Research And Development Of Distributed Intelligent Alarm Management System In Process Industry

With the continuous expansion of process industry and overuse of alarm points, the number of alarms increases exponentially. Repetitive alarms, invalid alarms and nuisance alarms occur in great numbers, which mask critical alarms and increases the probability of production accidents. In order to solve this problem, enterprises use alarm management system to evaluate and analyze alarms, so that operators can deal with alarms more efficiently. However, due to the increase of alarms, the traditional alarm management system based on relational database cannot suppress alarm overload effectively, facing severe challenges such as computing performance, storage capacity and effectiveness of alarm strategy.The distributed intelligent alarm management system applied in process industry was proposed according to the requirements of alarm management. It was built upon the distributed massive alarm storage and computing framework based on Hadoop. Meanwhile, efficient alarm management functions like alarm statistics, performance evaluation and alarm suppression were implemented. Storage capacity and query performance of alarm management system have been improved, while a new way of data mining in massive alarms is provided in current research. The main research results are as follows:(1)Alarm transmission mechanism was designed to solve the problem of heterogeneous system integration.(2) By utilizing features of Hadoop such as scalability and distribution, combined with reasonable design of column storage structure, distributed storage of mass alarm data was achieved.(3) According to the characteristics of massive alarm real-time query, a distributed parallel query model based on Coprocessor was proposed and a data mining model based on MapReduce was designed.(4) Mechanisms of alarm management and performance appraisal were proposed based on actual situation of alarm application.(5) Based on the characteristics of alarms and time-delay correlation of alarm occurrence, alarm similarity criteria which integrated time-delay correlation algorithm and attribute correlation algorithm were proposed to effectively identify homologous alarms. In order to suppress massive alarms, a new distributed dynamic alarm algorithm was designed to work with data mining model in parallel.