Alarm correlation mining is a key part of alarm management which plays an important role for chemical process safety. According to the traditional alarm correlation mining algorithms which is based on hierarchical clustering schemes

this paper proposed a non-regular alarm correlation calculation method to make up the deficiency that the traditional algorithms have

in dealing with the situation of delayed time change and asymmetry of alarms. Moreover

the value of alarm correlation was determined in the form of probability

which made the alarm correlation among different tags comparable. Furthermore

case studies based on simulation and real industrial process data were performed to demonstrate the effectiveness of the proposed methods in mining alarm correlation.