66th Annual Instrumentation Symposium for the Process Industries January 27-29, 2011

Human Machine Interface (HMI) Design: The Good,

The Bad, and The Ugly (and what makes them so)
Paul Gruhn, P.E.
ICS Triplex | Rockwell Automation, Houston, TX
[email protected], 713-353-2417
Abstract
Poor HMI designs have been identified as factors contributing to abnormal situations,
billions of dollars of lost production, accidents, and fatalities. Many HMIs actually impede rather
than assist operators. Many of the poor designs are holdovers due to the limitations of early
control systems and the lack of knowledge of system designers. However, with the advent of
newer and more powerful systems, these limitations no longer apply. Also, decades of research
has identified better implementation methods. Unfortunately, change is difficult and people
continue to follow poor design practices. In fact, some new designs are actually worse than older
designs! Just as a computer is not a typewriter, new HMI designs should not mimic those of old.
The problem is that many designers often simply dont know any better. This presentation will
review why certain HMI designs are poor (with many examples) and show how they can be
improved.
A Comparison
People are naturally proud of whatever they produce, whether its something as simple as a
menu, or as potentially complicated as an HMI. Its their baby and they love it, no matter what
anyone else may think! Unfortunately, many people dont know what they dont know. As a
result, theyre incapable of judging against established criteria; theyve never been educated or
trained on the subject at hand.
For example, when desktop publishing first became available to the masses on personal
computers, the result was a flood of very poorly designed newsletters and other printed materials.
People were producing documents without any knowledge of or training in graphic design.
Figure 1 is a reproduction of an actual restaurant menu. No doubt the restaurant owner thought
he did a wonderful job. Hopefully his cooking skills are better than his graphic design skills!
Good graphic design can be distilled down to four simple and basic concepts:
1. Contrast: Things that are different should look very different.
2. Repetition: Repeat visual elements.
3. Alignment: Every element should have some visual connection with another.
4. Proximity: Things that belong together should be placed together; those that are
different, should not.
These four concepts are very easy to remember using an acronym based on the first letter of
each of the four words. (Sorry!) Utilizing those four simple rules results in the menu shown in
Figure 2. Which do you think is easier to read? Which gives a better image of the establishment?
These concepts can also be used when designing HMIs!

HMI Design

Gruhn

66th Annual Instrumentation Symposium for the Process Industries January 27-29, 2011

Figure 1: Original Menu

Figure 2: Redesigned Menu

(Source: The Non-Designers Design Book)

Principles for User-Centered Design

More data does not equal more information. The information displayed should not center on
the sensors and technology that produce it. A proper user-centered design should integrate the
information in ways that fit the goals, tasks and needs of the user.
User-centered design can be achieved through the following principles and practices.
Organize technology around the users goals, tasks and abilities. Task analyses need
to be conducted to check if the system provides the needed information.
Organize technology around the way users process information and make decisions.
People first act to classify and understand a situation. Experts use pattern matching to
draw upon long-term memory in order to quickly understand a situation. Decision makers
must do more than simply perceive the state of their environment in order to have good
situation awareness. They must understand the integrated meaning in light of their goals;
they must understand the situation as a whole.
Keep the user in control and aware of the state of the system. High levels of
automation can actually put users out-of-the-loop leading to low levels of situation
awareness. If their awareness is compromised, their ability to be an effective decision
maker is also compromised.
HMI Design

Gruhn

66th Annual Instrumentation Symposium for the Process Industries January 27-29, 2011

Situation Awareness (SA)

