DREAM: A Method for Understanding the Causation of Single-Vehicle Crashes
Licentiate thesis, 2006
When conducting in-depth causation studies, it is essential that the analysis method can both define and classify contributing factors and be able to analyse how these factors may interact to produce a critical event. These issues are highly influenced by the underlying theory, e.g. the accident model. Consequently, the first study in this thesis surveyed the characteristics of four general accident models, i.e. sequential models, epidemiological models, energy transfer models and systemic models.
Based on the accident model survey, this thesis investigated the characteristics of accident models applied in five previous accident causation studies, three statistical studies and two case studies. In the investigation, it was found that the statistical studies had not defined an accident model. This together with other methodological deficiencies resulted in the misleading conclusion that human factors are responsible for up to 70-95% of traffic accidents.
The two case studies defined detailed accident models, forming the base for their causation analysis methods. However, because of an implemented driver model based on the information-processing concept, the accident models received features resembling the disadvantages of sequential accident models. As a result, every accident was analysed and explained always using the same structure of causation. Further, the results in the form of case reports provided limited possibilities to compare similar accidents and see common causation characteristics, e.g. through comparing contributory factors and their interactions.
In the second study of this thesis, a promising method for analysing causation, Driving Reliability and Error Analysis Method (DREAM), was presented and demonstrated in an in-depth investigation of 38 single-vehicle crashes (SVCs). DREAM is based on the principles of systemic accident models. Consequently, DREAM consists of a factor and a linkage structure that is flexible enough to analyse the various types of causation factors that together may contribute to an accident. The result of a DREAM analysis is illustrated as an accident pattern for each accident case.
The individual SVCs were compared to distinguish cases with similar crash circumstances. As a result, causation patterns were combined into four scenarios that demonstrate the most frequently occurring factors and links. The four scenarios described in greater detail the factors that contribute to accidents characterised by road departure, slipperiness, excessive speed or excessive driver manoeuvres.
Because of the flexible structure, the sequence of a DREAM analysis may first appear abstract and therefore difficult to immediately understand. Consequently, there is a risk that analysts with different competence may produce diversified results. However, this risk will likely be reduced with more distinct factor definitions, education and experience.