Systems thinking,the Swiss Cheese Model and accident analysis: A comparative systemic analysis of the Grayrigg train derailment using the ATSB,AcciMap and STAMP models

The Swiss Cheese Model (SCM) is the most popular accident causation model and is widely used throughout various industries. A debate exists in the research literature over whether the SCM remains a viable tool for accident analysis. Critics of the model suggest that it provides a sequential, oversimplified view of accidents. Conversely, proponents suggest that it embodies the concepts of systems theory, as per the contemporary systemic analysis techniques. The aim of this paper was to consider whether the SCM can provide a systems thinking approach and remain a viable option for accident analysis. To achieve this, the train derailment at Grayrigg was analysed with an SCM-based model (the ATSB accident investigation model) and two systemic accident analysis methods (AcciMap and STAMP). The analysis outputs and usage of the techniques were compared. The findings of the study showed that each model applied the systems thinking approach. However, the ATSB model and AcciMap graphically presented their findings in a more succinct manner, whereas STAMP more clearly embodied the concepts of systems theory. The study suggests that, whilst the selection of an analysis method is subject to trade-offs that practitioners and researchers must make, the SCM remains a viable model for accident analysis.     

An extension of the Human Factors Analysis and Classification System for use in open systems

The Human Factors Analysis and Classification System (HFACS), based upon Reason's model of human error in an organisational context, is currently the most widely used human factors accident analysis framework. However, it has been criticised for merely categorising accident data rather than analysing it. Previous research has established statistical associations between the levels and categories within HFACS but has not specified a mechanism by which one category influences subsequent behaviour. This paper extends the approach in two ways. Using the categories of control flaws derived from Leveson's Systems–Theoretical Accident Model and Processes (STAMP) approach, it describes the mechanisms by which categories within HFACS are associated with other categories lower in the organisational hierarchy. It also provides a mechanism by which active failures can promulgate across organisations. The revised methodology HFACS-STAMP is illustrated using the case study of the Uberlingen mid-air collision on 1 July 2002.     

The usual suspects? A novel extension to AcciMap using accident causation model tenets

Previous examination of the most widely used systems based accident causation models identified a series of core accident causation tenets. It is believed that these core tenets, referred to as ‘systems thinking tenets’, are the first step to a proactive approach to system safety. The article examines the Kimberly Ultramarathon fire and the extent that the systems thinking tenets can be applied using Rasmussen’s AcciMap technique. The findings suggest that indeed the tenets can be identified and further expose the specific system vulnerabilities that led to the Kimberly Ultramarathon accident. The tenets are a beneficial addition to the AcciMap technique providing the analyst a means to classify system properties leading to accidents. Implications for practice and future research steps are discussed.     

Application of a human error framework to conduct train accident/incident investigations

Accident/incident investigations are an important qualitative approach to understanding and managing transportation safety. To better understand potential safety implications of recently introduced remote control locomotive (RCL) operations in railroad yard switching, researchers investigated six railroad accidents/incidents. To conduct the investigations, researchers first modified the human factors analysis and classification system (HFACS) to optimize its applicability to the railroad industry (HFACS-RR) and then developed accident/incident data collection and analysis tools based on HFACS-RR. A total of 36 probable contributing factors were identified among the six accidents/incidents investigated. Each accident/incident was associated with multiple contributing factors, and, for each accident/incident, active failures and latent conditions were identified. The application of HFACS-RR and a theoretically driven approach to investigating accidents/incidents involving human error ensured that all levels of the system were considered during data collection and analysis phases of the investigation and that investigations were systematic and thorough. Future work is underway to develop a handheld software tool that incorporates these data collection and analysis tools.     

