Idustrial plant operating watching several computer screens. Photo.

Machine learning for predicting chattering alarms

Alarms systems play a vital role to ensure safety and reliability in the process industry. In a process plant environment, operators are notified by alarms if a process diverges from normal operating conditions. 

Alarms are triggered to warn the operator about irregular incidents. When several alarms go off simultaneously it is known as “alarm floods”, and the operator may miss the crucial alarms. Ideally, an alarm should therefore inform the operator about critical conditions only and provide guidance to a set of corrective actions associated with each alarm.

Associate Professor Nicola Paltrinieri at the Department of Mechanical and Industrial Engineering at NTNU and his colleagues recently benefited from this AUS project. Three scientific articles were recently published based on research conducted via the NIRD Service Platform and the Deep Learning Toolkit. An Advanced User Support (AUS) project contributed when a new method for assessing dynamic chattering alarms was developed. The results were used to train and evaluate a Deep Neural Network. The model was then tested against the ability to predict alarm chatter.

A modified approach based on run lengths distribution was developed to evaluate the likelihood of future alarm chatter. The method has allowed categorizing historical alarm events as alarms that will (or will not) show chattering in the future. Finally, categorized alarms have been used to train a Deep Neural Network by using TensorFlow, whose performance
has been evaluated against the ability to predict alarm chatter.