Bayesian Data Analysis of Severe Fatal Accident Risk in the Oil Chain

We analyze the risk of severe fatal accidents causing five or more fatalities and for nine different activities covering the entire oil chain. Included are exploration and extraction, transport by different modes, refining and final end use in power plants, heating or gas stations. The risks are quantified separately for OECD and non‐OECD countries and trends are calculated. Risk is analyzed by employing a Bayesian hierarchical model yielding analytical functions for both frequency (Poisson) and severity distributions (Generalized Pareto) as well as frequency trends. This approach addresses a key problem in risk estimation—namely the scarcity of data resulting in high uncertainties in particular for the risk of extreme events, where the risk is extrapolated beyond the historically most severe accidents. Bayesian data analysis allows the pooling of information from different data sets covering, for example, the different stages of the energy chains or different modes of transportation. In addition, it also inherently delivers a measure of uncertainty. This approach provides a framework, which comprehensively covers risk throughout the oil chain, allowing the allocation of risk in sustainability assessments. It also permits the progressive addition of new data to refine the risk estimates. Frequency, severity, and trends show substantial differences between the activities, emphasizing the need for detailed risk analysis.     

Risk Analysis and Management of Dam Safety

The current safety criteria for a high hazard dam focus on protecting the dam during a large flood. While protecting the dam does help to protect downstream people and property, the two objectives are not the same. Instead, the criteria should focus on lowering property damage (including damage to the dam) and preventing flood deaths. High hazard dams must survive a design flood in the current safety criteria. However, experts don't agree on the size of the peak flow that meets this criteria. Statistical hydrologists have proposed an alternative to using professional judgment to specify the design flood. Unfortunately, peak flow distributions cannot be estimated with confidence for extreme floods given available data. A major safety goal is to prevent deaths from floods. Preventing deaths is a major reason for constructing the spillway to handle extreme floods so that the dam doesn't fail due to overtopping. However, even if the dam doesn't fail, the spilled floods could cause many deaths. A better approach is to warn people to get them out of harm's way if a flood is coming. Retrofitting existing dams that could pass a “probable maximum flood” (PMF) when built is almost never a good use of funds. Instead, funds would be spent better by focusing on preventing damage from small floods, lowering the damage from medium-sized floods, and warning people in the event of a flood that could pose risks to life.     

Modeling Ship Transportation Risk

This article presents results from the Commission of the European Communities (CEC) project 'Safety of Shipping in Coastal Waters' (SAFECO). The project was performed by ten European partners during the period 1995-1998. The principal aim of the SAFECO project was to determine the influences that could increase the safety of shipping in coastal waters by analyzing the underlying factors contributing to the marine accident risk level. The work reported here focuses on the Marine Accident Risk Calculation System (MARCS) that was further developed during the SAFECO project. This paper presents the methods used by MARCS, as well as data and results from a 'demonstration of concept' case study covering the North Sea area. The estimated accident frequencies (number of accidents per year) were compared with historical accident data, to demonstrate the validity of the modeling approach. Reasonable (within a factor of 5) to good (within a factor of 2) agreement between calculated accident frequencies and observed accident statistics was generally obtained. However, significant discrepancies were identified for some ship types and accident categories. The risk model has particular problems with estimating the accident frequency for drift grounding in general and powered grounding for ferries. It was concluded that these discrepancies are related to uncertainties in several areas, specifically in the risk model algorithms, the traffic data, the error and failure probability data, and the historical accident statistics.     

Major Accidents (Gray Swans) Likelihood Modeling Using Accident Precursors and Approximate Reasoning

Compared to the remarkable progress in risk analysis of normal accidents, the risk analysis of major accidents has not been so well‐established, partly due to the complexity of such accidents and partly due to low probabilities involved. The issue of low probabilities normally arises from the scarcity of major accidents’ relevant data since such accidents are few and far between. In this work, knowing that major accidents are frequently preceded by accident precursors, a novel precursor‐based methodology has been developed for likelihood modeling of major accidents in critical infrastructures based on a unique combination of accident precursor data, information theory, and approximate reasoning. For this purpose, we have introduced an innovative application of information analysis to identify the most informative near accident of a major accident. The observed data of the near accident were then used to establish predictive scenarios to foresee the occurrence of the major accident. We verified the methodology using offshore blowouts in the Gulf of Mexico, and then demonstrated its application to dam breaches in the United Sates.     

