Improvement of Hazard Identification in Railway Software
Examensarbete för masterexamen
This thesis deals with the problem of creating more complete, less ambiguous HAZOP deviations. In order to conquer this task we develop a model for actions. An action is a process that changes at least one attribute of at least one object. The actual model is split into three sub models. The first one of them describes what objects are involved in an action what role they play. The second one analyses the conditions that must be satisfied for the action to start (precondition) and to come to an end (postconditions). Lastly we present the relation between actions regarding time. So parameter such as start time, end time, number of concurrent actions and so on are defined. Since not all HAZOP guide words create reasonable deviation we also give a framework defining what guide words can be applied to which part of each model. Our model creates a lot of HAZOP deviations, so it is impossible to take all of them into account. Therefore, we first eliminate redundancies, and also reduce the number of pairwise comparisons (comparison that deal with the interaction of two actions). This is done by using a similarity measure, that counts the number of commonly accessed objects of each two actions. We also give a grammar of our model, and compare this to a HAZOP deviation model that is based on objects in steady state. The comparison has shown that both grammars are equal, since the object grammar allows creation of arbitrarily many attributes, and the action grammar has no structure on its definition of conditions. Our model is evaluated on a new monitoring system for bogies in trains. The model created diminishes the time that is needed for a HAZOP, since the meeting and, therefore, the lengthy brainstorming session becomes dispensable. Furthermore, the systematic approach to create deviations is far more complete.
Information Technology , Informationsteknik