Professional Documents
Culture Documents
Human
Intervention: What Is the Best Fit
for the Best Performance?
Joel M. Haight, PhD, PE, and Vladislav Kecojevic, PhD
Dept. of Energy and Geo-Environmental Engineering, Penn State University, University Park, PA 16802; jmh44@psu.edu
(for correspondence)
In today’s complex industrial processes, automated provide judgment, logic, experience, and opinions. As
control systems are a necessity. However, is complete a component in the system, we are interactive, variable,
automation the answer? Whereas control system auto- and adaptable. We fill many roles because we can
mation provides predictable, consistent performance, it adapt and specialize, but our natural variability makes
is lacking in human judgment, adaptability, and logic. it possible for us to take actions that a system may not
Although humans provide these, we are unpredictable, be able to tolerate [3]. Therefore, our input is a “bless-
inconsistent, and subject to emotions and motivation. ing and a curse” [3] to those responsible for system
To maximize system performance, should we automate design or performance.
humans out of the system? . . . or . . . Do we maximize Human error is often implicated as a cause of or
human input and lose efficient, consistent, error-free contributor to process industry incidents that result in
system performance? The answer is likely somewhere in injuries, fires, spills, unplanned equipment downtime,
the middle of these two extremes and different for each and the like. Even with increasing automation, various
system and situation. This paper provides a review of sources report, that between 50 and 90% of all indus-
the existing literature covering control schemes and trial incidents are caused by human error. It is difficult
parameters that determine system performance. It at- to say what the real figure is because it depends on
tempts to help answer the questions “How can we min- one’s perspective. However, whatever your perspec-
imize human error while still maximizing system per- tive, it can be reasonably well argued that a large
formance?” and “What is the right human–machine percentage of industrial incidents are contributed to or
mix?” © 2005 American Institute of Chemical Engineers caused by human error [4].
Process Saf Prog 24: 45–51, 2005 As technology improves and efficient, productive
output becomes necessary for financial survival, we are
INTRODUCTION inclined to automate control systems. However, there is
Does automation of control systems in today’s in- a cost to the growth in this phenomenon. Twenty-two
dustry help to reduce human error? Intuitively, one years ago, this author, while learning oil production
might expect that if we engineered humans out of the operations, spent much time in the field with experi-
system, errors would decrease. This seems to be a enced first-line supervisors. It was impressive to watch
worthwhile goal. In fact, well-known management them in action. A supervisor would stop the truck upon
consultant, Walter Bennis, says the factory of the future hearing something out of the ordinary (a hiss, an un-
will have only two employees, a human and a dog. The familiar vibration). He would listen more closely, place
human is there only to feed the dog and the dog is his hand on a pump or piping, and then radio mainte-
there to bite the human if he or she touches any- nance to request the repair of a leak, a bad order
thing [1]. bearing, or some other problem. These supervisors
Human error is inevitable, so we may be tempted to relied on “sentient” knowledge (obtained from the
think like Mr. Bennis [2]. Although it may be appealing senses and experience—the concept of “gut feel” may
to automate humans out of the system, humans do partially describe this) to operate the process. They had
a “feel” for the system. With today’s reliance on com-
© 2005 American Institute of Chemical Engineers puter-controlled automation, it is less likely that oper-
Human Machine
Strengths Weaknesses Strengths Weaknesses
Can apply judgment Inconsistent Consistent No judgment
Adaptable Subject to errors Predictable Cannot be programmed for all
eventualities
Can apply sentient Unpredictable and Efficient No sentient knowledge
knowledge possibly unreliable
Interactive Subject to emotion and Uniform and Constrained by human limitations
motivations reliable in design, installation, and use
Table 3. Human roles in an automated system. pated human responses result when the designer has
not adequately thought through all potential responses.
Acknowledge control system signals (no change to Given this, it appears that an optimized system that
process) maximizes performance and minimizes human errors
Acknowledge control system signals (make required will dynamically operate somewhere between full au-
changes) tomation and complete manual control. Where on this
Record data and instrument readings and adjust as continuum, a system should operate, depends on the
needed application. Whatever the application though, to max-
Monitor system status and override as necessary imize system performance, a designer should maximize
Monitor system status and report findings with no the need for and use of what machines do best: accu-
change rate, consistent, fast, continuous, economic operation.
To ensure minimized human error, a designer must
integrate human input into the system in such a way
that the operator stays mentally and physically en-
automate, and to what level. Unfortunately, increasing gaged. As a “manager” of the interface, the operator
automation tends to promote the likelihood switching should have a monitoring role with override capabili-
to “habits of mind” [9]. To maximize system perfor- ties, receive adequate system status feedback with
mance, research indicates the need to take advantage enough time to respond, and trust system accuracy and
of the human and the control system strengths and then reliability [11]. This is a tall order and different for each
to create effective and active communication between system. Often, the level of interaction between the
them [10]. Table 3 shows the possible roles that the
operators and the control system develops over time.
human operator can fill in a human–machine system.
As personnel, training, and experience change over
HOW SHOULD AN ENGINEER INTEGRATE THE HUMAN AND time, the human input will change. An experienced,
AUTOMATED COMPONENTS? well-trained operator can become bored and less vigi-
Although human error may be caused by, among lant as he or she learns the system. Less vigilance often
other things, the work environment, research shows means more errors of omission in the process. This
that it is nonetheless inevitable [2]. Although minimiz- phenomenon is not fully in the control of the system
ing human error is not the only reason for automating designer.
a system, it is one of a number of the driving forces. An engineer must have an understanding not only of
Before determining how and to what level an engineer/ how the automation hardware and software operates,
designer should automate a system, it is necessary to he or she must also understand the inner workings of
present what “automation” means in the context of this the human operator; physical, mental, motivational,
article. One author defines automation as the execution emotional, training level, experience level, and the like.
by a machine agent (usually a computer) of a function
Often, an engineer will automate a system to a level
that was previously carried out by a human. They
that maximizes economic benefit, but then, almost as
explain that given this definition, what is currently
considered “automation” will change over time [7]. an afterthought, leaves the human operator to manage
A fear that workers have had in the past is that the system as best he or she can [7]. This can lead to
automated systems are meant to replace them. This fear increased errors, upsets, injuries, fires, and spills, for
may have been encouraged by the computer system example. The idea of allocating tasks between humans
“Hal” in the movie 2001, A Space Odyssey. Automation and machines within the context of a complex system
is not intended to remove the human; it merely changes is referred to as function allocation [6]. It is critical to
the nature of the role from one of doer to one of safe operation that designers carefully consider func-
overseer [7]. The operator manages the human–ma- tion allocation. Decisions related to task allocation
chine interface to keep system performance high [8]. A among humans and between humans and machines
problem occurs, however, when the designer does not are design-based decisions [6]. Because there are so
fully consider the ways in which the human role is many variables involved and their interrelationships are
changed by the automation. Unintended or unantici- complex, it is expected that optimization of these vari-