An Adaptive Optimization Approach To Process Alarm Thresholds Concerning Human Factors

The alarm management system can not only detect the abnormal state of the process productions, but also provide effective information for the operator to facilitate the processing, so as to reduce the incidence of accidents. However, an intractable problem of current industrial process production is acknowledged as numerous nuisance alarms which are identified to be possibly caused by unreasonable process alarm thresholds.At present, most process alarm thresholds optimization methods are rarely concerned with the influence of operators' perception and treatment capacities. In response, a novel adaptive optimization approach to alarm thresholds concerning human factors is introduced in this paper.Firstly, APRIORI algorithms are used to extract association rules between process alarms and treatment operations from historical data.Based on on-line recorded data, the nuisance alarm rate and the alarm treatment rate are calculated in terms of operators' perception and treatment capacities respectively which lay the foundation for quantitative human factor index.Secondly, human factor index is obtained through employing the nuisance alarm rate and the alarm treatment rate as the inputs of fuzzy reasoning strategies. An objective function for thresholds optimization is established in terms of the weighted FAR(False alarm rate) and MAR(Missed alarm rate), in which, kernel density methods are employed to estimate FAR and MAR, while weights are calculated based on human factor index.Finally, the TE process is employed for the application research. The comparison between the proposed and traditional methods in terms of alarm thresholds setting reveals that the proposed adaptive alarm thresholds optimization method enjoys better environmental adaptability,as well as potentially benefits the performance of alarm systems.