Real-Time Industrial Process Fault Diagnosis Based on Time Delayed Mutual Information Analysis

Causal relations among variables may change significantly due to different control strategies and fault types. Off line-based knowledge is not adequate for diagnosis, existing causal models obtained from data driven methods are mostly based on historical only. However, variable correlation would remain identical, could be very under certain industrial operation conditions. To deal with this problem, a diagnosis framework proposed information solely extracted process data. By method, mutual (MI) between each pair of first calculated obtain thresholds using data, as normal conditions contributed by random noises, which often neglected in analysis models. Once deviation detected, beyond these further investigated time delayed (TDMI) current so determine the logic them, represented propagation paths, can tracked all way back root cause. The method applied simulated Tennessee Eastman process. results show that difference diverse or response captured real time, path objectively identified, together Then, has been successfully whole year an process, proves feasibility application.