Situation awareness means a) being aware of what is happening around you, b)
understanding what that information means to you now, and c) understanding what that
information means to you in the future. Situation awareness relates to the goals and objectives of
a specific job or function.
Situation awareness consists of the following three levels.
1. Perception of needed data. The status, attributes and dynamics of elements in the
environment.
2. Comprehension of the current situation. Comprehension is based on a synthesis of the
disjointed Level 1 elements and a comparison of that information against ones goals.
Novices may not have the knowledge base to draw on and may be at a disadvantage when
trying to develop Level 2 SA.
3. Projection of future status. The ability to predict what will happen. This can only be
achieved by having a good understanding of Level 2 SA, along with a highly developed
mental model of the system.
By understanding how operators select and use goals, designers can better understand how
information is perceived. Without understanding the users goals on SA, the information
presented has no meaning.
Demons of Situation Awareness
Situation awareness depends upon human information processing which is a complex
process. Factors that undermine SA are referred to by Endsley et al as demons. They consist of
the following.
1. Attention tunneling. Fixating on one set of information to the exclusion of others. This
is the most common type of SA failure.
2. Working memory trap. Relying on limited short term memory.
3. Workload, anxiety, fatigue, and other stressors. Reduction in a persons capacity to
process information.
4. Data overload. Overwhelming amounts of data can reduce SA. For example, text moves
through our mental pipeline much more slowly than information presented graphically.
5. Misplaced salience. Salience is the compellingness of certain forms of information. The
color red, movement, and flashing lights can distract a person away from information that
may be more important.
6. Complexity creep. Its difficult to develop an accurate mental model of how a system
works when there are too many features.
7. Incorrect mental models. Using the wrong mental model will lead to misinterpretation
of information.
8. Out-of-the-loop syndrome. Automation can actually undermine SA by taking people
out-of-the-loop. They may develop poor SA on how the automation is performing, as
well as the elements the automation is supposed to be controlling.

HMI Design

Gruhn

66th Annual Instrumentation Symposium for the Process Industries January 27-29, 2011

Typical Practice in the Process Industry

Unfortunately, many HMIs are developed with little user input. Many users do not have an
HMI design guideline. This means every HMI they receive on every project will be different.
This lack of standardization will naturally cause problems. Many HMIs are created by white
collar, college educated engineers; yet the systems are used by blue collar, high school educated
operators. The mental models used by engineers and operators differ. Operators are not asked
what their goals are, how theyre expected to operate the plant efficiently, or what information
they need in order to do that. Most HMI screens are based on P&IDs (Piping and Instrumentation
Drawings) and animated with live data (fields with numbers). This is Level 1 SA. This generally
does not provide the operator with the relevant information he needs in order to understand how
the plant is operating (Level 2 SA) or how it may be trending (Level 3 SA). The result is HMIs
like the one shown in Figure 3.

Figure 3: A Poor HMI (Source: The High Performance HMI Handbook)

In many such HMIs fan and pump impellors are spinning, flames are animated, valves that
are open and pumps that are on are shown in one color, valves that are closed and pumps that are
off are shown in another. Different process lines are shown in different colors. The overuse of
animation and color results in misplaced salience. Raw data (not useable Level 2 information) is
presented in many different fields. Too much raw data results in both working memory trap
problems and data overload. Just what should all the values be? How can this data be turned into
HMI Design

Gruhn

66th Annual Instrumentation Symposium for the Process Industries January 27-29, 2011

usable information (e.g., is the plant running optimally)? How can this information be projected
into the near future (e.g., are we approaching any alarm or trip levels)? What level of experience
and training are required for someone to have a correct mental model of the system?
More recent guidelines stress the use of low contrast gray backgrounds and limited use of
color. Some designers still violate the basic principles and screens such as the one shown in
Figure 4 continue to be produced.

Figure 4: A More Recent, But Still Ineffective HMI

Operator error has been cited as a major factor in many accidents. Many of the errors actually
stemmed from poor situation awareness caused by what may be better termed design-induced
error. Its not that people make bad decisions or act poorly; they most often simply
misunderstand the situation theyre in.
For example, a poor HMI was considered one of many contributors to the 2005 accident in
Texas City, TX. The accident involved overfilling a distillation column and a blow down vessel
(when no more than 10 feet of liquid was supposed to be present in the distillation column). How
could the operator have allowed this to happen? The level transmitter indicated a high level
alarm, but the high level switch did not. The switch was not working properly, the transmitter
was not calibrated properly, and the transmitter was incapable of measuring a level beyond 10
feet. One screen showed the amount of material going into the column, another showed the
amount of material going out. No single screen showed the information the operator really
needed the difference between the two the material balance. Had this useful piece of
information been presented along with the range of values it should be, and that it had been
exceeded by orders of magnitude the conditions leading up to the accident may never have
developed. The operator was completely unaware that anything was wrong.
HMI Design

