Design of incident message analysis methods for a life safety automated control system

The life safety automated control system’s software must consolidate several reports of the same incident with an aim to form generalized operational view of the crisis. This task is often solved with the imperative rules by an alarm reports consolidation, or with the use of machine learning or neural networks principles. A decision-making architecture with imperative rules is extremely resource intensive, in contrast with the last two methods that are able to reduce the amount of analytical tasks. However, the drawback of the machine learning and neural networks is the impossibility to prove the correctness of the decisions offered by the decision-making system, which is applied at the situational center. The proposed method is based on the operator’s decisions about the message association with the incident. The system automatically forms the assumptions of the logic rules for this association. These hypotheses then will be assessed by the system with gaining or losing confidence level. The final estimation concerning confidence level of hypothesis will be correlated with the subsequent operator’s decisions and refined by generalization, i.e. discarding of excessive wide preconditions or by narrowing down decision context, i.e. by adding required preconditions. The resulting set of the hypotheses formalized using Web Ontology Language is presented to the expert for approving or rejecting. The positive effect is to reduce the experts’ scope of work, since experts have only to assess the automatically formed hypothesis instead of defining many imperative rules.