Graphical Representation of Industrial Alarm Data

Combining Knowledge Modeling and Machine Learning for Alarm Root Cause Analysis

“…Alarm flooding is an extraordinary plant state, where the rate of incoming alarms exceeds the human reception capacity; the operator is overtaxed and as a result not able to make diagnoses and take actions in required response time Izadi et al (2009); Kondaveeti et al (2010). Sequential structures of alarms can be modelled with causal networks to analyze how such alarm floods evolve.…”

A method for pattern mining in multiple alarm flood sequences

“…In (Duan et al, 2014b) methods for root cause diagnosis of plant-wide oscillations were summarized and compared, of which data-driven causality analysis as an important branch was reviewed. Alarm correlation analysis methods were also proposed in (Kondaveeti et al, 2010;Nishiguchi and Takai, 2010;Yang et al, 2012Yang et al, , 2013; these methods could help restore the connections between alarm tags directly without using process data.…”

“…Usually a large proportion of an alarm flood are nuisance alarms, of which chattering alarms form an important part. Methods such as high density alarm plots, calculation of a chattering index, delay-timers, and dead-bands can be used to visualize, quantify, and reduce such chattering alarms (Kondaveeti et al, 2010(Kondaveeti et al, , 2013Izadi et al, 2009). However, by applying delay-timers and dead-bands, often one cannot totally suppresses alarms during alarm floods; the remaining alarms are mainly consequence alarms, which can be caused by three reasons: (1) process state changes such as start-up and shutdown, (2) bad alarm configurations such as redundant measurements on a single process and (3) causal relationships among measured variables.…”

A Novel Approach to Alarm Causality Analysis Using Active Dynamic Transfer Entropy

“…In the past two decades, the process modeling, process monitoring, and control systems, such as Distributed Control System (DCS) and Supervisory Control and Data Acquisition (SCADA), have evolved to incorporate various software tools for this purpose. , As a result, the performance of industrial alarm systems has improved significantly. However, in practice, alarm flooding problems occur due to an excessive number of alarms. − A typical alarm flooding problem includes invalid and repeated alarms, making it difficult for operators to prioritize and separate critical alarms. Alarm flooding problems can be resolved by finding the root-cause of alarms from many correlated alarms using causality analysis and fault diagnosis methods. , …”