Gruhn

66th Annual Instrumentation Symposium for the Process Industries January 27-29, 2011

A Medical Equivalent
Imagine that youre going to see your doctor for a checkup. You lie down on a bed, a
technician connects various sensors and monitors to you, and the technician then views a display
like the one shown in Figure 5.
Just as your plant process is represented as a P&ID with spinning pumps and numbers
representing raw data, this medical screen shows a graphic of a body, complete with a pumping
heart and lungs that expand and contract. Raw data is displayed. Is this person healthy? What
should the numbers be? How long does it take you to scan and interpret the information? Are the
numbers meaningful to you? How much training and experience would you need to understand
this raw data?

Figure 5: Sample Medical Display #1

Figure 6: Sample Medical Display #2

It might be easier to interpret the information if you knew what range each number should be
within. Figure 6 represents a slight improvement, but how long does it take you to scan the
information and determine if each number is within the proper range, and by how much? Might
there be a better way still of displaying this information?
Text fields are an inefficient way
for humans to interpret information.
Humans can interpret more
information more quickly if its
represented graphically. The actual
numerical values dont even matter if
theyre within normal limits. Figure 7
shows the Star Trek medical display
from the mid 1960s. It takes less
than two seconds to interpret this
display! This is information (Level 2
SA) rather than raw data (Level 1
Figure 7: Sample Medical Display #3
(Source: The High Performance HMI Handbook)
SA) that requires interpretation.

HMI Design

Gruhn

66th Annual Instrumentation Symposium for the Process Industries January 27-29, 2011

A More Effective HMI Design Process

An enormous amount of data can be made available to operators today. However, the amount
of data available can easily exceed an operators ability to effectively process the information.
The following decisions need to be made.
1. Among the vast quantity of data, what is the information that the operator needs?
2. What is the best way to present that information to the operator?
3. How should that information be organized?
4. What information should be emphasized?
An effective interface design process can help answer those questions. The process consists
of the following three steps.
1. requirements analysis
2. design
3. measurement and feedback
The requirements analysis is the most important phase. There must be an operational
concept of how the system will be used. Environmental conditions and user characteristics must
be defined. Operational requirements consider how the users work and accomplish their jobs. A
goal directed task analysis is used to identify high level goals (not the pre-conceived tasks to
accomplish them), decisions, types of cognitive processes users employ, their need for
interaction with others, and their SA requirements. This can be determined through interviews
and observation. The result is a system requirements document.
Part of the design process consists of a function analysis and function allocation. Define
what functions need to be done to accomplish the job, such as monitoring of information,
generation of options, selection of actions, or the implementation of tasks. Function allocation
specifies who will then perform each function, such as the human, the machine, or a combination
of both. Information should be displayed based on the operators goals, not based on the
technology providing the data (readouts of temperature and pressure values). Research has
shown that when automation aids in or takes over task implementation, overall performance
improves. However, when automation is used to generate options, performance declines.
Displays should provide Level 2 SA information that supports comprehension, rather than
Level 1 data that requires cognitive effort. For example, display a level deviation directly rather
than require the operator to mentally calculate the difference between the direct reading and the
desired value. Provide assistance for Level 3 SA projection of future status through the use of
trend displays. The operator can then see where the process is heading. The operator can then be
proactive and recognize impending problems, rather than being reactive and responding to
alarms and problems after the fact. Think of ways to keep the operator active and in-the-loop,
rather than as a passive observer. For example, if you were to ride as a passenger to a new and
distant location, how comfortable would you be navigating your way back if you had to drive
yourself? Passive processing leads to poor situation awareness.
In terms of measurement and feedback, early prototypes allow input from both users and
designers so that the iterative process of design convergence can occur. An evaluation of user
workload and SA can be done, although objective measures can be difficult to obtain. Situation
awareness is an internalized mental construct and creating measures to access and describe it are
difficult. Events can be incorporated in the simulation and the individuals response (or nonresponse) analyzed. Unfortunately, there is not always a direct relationship between good SA and
good performance. Operators and/or observers can be asked to rate the quality of SA, although
HMI Design

Gruhn

66th Annual Instrumentation Symposium for the Process Industries January 27-29, 2011

such subjective measures also have shortcomings. Simulations can be halted at random intervals
and operators can be asked a variety of relevant questions to determine their SA.
One example of an effective screen is shown in Figure 8.

