Leverage human strengths in operational design

Have machines do what they do well, and do the same for your human operators.

By Jim Montague

CT1610 Yokogawa Header15

In the search to optimize process operations, it's become painfully obvious that one of the main constraints on efficiency and productivity is their human operators. This depressing and ironic situation may seem inescapable, especially when people interact with increasingly complex systems, faster operations, alarm floods, staff limits and declining experience levels.

Despite these persistent obstacles, there are remedies and ways to alleviate negative human impact, according to David Strobhar, P.E., chief human factors engineer at Beville Engineering Inc., a founding member of the Center for Operator Performance. Strobhar presented "Human Factor, Limiting Factor" this week at the Yokogawa Users Conference and Exhibition 2016 in Orlando.

"In human-machine systems, it's important to have people do what they do well, and have machines do what they do well," said Strobhar. "So have people recognize patterns, detect stimuli against high noise backgrounds, sense unusual and unexpected events, remember principles and strategies, reason, and continue to function even when overloaded or presented with incomplete information. Conversely, have machines sense extreme stimuli, monitor for pre-specified events, store and retrieve coded information, make rapid and consistent responses to signals, exert force, and perform repetitive actions over long periods."

Balancing workload and performance

Because inexpensive alarm functions proliferated in process control systems in recent years, Strobhar reported that classic alarm floods also have multiplied. "Workloads increase quickly with upsets like alarm floods," said Strobhar. "However, operators can only detect about 75 alarms per minute and can only process about 25 per minute, so even limiting low-priority alarms typically leaves too many for them to process. The traditional response was simply adding more operators, but this doesn't work either because you end up with under-loaded operators. And bored operators are poor operators."

Because overloaded and under-loaded operators are both ineffective and potentially dangerous, Strobhar explained it's crucial for process user companies to base their console staffing levels on steady-state workloads, and design their applications, process controls and automation with upsets in mind. Not coincidentally, he added, this is one of the main missions and research topics of the Center for Operator Performance and its members, including Chevron, Koch Group, Citgo, Nova Chemicals, Yokogawa, Emerson Process Management, Beville and Wright State University, where the center is located.

"Most of our research has been on display areas and developing tools for evaluating them," added Strobhar. "In fact, we are just about to publish our 'Data to Information, Refining Edition' handbook that provides design templates, and covers types of data, data representation elements, arranging elements, displaying visual information, dynamic graphic symbols, designing displays, common processing equipment, refining in hydrocarbon processing units, and putting it all together."

In general, Strobhar added that designing process applications for upsets requires developers to address the following major areas:

  • Displays should take into account operator situations, awareness graphics, content selection tools, data-to-information handbook, display evaluation toolkits, radar plots and background color;
  • Training should take into account decision-making exercises, shadowboxing, training strategies and mental models to help operators make better decisions;
  • Support systems should take into account event prediction, procedure warnings, procedure modularization, fatigue factors and line labeling;
  • Workspaces should evaluate screen clutter, too-large screens, and unneeded screens; and,
  • Alarm management, which should include alarm rates, presentation, comparisons of expert versus novice operator performance, and alarm tones.

"We've found that during emergency procedures, only 10% of the information presented on displays is unique, and because operator errors increase quickly when operators work more than several consecutive workdays, it's important to focus on improving data-volume procedures for them," added Strobhar. "We're also going to study alarm tones, and hope to study “safe-park” methodologies for stabilizing applications.

“In any event, while operators can be the limiting component in optimizing system performance, it's also true that people have characteristics, just like equipment, and we can design to account for those characteristics."