Situation Awareness – A Primer for Process Operations

This title, Situation Awareness – A Primer for Process Operations, caught my eye and decided to end the day in this session. The abstract:

As more and more emphasis is placed on supporting operator performance in the process industries, it is important to understand how concepts from outside process control can be used to inform improvement efforts. Developed primarily in the context of aviation and military applications, Situation Awareness (SA) is becoming a popular term when discussing monitoring in process control settings. This session will provide some background on the concept of situation awareness and takes a critical look at how it can be used to provide a new perspective on improving process operations.

An informal definition of Situational Awareness (SA) is to know what’s going on. The origins go back to World War I, but the concept took off in the 1980s when human factors research came into increased focus.

Situational awareness matters because decisions are mostly situation dependent. It’s rare that they are pure rule based– if X then Y. Poor situational awareness has led to many industrial accidents. SA is also linked to productivity in routine operations.

SA leads directly to decision making. The levels of SA are What? So what?, and What’s Next? The goal of process monitoring is how optimally operators can sample available data to make a decision. It’s not about knowing all things at all times, but knowing the right things at the right time.

Level 1 SA is about knowing relevant data as it becomes available. Analog presentation formats provide better visual processing than digital formats. Salience mapping is giving more visual priority to important stuff than lesser stuff. The lesser can be present, but faded into the background. Practice good visual hygience and alarm sound design.

The biggest human performance bottleneck is attention. Our brains aren’t designed to multitask. Level 1 SA supports control of attention. This is done through the design of alarms and alerts and from the top down provide big picture overviews and operator configurable display areas.

Level 2 SA is where operators make sense of the information they’ve received. Think of putting the pieces together. Operating context is critical. What are the production priorities, modes, activities, and other soft (non sensor derived). Next is the operators mental model. The mental model is an operator’s representation of the process. This is the basis of open-loop, predictive control and the basis for diagnosis of problems.

Schema are representation of patterns of data values or process behaviors. Recognizing the situation is often the hardest part. Making correct decisions and taking actions usually follow correct recognition. It’s important to support pattern recognition and recognition of patterns in alarms.

An example was presented showing four process variables on a distillation column. They were on an axis where they made a square when operating at the correct levels. The square skewed to show upset conditions such as column flooding.

Think about automation as a team player–make it observable, directable. Provide support for dealing with uncertainty.–additional related information, relying on defaults, bet-hedging/contingency planning, narrowing/safe-states, thresholding.

Support teamwork to understand what others are doing and why. Show others’ assessments to help cross-check assessments and decisions.

3 comments

  1. Jonas Berge says:

    Pervasive sensing by deploying a second layer of automation for missing measurements beyond the P&ID provides a way to add important measurements and feedback to the operators to improve situational awareness. As such, enabling them to make better informed decisions based on actual information rather than having to deduce or infer information. Transmitters using WirelessHART technology can easily be installed in older plants. Read on…
    http://www.iaasiaonline.com/more.php?id=2222&cat=im

  2. Scott Turner says:

    I guess that statistical process control fits nicely into this, especially multivariate statistical process control. Using statistical process control you can see how your overall process is deviating using eliptical action and warning limits on scores plots.I imagine geometric modelling is also useful for pattern recognition.

Leave a Reply