Have we reached the organisational ceiling? a review of applied accident causation models,methods and contributing factors in construction

Work-related fatalities continue to represent a significant issue within the construction industry. Contemporary accident causation models are underpinned by systems thinking, however, it is unclear whether these theories have translated into the construction domain. This article presents the findings of a review that was conducted to determine if the construction accident analysis literature has applied a systems thinking approach to understand accident causation. Specifically, the review examined two key aspects: first the types of models and methods that have been applied to analyse construction incidents, and second, the types of contributing factors identified as playing a role in construction incidents. The findings are summarised using Rasmussen’s Risk Management Framework. The review revealed that contemporary models of accident causation have not yet been applied in construction-related research. It is concluded that the models and methods applied in the construction literature predominantly identify contributing factors associated with the company, management, and frontline work levels of the system, rather than considering broader, system-wide factors. Further, the relationships between individuals and organisations operating at each level of the system have not been typically examined. Further research underpinned by systems thinking is required to better understand accident causation in the construction domain.

• Highlights

• The review revealed that contemporary models of accident causation have not yet been applied in construction-related research;

• The models and methods applied in the construction literature predominantly identify contributing factors associated with the company, management, and front line work levels of the system, rather than considering broader, system-wide factors;

• Relationships between individuals and organisations operating at each level of the system have not yet been examined;

• Despite the efforts of regulators and construction entities, the consistently high rate of fatal accidents in construction highlights the challenges surrounding the practical management of safe production within complex and dynamic working environments;

• Accident analysis methods used in this industry have not kept pace with advances in the field of safety science.

     

Routes to failure: analysis of 41 civil aviation accidents from the Republic of China using the human factors analysis and classification system

The human factors analysis and classification system (HFACS) is based upon Reason's organizational model of human error. HFACS was developed as an analytical framework for the investigation of the role of human error in aviation accidents, however, there is little empirical work formally describing the relationship between the components in the model. This research analyses 41 civil aviation accidents occurring to aircraft registered in the Republic of China (ROC) between 1999 and 2006 using the HFACS framework. The results show statistically significant relationships between errors at the operational level and organizational inadequacies at both the immediately adjacent level (preconditions for unsafe acts) and higher levels in the organization (unsafe supervision and organizational influences). The pattern of the 'routes to failure' observed in the data from this analysis of civil aircraft accidents show great similarities to that observed in the analysis of military accidents. This research lends further support to Reason's model that suggests that active failures are promoted by latent conditions in the organization. Statistical relationships linking fallible decisions in upper management levels were found to directly affect supervisory practices, thereby creating the psychological preconditions for unsafe acts and hence indirectly impairing the performance of pilots, ultimately leading to accidents.     

Understanding the human factors contribution to railway accidents and incidents in Australia

Forty rail safety investigation reports were reviewed and a theoretical framework (the Human Factors Analysis and Classification System; HFACS) adopted as a means of identifying errors associated with rail accidents/incidents in Australia. Overall, HFACS proved useful in categorising errors from existing investigation reports and in capturing the full range of relevant rail human factors data. It was revealed that nearly half the incidents resulted from an equipment failure, most of these the product of inadequate maintenance or monitoring programs. In the remaining cases, slips of attention (i.e. skilled-based errors), associated with decreased alertness and physical fatigue, were the most common unsafe acts leading to accidents and incidents. Inadequate equipment design (e.g. driver safety systems) was frequently identified as an organisational influence and possibly contributed to the relatively large number of incidents/accidents resulting from attention failures. Nearly all incidents were associated with at least one organisational influence, suggesting that improvements to resource management, organisational climate and organisational processes are critical for Australian accident and incident reduction. Future work will aim to modify HFACS to generate a rail-specific framework for future error identification, accident analysis and accident investigation.     

A systems approach to accident causation in mining: An application of the HFACS method

