A generic inference engine for alarm filtering in automated production systems

The goal of this paper is to propose new alarm filtering techniques for Intelligent Alarm Processing Systems (IAPS). A basic idea of our approach of alarm filtering is to use a type of knowledge model (the functional graph) to define synthesis alarms. Using causal inhibition principle, detail alarms are inhibited by synthesis alarms to reduce the number of alarm presented to the operator. Moreover, we propose a hierarchy of concepts to characterise the status of an alarm. This hierarchy enable us to propose integrated filtering techniques that take into account both filtering due to system properties (validation, causal inhibition) and the interaction of the operator with the alarm system to sort presented alarms (acquitement, suppression).