Establishing Confidence in Safety Assessment Evidence

With the increased complexity and higher safety commitment of modern safety–critical systems, safety assessment models of these systems are increasingly complicated and obscure. In practice, however, there is insufficient guidance on how to improve the understanding and evaluation of these models, while they are often used as important items of evidence in safety cases. This significantly threatens the confidence we can have in the soundness of safety cases.
In this thesis, a coherent, structured approach to establishing confidence in safety assessment evidence is developed. Firstly, a means for the structured documentation of the core data elements of safety assessment models is defined, to support the development of both primary safety arguments and confidence arguments. Secondly, a model of evidence is developed to support the interfacing of safety assessment evidence with safety arguments. Thirdly, a
structured cross-model inconsistency analysis method is proposed as a means of scrutinizing potentially inadequate models. Finally, an expanded argument construction process is established to add rigour to safety case development, and a number of argument patterns are designed to guide and inspire structured justification of the adequacy of safety assessment models as evidence for safety critical systems.
The evaluation of the approach is carried out primarily through examples and cases studies. It is demonstrated that the approach is feasible and the confidence issue in safety assessment evidence is addressed more explicitly and more rigorously by using the approach.