This project aimed to provide a greater understanding of the systemic factors involved in mining accidents, and to examine those organisational and supervisory failures that are predictive of sub-standard performance at operator level. A sample of 263 significant mining incidents in Australia across 2007–2008 were analysed using the Human Factors Analysis and Classification System (HFACS). Two human factors specialists independently undertook the analysis. Incidents occurred more frequently in operations concerning the use of surface mobile equipment (38%) and working at heights (21%), however injury was more frequently associated with electrical operations and vehicles and machinery. Several HFACS categories appeared frequently: skill-based errors (64%) and violations (57%), issues with the physical environment (56%), and organisational processes (65%). Focussing on the overall system, several factors were found to predict the presence of failures in other parts of the system, including planned inappropriate operations and team resource management; inadequate supervision and team resource management; and organisational climate and inadequate supervision. It is recommended that these associations deserve greater attention in future attempts to develop accident countermeasures, although other significant associations should not be ignored. In accordance with findings from previous HFACS-based analyses of aviation and medical incidents, efforts to reduce the frequency of unsafe acts or operations should be directed to a few critical HFACS categories at the higher levels: organisational climate, planned inadequate operations, and inadequate supervision. While remedial strategies are proposed it is important that future efforts evaluate the utility of the measures proposed in studies of system safety.     

Use of CCTV to determine road accident factors in urban areas

This paper sets out to assess whether there is a potential use for images collected through the increasingly ubiquitous use of CCTV cameras in urban areas as a means of increasing understanding of the causes of road traffic accidents. Information on causation and contributory factors is essential as a means of understanding why accidents occurred and how the occurrence of similar events may be prevented in the future. CCTV records of accidents could provide an independent perspective on an accident and have the potential to increase both the quality and quantity of information available to the safety researcher.

This study focuses on an area of central Leeds in the UK and shows that an existing CCTV camera system used for urban traffic management reasons has the potential to record around a quarter of the accidents which occur in the area, based on patterns of past occurrence. Most city centres in the UK will have similar camera systems set-up. By the introduction of additional strategically placed cameras and replacement of existing cameras with ones dedicated to accident recording, this figure could be increased substantially.

The paper also considers how effective cameras and video records will be as a means of identifying contributory factor information once an accident is recorded. The contributory factor classification used by a recently introduced system in Britain is assessed in terms of how visible each of the factors is likely to be on video and their relative frequency of occurrence. It is concluded that CCTV has a high potential to provide corroborative evidence about many of the most commonly occurring factors, and to throw further light on accident causation.     

Machine learning techniques applied to the determination of road suitability for the transportation of dangerous substances

This article describes a methodology to model the degree of remedial action required to make short stretches of a roadway suitable for dangerous goods transport (DGT), particularly pollutant substances, using different variables associated with the characteristics of each segment. Thirty-one factors determining the impact of an accident on a particular stretch of road were identified and subdivided into two major groups: accident probability factors and accident severity factors. Given the number of factors determining the state of a particular road segment, the only viable statistical methods for implementing the model were machine learning techniques, such as multilayer perceptron networks (MLPs), classification trees (CARTs) and support vector machines (SVMs). The results produced by these techniques on a test sample were more favourable than those produced by traditional discriminant analysis, irrespective of whether dimensionality reduction techniques were applied. The best results were obtained using SVMs specifically adapted to ordinal data. This technique takes advantage of the ordinal information contained in the data without penalising the computational load. Furthermore, the technique permits the estimation of the utility function that is latent in expert knowledge.     

Multidisciplinary perspective on accident investigation

The increasing complexity of many computer-controlled application processes is placing increasing demands on the investigation of adverse events. At the same time, there is a growing realisation that accident investigators must consider a wider range of contributory and contextual factors that help to shape human behaviour in the causes of safety-related incidents. A range of techniques have been developed to address these issues. For example (as we show in this paper), task modelling techniques have been extended from human computer interaction and systems design to analyse the causes and consequences of operator ‘error’. Similarly, barrier analysis has been widely used to identify the way in which defences either protected or failed to protect a target system from potential hazards. Many barriers fail from common causes, including misconceptions that can be traced back to early stages in the development of a safety-critical system. For instance, unwarranted assumptions can be made about the impact of training on operator behaviour in emergency situations. Similarly, barrier analysis can also be used before a system has been designed to inform the system model and make it more tolerant to errors by incorporating human and technical barriers into the design. Task models often uncover deep-rooted problems, for instance, in workload allocation across many different aspects of an interactive control system. It can be difficult to use barrier and task analysis to trace these common causes that lie behind the failure of many different defences. In order to deal with this complex combination of contributory factors and systems, we promote the use of abstraction (via models) as a way of representing these components and their interrelations whether it is design, construction or investigation. We use, to formally model an abstraction of the system. Additionally, the system model (described using a dialect of high-level Petri-nets) allows to reason about the system and to check conformance with the other models (task model, safety case and barriers). This paper, therefore, shows how an analysis of safety case arguments can be used to support the application of barrier, task, error and system analysis during the investigation of a command and control failure. The intention, in this paper, is to show that if an accident involved the failure of multiple barriers, it is also possible to trace the common causes of those failures back to the assumptions and arguments that are embodied within a safety case. Many countries require that safety cases demonstrate a system is ‘acceptably safe’ before they grant regulatory approval. These documents and the associated analytical techniques, therefore, provide a rich source of information about why command and control failures occurred. We demonstrate our approach on a fatal mining accident case study.     

