Accident analysis framework

A generalised process model constitutes the backbone of our accident-analysis framework in Chapter 6. Process models help us in understanding how a production system gradually deteriorates from a normal state into a state where an accident occurs. Time is thus a basic factor. In contrast to causal-sequence models, process models make a clear distinction between the accident sequence on the one hand and the underlying causal or contributing factors on the other hand. 

We will later apply the accident-analysis framework in a review of different types of methods used in the collection and analysis of data of accident risks. We will start at the output side of the model by reviewing the different types of classification systems used to document the consequences of accidents and different measures of loss. We will then continue by looking into the classification systems used to document incidents and deviations. Finally, we will review the different classification systems for contributing factors and root causes. Our aims will be twofold first, to be complete, i.e. by presenting all alternative means of measuring and classification, and second, to give specific advice on the preferred method. The reader will find recommended alternatives in shaded tables and checklists. 

Table 6.1 Application of the accident-analysis framework to illustrate the development of different types of losses...

The observant reader will note that the four types of accident experience have their counterparts in the different elements of the accident analysis framework of Figure 6.4. When we move upwards in Van Court Hare s hierarchy, we can also expect that the measures will involve a higher level in the causal-factor hierarchy. 

Figure 12.1 Overview of different means of collecting data on accident risks and their ideal scope seen in relation to the accident-analysis framework of Chapter 6.

We now move further to the left in the accident-analysis framework (Figure 16.1). We will look into SHE performance indicators based on information about contributing factors and root causes. The organisation and SHE management system are in focus. These indicators have many similarities to the audit methods described in Section 14.2. 

Figure 21.2 illustrates how the starting point, the directions and the scope of each method fit into the accident-analysis framework of Chapter 6. Two of the methods. Fault tree analysis and Comparison analysis are deductive in that they start with the unwanted event. They proceed by analysing the underlying incidents and deviations (Fault tree analysis) or contributing factors (Comparison analysis). Several of the methods are mainly inductive in that they start with a deviation and proceed by studying the effects of this deviation. This applies to HAZOP, Failure mode and effect analysis. Event tree analysis and CRIOP, although they also have a component of causal analysis. Coarse analysis and Job-safety analysis start with the hazard and use a combination of inductive and deductive analyses. 

The accident-analysis framework of Chapter 6 provides a basis for tying together information from these different activities, see Section 15.1. 

As an alternative, the current paper presents an approach for analysis of aviation incidents that takes a multi-agent perspective, and is based on formal methods. The approach is an extension of the approach introduced in the work of Bosse and Mogles [4], which was in turn inspired by Blom, Bakker, Blanker, Daams, Everdij and Klompstra [1]. Whereas this approach mainly focuses on the analysis of existing accidents (also called accident analysis or retrospective analysis), the current paper also addresses analysis of potential future accidents (called risk analysis or prospective analysis). This is done by means of a multi-agent simulation framework that addresses both the behaviour of individual agents (operators, pilots) as well as their mutual communication, and interaction with technical systems. By manipulating various parameters in the model, different scenarios can be explored. Moreover, by means of automated checks of dynamic properties, these scenarios can be assessed with respect to their likelihood of the occurrence of accidents. The approach is illustrated by a case study on a runway incursion incident at a large European airport in 1995. 

An accident analysis technique should provide a framework or process to assist in understanding the entire accident process and identifying the most important systemic causal factors involved. This chapter describes an approach to accident analysis, based on STAMP, called CAST (Causal Analysis based on STAMP). CAST can be used to identify the questions that need to be answered to fully understand why the accident occurred. It provides the basis for maximizing learning from the events. 

Reinach, S. Viale, A. 2006 Application of a Human Error Framework to Conduct Train Accident/ Incident Investigations. Accident Analysis Prevention, 38, 396-406. 

It is worth to notice that this work applies a retrospective method for data collection based on CREAM taxonomy, but certainly not following the procedures developed by Hollnagel to perform accident analysis. It means that the CREAM retrospective technique is not used whatsoever, considering that the accidents collected were submitted to extensive investigation and the causes and contributing factors are already exposed in the reports, being the current gap the need for interpretation and classification of the data under a common framework. 

How to analyse the causes of accidental events involving releases of pollution Chapter 5 gives an overview of accident models that have been developed to support analyses of this type. Chapter 6 presents one particular analysis framework that suits this purpose. Chapter 8 focuses... 

Table 15.1 presents a proposed smallest efficient data set on accidents and near accidents. It is based on the framework for accident analysis presented in Chapter 6 and takes the author s experiences from evaluations of SHE information systems into consideration. In this proposal, no additional information is requested from the supervisor other than that collected on a traditional accident-investigation form. The reason why a SHE expert is needed to feed additional data is to secure the reliability of the data. In addition, the SHE expert will be responsible for checking the quality of the data from the supervisors. 

Figure 16.1 presents an overview of different SHE performance indicators. This overview is based on the framework for accident analysis in Chapter 6. Loss-based SHE performance indicators will be our starting point. Among these we find the most commonly used indicator, the lost-time injury frequency rate. We proceed by reviewing process-based indicators, similar to those developed in the fertiliser-plant case in Chapter 4. Finally, we will look into indicators relating to causal factors, i.e. indicators based on information about the organisation and SHE management system. 

The first three methods analyse the risk of occupational accidents. They have a joint theoretical basis in the framework for accident analysis described in Chapter 6. The methods serve different purposes. Often, Coarse and Job-safety analysis are used in combination by starting with the Coarse analysis. A Job-safety analysis will follow when severe hazards have been identified and there is a need to go into details on how they may result in harm. We will... 

Reinach, S., and A. Viale. 2006. Application of a human error framework to conduct train accident/incident investigations. Accident Analysis and Prevention 38 396-406. 

As the previous ehapter discussed nuelear power reactor operation and how to perform a PSA on it, this chapter attempts to apply a similar framework to chemical processing. The problem is the diversity of chemical processing that blurs the focus. This chapter begins by showing that accidents in the chemical process industry cost lives and dollars. Descriptions of deadly chemical accidents arc presented to show the chain of sequences that were involved to suggest how their PSA may be structured. Background on selected hazardous chemical process is presented followed by descriptions of how their PSA have structured. The chapter concludes by applying FTAPSUIT to a pressure vessel rupture analysis. 

Several examples have already been provided of the use of cognitive models of error to evaluate the possible causes of accidents that have already occurred. This form of retrospective analysis performs a vital role in providing information on the recurring underlying causes of accidents in which human error is implicated. The advantage of an analytical framework driven by a model of human error is that it specifies the nature of the questions that need... 

---