoring of helmsman”, “narrowing attention”, “missed observation” and “lapsing/micro sleep” impact the probability of fatigue in the crew the most. When “vessel certifications”, “manning resources” and “quality control” are in OK states the probability of fatigue is reduced from 0.23 to 0.20, and subsequently, the probability of grounding decreases to 1.25 × 10−3 per year for a random vessel (from 1.27 × 10−3). The probability of an adequate safety culture onboard also increases from 0.45 to 0.61. A company may use the model to calculate its level of fatigue in the BMT and the corresponding probability of grounding. Also, the model may be used to assess the efficiency of risk reducing measures. The probability of grounding can be calculated with and without measures implemented so that any impact on the probability can be assessed. In general, adequate manning is crucial for coping with fatigue. Fatigue reducing measures regularly discussed in the literature, for instance, eating regular meals, having enough sleep, adequate resting periods, reduced administrative tasks, and free time, all depend on sufficient manning levels onboard. One cannot, for instance, address adequate resting periods for the crew without taken into account the manning levels onboard. Further on, the best and the worst case scenario show that there are significant gains in improving organizational and intermediate management factors (preconditions of unsafe acts). In the best case scenario, the probability of fatigue drops to 1% (from 23%), and in the worst case it increases to 77%. This article shows that 6–6 and 12–12 watch schemes reduce the probability of fatigue being present in the BMT more than the 8–4–4–8 watch scheme when other fatigue-related factors also are taken into consideration. Earlier research has crudely concluded that 8–4–4–8 causes less fatigue, which is true, but only if other influences are held constant in the two watch schemes. This article shows that in relation to fatigue, “adequate manning level” is crucial for the success of 8–4–4–8. However, one may argue that the changes between the various schemes are of less importance compared to the effects from the eight major fatigue-related factors. For instance setting evidence on “variable working hours” increases the probability of fatigue to 0.35, while 8–4–4–8 raises the probability of fatigue from 0.23 to only 0.24. Still, rather than proving superiority of one scheme over another, this example demonstrates the strength of using BN, i.e., once the model is developed, several aspects can easily be looked into simultaneously, because the BN provides a holistic approach to a problem. Other aspects must also be taken into account when choosing work scheme, for instance the size of the ship, the bridge’s user friendliness, etc. The results from this study should not be generalized without constraints. Our input data is based on larger vessels with at least seven crewmembers, serving routes of at least one day long. The number of accidents analyzed was 93. Future research should increase the amount of accident analyses and also standardize the accident investigation reports which would help in quantifying the data. An online version of the BN could be developed which would enable ship managements around the world to judge their crew’s fatigue levels, and see the likely results of various risk reducing measures before implementing them.