Accident investigation reporting deficiencies related to organizational factors in machinery space fires and explosions

Careful accident investigation provides opportunities to review safety arrangements in socio-technical systems. There is consensus that human intervention is involved in the majority of accidents. Ever cautious of the consequences attributed to such a claim vis-à-vis the apportionment of blame, several authors have highlighted the importance of investigating organizational factors in this respect. Specific regulations to limit what were perceived as unsuitable organizational influences in shipping operations were adopted by the International Maritime Organization (IMO). Guidance is provided for the investigation of human and organizational factors involved in maritime accidents. This paper presents a review of 41 accident investigation reports related to machinery space fires and explosions. The objective was to find out if organizational factors are identified during maritime accident investigations. An adapted version of the Human Factor Analysis and Classification System (HFACS) with minor modifications related to machinery space features was used for this review. The results of the review show that organizational factors were not identified by maritime accident investigators to the extent expected had the IMO guidelines been observed. Instead, contributing factors at the lower end of organizational echelons are over-represented.     

A combined M5P tree and hazard-based duration model for predicting urban freeway traffic accident durations

The duration of freeway traffic accidents duration is an important factor, which affects traffic congestion, environmental pollution, and secondary accidents. Among previous studies, the M5P algorithm has been shown to be an effective tool for predicting incident duration. M5P builds a tree-based model, like the traditional classification and regression tree (CART) method, but with multiple linear regression models as its leaves. The problem with M5P for accident duration prediction, however, is that whereas linear regression assumes that the conditional distribution of accident durations is normally distributed, the distribution for a “time-to-an-event” is almost certainly nonsymmetrical. A hazard-based duration model (HBDM) is a better choice for this kind of a “time-to-event” modeling scenario, and given this, HBDMs have been previously applied to analyze and predict traffic accidents duration. Previous research, however, has not yet applied HBDMs for accident duration prediction, in association with clustering or classification of the dataset to minimize data heterogeneity. The current paper proposes a novel approach for accident duration prediction, which improves on the original M5P tree algorithm through the construction of a M5P-HBDM model, in which the leaves of the M5P tree model are HBDMs instead of linear regression models. Such a model offers the advantage of minimizing data heterogeneity through dataset classification, and avoids the need for the incorrect assumption of normality for traffic accident durations. The proposed model was then tested on two freeway accident datasets. For each dataset, the first 500 records were used to train the following three models: (1) an M5P tree; (2) a HBDM; and (3) the proposed M5P-HBDM, and the remainder of data were used for testing. The results show that the proposed M5P-HBDM managed to identify more significant and meaningful variables than either M5P or HBDMs. Moreover, the M5P-HBDM had the lowest overall mean absolute percentage error (MAPE).     

Analytical HFACS for investigating human errors in shipping accidents

Despite the innovative trends in marine technology and the implementation of safety-related regulations, shipping accidents are still a leading concern for global maritime interests. Ensuring the consistency of shipping accident investigation reports is recognized as a significant goal in order to clearly identify the root causes of these accidents. Hence, the goal of this paper is to generate an analytical Human Factors Analysis and Classification System (HFACS), based on a Fuzzy Analytical Hierarchy Process (FAHP), in order to identify the role of human errors in shipping accidents. Integration of FAHP improves the HFACS framework by providing an analytical foundation and group decision-making ability in order to ensure quantitative assessment of shipping accidents.     

