Alarm Management for Batch Processes

One of the challenges of implementing effective alarm management is in dealing with batch processes. Batch processes are typically characterized by a batch lot number and consist of several process steps/phases, most or all of which involve non-steady state, nonlinear, time varying operations.

TiPS, Inc. is pleased to have a guest subject matter expert authoring this blog, Joseph Alford, Ph.D, CAP, who has been involved in alarm management throughout his 35 year automation career in the pharmaceutical industry. Joe has often led initiatives that created novel solutions for batch processes in order to meet customer requirements. He has published four alarm management articles in various technical journals since 2005, and is a co-author of the ANSI/ISA 18.2 standard on Management of Alarms for the Process Industries. Most recently, he co-chaired the ISA committee that published a technical report (ISA-TR18.2.6-2012) titled “Alarm Systems for Batch and Discrete Processes.”

Individual process alarms are usually not relevant for all batch steps and so need to be programmatically suppressed under appropriate conditions. When alarms are enabled, some of their attributes, e.g., setpoints (i.e., limits), priority, and category (i.e., safety, environmental, product quality, etc.) may be functions of specific process steps and sometimes may even change within a step. Also, the control rooms for batch processes are often not continuously manned due to manual operations occurring in the field, leading to requirements for remote alarming and alarm acknowledgement.

When asking plant support staff what user requirements they have of their alarm system(s), a common response includes one that alarm records exist that can be queried and sorted based on batch lot number, process step/phase, alarm category, and/or relative time (instead of or in addition to calendar time). Further, automated utilities should be provided to, e.g., automatically generate and present Pareto charts of the most frequently occurring alarms. Another common desire is for alarm records that contain intelligible descriptors rather than cryptic loop tags such as FIC101.

Not many industrial plants have automation systems that provide all of the above functionality. As a result, auditors often find files and/or databases in historians containing thousands of alarm records, with little use of this data occurring among plant support personnel. These records frequently remain as unutilized data, with little effort to extract its information and knowledge content. Lack of data mining tools is not the only concern. Even when appropriate tools exist, users can be hesitant to deal with alarm record databases since so many of the alarm records represent nuisance alarms (i.e., NOT representing an abnormal situation requiring an operator response). Nuisance alarms can be especially prevalent for batch processes, due to the many time varying phases and transitions between them. Dynamic alarming is needed for many batch processes, while more static alarming is often satisfactory for continuous processes operating at near steady state.