A Reinforcement Learning-Based Approach for Early Prediction of Industrial Alarm Floods

An industrial alarm flood indicates the emergence of a major problem and requires immediate and effective measures. Without contextual guidance during an alarm flood, it is difficult to respond efficiently, especially when the alarm rate is significantly high. Thus, a reinforcement learning (RL) based approach is proposed in this work for early prediction of an industrial alarm flood, which can assist industrial operators in identifying the root cause and taking corrective actions at an early stage. The proposed method is mainly implemented in the following steps: 1) An association rule mining approach is employed to eliminate irrelevant alarms and extract alarm patterns from the historical alarm flood sequences; 2) The early prediction problem is framed as a partially observable Markov decision process (POMDP), and the double deep Q-network (DDQN) algorithm is adopted to ensure both accuracy and earliness. The effectiveness of the proposed method is demonstrated using real industrial data from an oil refinery.