Figure 8: An Effective High Performance HMI (Source: The High Performance HMI Handbook)

Just as a Doctor does not need to see a graphic of the human body in order to understand the
health of a patient (Figures 5-7), an operator does not need to see a graphic of a plant P&ID in
order to understand the health of the process hes trying to control (no matter what he may be
used to or may want). Figure 8 represents Level 2 and 3 SA information. Analog bar indicators
show the current status of the process better than displaying raw data and taxing the operator to
interpret it all. Trend graphs show both the current and projected performance of the process.
The operator can now actively take corrective action before the process approaches or exceeds
normal limits. Displays are low contrast gray with very limited use of color. Colors such as
yellow and red should be used in limited amounts and for alarming only. Whether a pump is on
or off, or whether a valve is open or closed, should not be called out to the operator, unless the
item is in a state that it should not be. Clicking on buttons or other areas of the display can take
the operator to more detailed screens. The number of mouse clicks to get to any particular
information should be limited.

HMI Design

Gruhn

66th Annual Instrumentation Symposium for the Process Industries January 27-29, 2011

Characteristics of Poor HMIs

P&ID representation
Presentation of raw data as numbers (temperatures, pressures, etc.)
No trends
Flashing/spinning graphics
Bright colors, 3-D shadows
Color coding of piping and vessel contents
Measurement units in large, bright text
Lots of crossing lines
Alarm related colors for non-alarm related elements
Inconsistent colors

Characteristics of Effective HMIs

Depiction of process status and values as information, not numbers

Layout consistent with operators model of the process (not a P&ID)
Key Performance Indicators as trends
No gratuitous information
Gray backgrounds, low contrast
Very limited use of color (for alarming)
Consistent visual and color coding
Gray process lines
Measurement units in low contrast lettering, if used at all

Conclusions
Douglas Adams said, Human beings, who are almost unique in having the ability to learn
from the experience of others, are also remarkable for their apparent disinclination to do so.
Change is difficult. The suggestions in this paper will not be accepted by many (no matter what
the body of research may suggest) and implementing the suggestions will not be easy. In order to
do so, there will need to be involvement and buy-in from management, technical staff, and
operators. Perhaps the basic education and information presented in this paper will be a useful
first step. Further details are provided in the references cited.
Acknowledgments
The author does not claim to be the original developer of the bulk of this work and gratefully
acknowledges the authors listed in the reference section.

HMI Design

Gruhn

66th Annual Instrumentation Symposium for the Process Industries January 27-29, 2011

Author Bio
Paul Gruhn is the Training Manager at ICS Triplex (a Rockwell Company) in
Houston, Texas. Paul is an ISA Fellow, a member of the ISA 84 standard
committee, the developer and instructor of ISA courses on safety systems, and
co-author of the ISA textbook on the subject. He has a B.S. degree in
Mechanical Engineering from Illinois Institute of Technology, is a licensed
Professional Engineer (PE) in Texas, a Certified Functional Safety Expert
(CFSE), and an ISA 84 Expert.
References
1. Designing for Situation Awareness: An Approach to User-Centered Design by Mica
Endsley, Betty Bolt, and Debra Jones, CRC Press, 2003
2.

The High Performance HMI Handbook: A Comprehensive Guide to Designing,

Implementing and Maintaining Effective HMIs for Industrial Plant Operations, by Bill
Hollifield, Dana Oliver, Ian Nimmo, and Eddie Habibi, PAS 2008

3.

Human-Machine Interface Design for Process Control Applications, Jean-Yves Fiset, ISA,
2009

4. Investigation Report, Refinery Explosion and Fire, Report No. 2005-04-I-TX, March 2007,
Chemical Safety Board
5. The Non-Designers Design Book, Robin Williams, Peachpit Press, 2008

Further Information
1. PAS web site: www.pas.com
2. User Centered Design Services web site: www.mycontrolroom.com
3. Abnormal Situation Management Consortium web site: www.asmconsortium.net

HMI Design

10

Gruhn