Alarm clustering analysis and ACO based multi-variable alarms thresholds optimization in chemical processes

In chemical process, excessive alarms, high false alarm rate (FAR), and high missed alarm rate (MAR) generated by unreasonable setting to variable alarm thresholds are the main causes of affecting operation stability and device safety. In this paper, a clustering analysis based method was proposed to optimize the variable alarm thresholds. Variables are first clustered into groups using standardized Euclidean distance before variable weights are given by entropy weight method. Second, the probability density functions of the variables are fitted with process data under normal and abnormal conditions. An objective function about the FAR, MAR, and average alarm delay (AAD) is then established with variable weight and alarm delay. Finally, the objective function is optimized to find the optimal alarm thresholds using ant colony optimization (ACO) method. Case study of an industrial atmospheric-vacuum crude distillation shows that the proposed method can effectively reduce FAR and MAR.