The relationship between job insecurity and accident under-reporting: A test in two countries

Although the problem of accident under-reporting is increasingly recognized in the literature, less is known regarding work environment variables that predict the severity of such under-reporting. Data on perceived job insecurity and on the numbers of both accidents that were experienced and accidents that were actually reported over a period of one year were obtained from 786 employees in 24 US organizations and 563 employees in 20 Italian organizations in sectors where safety is highly relevant. Analysis of these data suggested that not only is job insecurity related to the likelihood of experiencing an accident, but also perceptions of job insecurity may serve to inhibit the reporting of accidents to appropriate company officials. Overall, the rate of accidents was lower in Italy than the United States. Moreover, in both countries, when job insecurity was low, there was little difference between the total number of experienced accidents and the number that employees reported. However, as job insecurity increased, the under-reporting of accidents increased. The implications of these results are discussed in light of the globally increasing prevalence of job insecurity in today's workplace.     

Human Factors Analysis for Maritime Accidents Based on a Dynamic Fuzzy Bayesian Network

Human factors are widely regarded to be highly contributing factors to maritime accident prevention system failures. The conventional methods for human factor assessment, especially quantitative techniques, such as fault trees and bow-ties, are static and cannot deal with models with uncertainty, which limits their application to human factors risk analysis. To alleviate these drawbacks, in the present study, a new human factor analysis framework called multidimensional analysis model of accident causes (MAMAC) is introduced. MAMAC combines the human factors analysis and classification system and business process management. In addition, intuitionistic fuzzy set theory and Bayesian Network are integrated into MAMAC to form a comprehensive dynamic human factors analysis model characterized by flexibility and uncertainty handling. The proposed model is tested on maritime accident scenarios from a sand carrier accident database in China to investigate the human factors involved, and the top 10 most highly contributing primary events associated with the human factors leading to sand carrier accidents are identified. According to the results of this study, direct human factors, classified as unsafe acts, are not a focus for maritime investigators and scholars. Meanwhile, unsafe preconditions and unsafe supervision are listed as the top two considerations for human factors analysis, especially for supervision failures of shipping companies and ship owners. Moreover, potential safety countermeasures for the most highly contributing human factors are proposed in this article. Finally, an application of the proposed model verifies its advantages in calculating the failure probability of accidents induced by human factors.     

Risk Prevention and Policy-Making in Automatic Systems

Accidents with automatic production systems are reported to be on the order of one in a hundred or thousand robot-years, while fatal accidents are found to occur one or two orders of magnitude less frequently. Traditions in occupational safety tend to seek for safety targets in terms of zero severe accidents for automatic systems. Decision-making requires a risk assessment balancing potential risk reduction measures and costs within the cultural environment of a production company. This paper presents a simplified procedure which acts as a decision tool. The procedure is based on a risk concept approaching prevention both in a deterministic and in a probabilistic manner. Eight accident scenarios are shown to represent the potential accident processes involving robot interactions with people. Seven prevention policies are shown to cover the accident scenarios in principle. An additional probabilistic approach may indicate which extra safety measures can be taken against what risk reduction and additional costs. The risk evaluation process aims at achieving a quantitative acceptable risk level. For that purpose, three risk evaluation methods are discussed with respect to reaching broad consensus on the safety targets.     

Learning from Nuclear Accident Experience

Statistical procedures are developed to estimate accident occurrence rates from historical event records, to predict future rates and trends, and to estimate the accuracy of the rate estimates and predictions. Maximum likelihood estimation is applied to several learning models and results are compared to earlier graphical and analytical estimates. The models are based on (1) the cumulative number of operating years, (2) the cumulative number of plants built, and (3) accidents (explicitly), with the accident rate distinctly different before and after an accident. The statistical accuracies of the parameters estimated are obtained in analytical form using the Fisher information matrix. Using data on core damage accidents in electricity producing plants , it is estimated that the probability for a plant to have a serious flaw has decreased from 0.1 to 0.01 during the developmental phase of the nuclear industry. At the same time the equivalent frequency of accidents has decreased from 0.04 per reactor year to 0.0004 per reactor year, partly due to the increasing population of plants.     

Maritime Transportation Risk Assessment of Tianjin Port with Bayesian Belief Networks

This article develops a Bayesian belief network model for the prediction of accident consequences in the Tianjin port. The study starts with a statistical analysis of historical accident data of six years from 2008 to 2013. Then a Bayesian belief network is constructed to express the dependencies between the indicator variables and accident consequences. The statistics and expert knowledge are synthesized in the Bayesian belief network model to obtain the probability distribution of the consequences. By a sensitivity analysis, several indicator variables that have influence on the consequences are identified, including navigational area, ship type and time of the day. The results indicate that the consequences are most sensitive to the position where the accidents occurred, followed by time of day and ship length. The results also reflect that the navigational risk of the Tianjin port is at the acceptable level, despite that there is more room of improvement. These results can be used by the Maritime Safety Administration to take effective measures to enhance maritime safety in the Tianjin port.     

