FLINK STREAMING PROCESSING SYSTEM FOR CONDITION MONITORING AND DIAGNOSIS

In order to enhance the ability of fault diagnosis based on data streaming collected as equipment operating, Flink streaming processing systemis designed and developed. The distributed parallel processing flowgraph was designed for multi-channel sensory data. The operators of the flowgraph were developed. Therefore, Flink application transformed the input streaming into diagnosis output streaming. The Flink clusters were built to carry out functional and performance testing. The results indicate that the delay time is less than 1 ms, data integrity rate is 100%, disorderly data restoration rate is 100%, and the recovery time for job manager fault and task manager fault are 62 s and 26 s respectively. The Flink-based system satisfies the requirements of low latency, exactly-once consistency, and fault tolerance.