Human Performance Models for Response to Alarm Notifications in the Process Industries: An Industrial Case Study

Abstract

The Engineering Equipment and Materials User Association (EEMUA) recently published recommendations for average and peak alarm rates of 1 alarm per 10 minutes during normal operations and no more than 10 alarms per 10 minutes following an upset condition, respectively. However, these recommendations have been made in the absence of human performance studies or theory. This paper presents two studies –an analytical Keystroke-Level Modeling study and a Markov modeling study– that provide an initial human performance context for the EEMUA recommendations. Results suggest that these EEMUA recommendations are in line with human performance limits. Several limitations of the two studies for generalizing from either the KLM or Markov modeling results are discussed. Practical implications presented include the need for advanced alarm reduction techniques and consideration of peak alarm rates for upset conditions during the alarm rationalization process. Future research directions are also discussed, such as establishing the duration for which the EEMUA recommendation for peak alarm rates can be endured by a human operator.