Evacuation Risks: A Tentative Approach for Quantification

This study tries to assess the risk of deaths and injuries from motor vehicle accidents associated with an evacuation of population groups in case of nuclear plant accidents. The risk per person–km is evaluated using: (a) data from previous evacuation: information from Soufriere evacuation (Guadeloupe Island 1976) and Mississauga (1979), added to Hans and Sell's data: no road accident occurred for a sample of 1,500,000 persons; (b) national recording system for motor vehicle accident: the rates of 2.2 10 ^-8 deaths per person–km and 32 10^-8 injuries per person–km is calculated as an average. These last rates in France overestimate the number of casualties. A reasonable hypothesis is to assume that the probability of road accident occurrence follows a Poisson distribution, as these events are independent and unfrequent, as no accident was observed in a sample of 1,500,000 persons the probability is between 0 and an upper value of 0.24 10^-8 deaths per person-km and 3.29 10^-8 injuries per person–km. The average and maximum population involved within different radii around French and U.S. Nuclear power sites are taken as a sample size in order to study the total risk of deaths and injuries in the hypothesis of an evacuation being necessary to protect the populations.     

Workplace Accidents and Workplace Safety: On Under‐reporting and Temporary Jobs

Statistics on workplace accidents do not always reflect workplace safety because workers under‐report for fear of job‐loss if they report having had an accident. Based on an analysis of fatal and non‐fatal workplace accidents and road accidents in 15 EU‐countries over the period 1995–2012, we conclude that there seems to be cyclical fluctuations in reporting of non‐fatal workplace accidents. Workers are less likely to report a workplace accident when unemployment is high. Furthermore, analyzing data from Italy and Spain on both workplace accidents and commuting accidents, we conclude that workers on temporary jobs are likely to under‐report accidents.     

Managing the Fukushima Challenge

The Fukushima Daiichi accident raises a fundamental question: Can science and technology prevent the inevitability of serious accidents, especially those with low probabilities and high consequences? This question reminds us of a longstanding challenge with the trans‐sciences, originally addressed by Alvin Weinberg well before the Three Mile Island and Chernobyl accidents. This article, revisiting Weinberg's issue, aims at gaining insights from the accident with a special emphasis on the sociotechnical or human behavioral aspects lying behind the accident's causes. In particular, an innovative method for managing the challenge is explored referring to behavioral science approaches to a decision‐making process on risk management; such as managing human behavioral risks with information asymmetry, seeking a rational consensus with communicative action, and pursuing procedural rationality through interactions with the outer environment. In short, this article describes the emerging need for Japan to transform its national safety management institutions so that these might be based on interactive communication with parties inside and outside Japan.     

Risk Management of Domino Effects Considering Dynamic Consequence Analysis

Domino effects are low‐probability high‐consequence accidents causing severe damage to humans, process plants, and the environment. Because domino effects affect large areas and are difficult to control, preventive safety measures have been given priority over mitigative measures. As a result, safety distances and safety inventories have been used as preventive safety measures to reduce the escalation probability of domino effects. However, these safety measures are usually designed considering static accident scenarios. In this study, we show that compared to a static worst‐case accident analysis, a dynamic consequence analysis provides a more rational approach for risk assessment and management of domino effects. This study also presents the application of Bayesian networks and conflict analysis to risk‐based allocation of chemical inventories to minimize the consequences and thus to reduce the escalation probability. It emphasizes the risk management of chemical inventories as an inherent safety measure, particularly in existing process plants where the applicability of other safety measures such as safety distances is limited.     

Accident Epidemiology and the U.S. Chemical Industry: Accident History and Worst-Case Data from RMP*Info

This article reports on the data collected on one of the most ambitious government-sponsored environmental data acquisition projects of all time, the Risk Management Plan (RMP) data collected under section 112(r) of the Clean Air Act Amendments of 1990. This RMP Rule 112(r) was triggered by the Bhopal accident in 1984 and led to the requirement that each qualifying facility develop and file with the U.S. Environmental Protection Agency a Risk Management Plan (RMP) as well as accident history data for the five-year period preceding the filing of the RMP. These data were collected in 1999-2001 on more than 15,000 facilities in the United States that store or use listed toxic or flammable chemicals believed to be a hazard to the environment or to human health of facility employees or off-site residents of host communities. The resulting database, RMP*Info, has become a key resource for regulators and researchers concerned with the frequency and severity of accidents, and the underlying facility-specific factors that are statistically associated with accident and injury rates. This article analyzes which facilities actually filed under the Rule and presents results on accident frequencies and severities available from the RMP*Info database. This article also presents summaries of related results from RMP*Info on Offsite Consequence Analysis (OCA), an analytical estimate of the potential consequences of hypothetical worst-case and alternative accidental releases on the public and environment around the facility. The OCA data have become a key input in the evaluation of site security assessment and mitigation policies for both government planners as well as facility managers and their insurers. Following the survey of the RMP*Info data, we discuss the rich set of policy decisions that may be informed by research based on these data.     