Relationships between accident investigations, risk analysis, and safety management

Several different approaches to achieve safety are in common use, and examples are accident investigations (AI), risk analysis (RA), and safety management systems (SMS). The meaning of these concepts and their practical applications vary quite a lot, which might cause confusion. A summary of definitions is presented. A general comparison is made of application areas and methodology. A proposal is made how to indicate parameters of variation. At one end of the scale there are organisations, which are highly organised in respect to safety. At the other end are small companies with informal safety routines. Although the three concepts differ in a number of respects, there are many links between them which is illustrated in a model. A number of relations have been described mainly concerned with more advanced organisations. Behind the practical safety work, there are varying sets of more or less explicit explanations and theories on safety and accident causation. Depending on the theory applied, the relations between approaches can be more or less clear and essential.     

Investigations of Human and Organizational Factors in hazardous vapor accidents

This paper presents a model to assess the contribution of Human and Organizational Factor (HOF) to accidents. The proposed model is made up of two phases. The first phase is the qualitative analysis of HOF responsible for accidents, which utilizes Human Factors Analysis and Classification System (HFACS) to seek out latent HOFs. The hierarchy of HOFs identified in the first phase provides inputs for the analysis in the second phase, which is a quantitative analysis using Bayesian Network (BN). BN enhances the ability of HFACS by allowing investigators or domain experts to measure the degree of relationships among the HOFs. In order to estimate the conditional probabilities of BN, fuzzy analytical hierarchy process and decomposition method are applied in the model. Case studies show that the model is capable of seeking out critical latent human and organizational errors and carrying out quantitative analysis of accidents. Thereafter, corresponding safety prevention measures are derived.     

What you find is not always what you fix—How other aspects than causes of accidents decide recommendations for remedial actions

In accident investigation, the ideal is often to follow the principle “what-you-find-is-what-you-fix”, an ideal reflecting that the investigation should be a rational process of first identifying causes, and then implement remedial actions to fix them. Previous research has however identified cognitive and political biases leading away from this ideal. Somewhat surprisingly, however, the same factors that often are highlighted in modern accident models are not perceived in a recursive manner to reflect how they influence the process of accident investigation in itself. Those factors are more extensive than the cognitive and political biases that are often highlighted in theory. Our purpose in this study was to reveal constraints affecting accident investigation practices that lead the investigation towards or away from the ideal of “what-you-find-is-what-you-fix”. We conducted a qualitative interview study with 22 accident investigators from different domains in Sweden. We found a wide range of factors that led investigations away from the ideal, most which more resembled factors involved in organizational accidents, rather than reflecting flawed thinking. One particular limitation of investigation was that many investigations stop the analysis at the level of “preventable causes”, the level where remedies that were currently practical to implement could be found. This could potentially limit the usefulness of using investigations to get a view on the “big picture” of causes of accidents as a basis for further remedial actions.     

Representing context, cognition, and crew performance in a shutdown risk assessment

This paper presents a potentially practical treatment of dynamic operator-system interactions. The approach employs a dynamic event tree framework to explicitly address plant dynamics, system indications, crew-plant interactions, time available, crew cognition, errors of commission (mistakes), and multiple planning and diagnosis possibilities. The approach is applied in an analysis of a hypothesize medium break loss of coolant accident for a test reactor that occurs during plant shutdown. This transient was selected on the basis of: a significant cognitive component being present, high consequences being associated with operator actions, and the importance of event timing to scenario progression. The results presented show how quantitative risk predictions are affected by the treatment of dynamics, and demonstrate how non-proceduralized recovery actions and a number of performance shaping factors (e.g., crew experience, stress, and confidence) can be explicitly treated. Insights and lessons learned regarding the performance of a dynamic assessment are also presented.