A Review of Plant Specific PRAs

A large number of PRA studies have been completed for specific plants at specific sites. From these studies, taken individually or collectively, many significant insights have evolved into items important to risk and safety. The content of this paper is primarily based on the material contained in the EPRI funded review of five PRA studies: Big Rock Point, Zion, Limerick, Grand Gulf, and Arkansas Nuclear One. The first three were the utility sponsored studies publicly available at the time of project initiation while the other two were deemed representative of the NRC's RSSMAP and IREP programs respectively. The results of PRA studies are usually expressed in terms of core melt frequencies, radionuclide release frequencies, and frequencies of occurrence of different reactor accident consequences (e.g., early and latent fatalities) depending on the level of PRA. These subjects are prominently addressed in this paper. One of the results of a PRA study is identification of a relatively small number of accident sequences that represent the dominant contributors to core melt. An analysis of the salient features of the dominant accident sequences from eleven PRA's yielded a characterization of accident sequence categories discussed at some length. Impact of external events is discussed very briefly. Next to an explicit quantification of public risk or core melt frequency, the identification of specific safety concerns and the evaluation of possible solutions to implement risk management are probably the best recognized and most widely used applications of PRA. Several illustrative examples are briefly discussed. Human interactions are extremely important contributors to safety and reliability of the plants. A review of PRA studies concluded that it was necessary to account for five types of human interactions; some of which may mitigate while others may exacerbate an accident sequence.     

Methodology for the Calculation of Annualized Incremental Risks in Systems of Dams

In the past few years, the field of dam safety has approached risk informed methodologies throughout the world and several methodologies and programs are appearing to aid in the systematization of the calculations. The most common way of implementing these calculations is through the use of event trees, computing event probabilities, and incremental consequences. This methodology is flexible enough for several situations, but its generalization to the case of systems of several dams is complex and its implementation in a completely general calculation methodology presents some problems. Retaining the event tree framework, a new methodology is proposed to calculate incremental risks. The main advantage of this proposed methodology is the ease with which it can be applied to systems of several dams: with a single risk model that describes the complete system and with a single calculation the incremental risks of the system can be obtained, being able to allocate the risk of each dam and of each failure mode. The article shows how both methodologies are equivalent and also applies them to a case study.     

A Causal Analysis of Large Vehicle Accidents Through Fault-Tree Analysis

Recent trends indicate that vehicle miles traveled for large trucks is increasing at a higher rate than for other vehicles. The resulting competition between large trucks and other vehicles for highway space can be expected to result in more multivehicle collisions involving large trucks. This paper presents the result of an investigation of the causes and the mechanism related to large vehicle accidents. A fault-tree analysis of large vehicle accidents identifies the individual roles played by driver, vehicle, and environmental factors, as well as their interactions in the accident mechanism. Using accident data for 1984-1986, the probabilities for different basic events in the fault tree were assessed. The most likely events leading to a large vehicle accident, as well as the most effective counter measures, were then identified. The result indicate that the most prevalent form of accidents due to driver-related failures is when a normal driver makes an error in judgment and is unsuccessful in his or her evasive action. For vehicle-related failures, the predominant type of failure is equipment failure, and for environmental-related failures, excessive demand on driver and vehicle performance created by the environmental or roadway factors.     

Risk-Based Ranking of Dominant Contributors to Maritime Pollution Events

This report describes a conceptual approach for identifying dominant contributors to risk from maritime shipping of hazardous materials. Maritime transportation accidents are relatively common occurrences compared to more frequently analyzed contributors to public risk. Yet research on maritime safety and pollution incidents has not been guided by a systematic, risk-based approach. Maritime shipping accidents can be analyzed using event trees to group the accidents into "bins," or groups, of similar characteristics such as type of cargo, location of accident (e.g., harbor, inland waterway), type of accident (e.g., fire, collision, grounding), and size of release. The importance of specific types of events to each accident bin can be quantified. Then the overall importance of accident events to risk can be estimated by weighting the events' individual bin importance measures by the risk associated with each accident bin.     

Probabilistic Analysis of Domino Effects by Using a Matrix-Based Simulation Approach

Major industrial accidents occurring at so-called major hazard installations may cause domino accidents which are among the most destructive industrial accidents existing at present. As there may be many hazard installations in an area, a primary accident scenario may potentially propagate from one installation to another, and correlations exist in probability calculations of domino effects. In addition, during the propagation of a domino effect, accidents of diverse types may occur, some of them having a synergistic effect, while others do not. These characteristics make the analytical formulation of domino accidents very complex. In this work, a simple matrix-based modeling approach for domino effect analysis is proposed. Matrices can be used to represent the mutual influences of different escalation vectors between installations. On this basis, an analysis approach for accident propagation as well as a simulation-based algorithm for probability calculation of accidents and accident levels is provided. The applicability and flexibility of this approach is discussed while applying it to estimate domino probabilities in a case study.