Copyright (c) 2005 IFAC.

All rights reserved

16th Triennial World Congress, Prague, Czech Republic

FROM ELECTRIC MOTORS TO FLEXIBLE MANUFACTURING:

CONTROL TECHNOLOGY DRIVES INDUSTRIAL AUTOMATION

Sujeet Chand
Vice President and Chief Technical Officer
Rockwell Automation

Abstract: Industrial Automation has evolved from stand-alone, hard-wired relay panels to
a contemporary, networked system that supports flexible manufacturing and enterprise
integration. From precision control of machines and robots on the factory floor to flexible
coordination of multiple cells of automation, advancements in control technologies have
driven the evolution of industrial automation. The confluence of five technologies,
control, computing, communications, software, and materials, is shaping the future
direction of industrial automation systems. This paper summarizes the major technical
trends, and highlights the continuing opportunities and challenges for the application of
control technologies in industrial automation. Copyright © 2005 IFAC

Keywords: factory automation, industry automation, control engineering

The advent of the Internet has greatly accelerated the

1. INTRODUCTION: EVOLUTION OF availability of information from factory control at the
CONTROL IN INDUSTRIAL AUTOMATION right time, at the right place and in the hands of the
right people to enable “e-manufacturing.” The need
One of the earliest known control schemes could be for access to more and more data has driven the
found in water clocks of the Greeks and Arabs. migration of networks like Ethernet down into the
Developments in control did not progress much until factory floor. In fact, the rapid adoption of Ethernet
the dawn of the industrial revolution with control along with the growth in information and
systems for temperature, level, pressure, and the communication technology, has enabled the “shop
centrifugal governor made popular by Watt’s steam floor” or the factory to be integrated with the “top
engine. With the developments in general purpose floor”, or business enterprise. Several enterprises
technology such as electricity, relays hardwired to serving a common business purpose are now tightly
predetermined logic became a popular means of coupled to form a “supply chain.” The focus has now
control in factories. Following the rapid advances in expanded to achieve new efficiencies in the entire
electronics and the advent of integrated circuits, value chain through collaboration by integrating the
Programmable Logic Controllers (PLCs) provided an supply chain through the use of standards such as web
alternative to hardwired relays in industrial services, and by facilitating information flow
applications, and gave birth to what is widely throughout the enterprise through use of open
accepted today as factory automation. The PLCs were architectures and associated protocols.
programmed with relay ladder logic, which provided a
flexible mechanism to make changes to the operation Collectively, these technologies are helping
of machines in the factory without rewiring relays. manufacturers meet today's most challenging
The advent of digital communication networks on the manufacturing demands such as improving production
factory floor allowed distributed control, i.e., control uptime, meeting regulatory compliance, optimizing
to be divided into cells that are geographically manufacturing, and maintaining an aggressive time-
distributed, with significant reduction in the size of to-market goal. And, control technologies ranging
individual programs. In multi-cell systems, operation from automatic control to autonomous control, are at
of multiple distributed controllers was coordinated the core of industrial automation. The breadth of
with interlocks of data or I/O points in hierarchical applications of control technologies spans all three
structures of master controllers. dimensions of today’s collaborative enterprise (Fig. 1):
(1) Value Chain (suppliers - customers), (2)

1
Enterprise (shop floor - business) and (3) Product Life example, High Temperature Superconductivity (HTS)
Cycle (design - support). materials will drive new motors that are smaller,
lighter, and much more efficient than today’s
conventional high-efficiency induction motors.

Also, improvements in chip-level integration of

sensors, processing logic and associated architecture
are enabling the evolution of smart, multi-function
devices. One example is an integrated sensor for
monitoring the fluid health of a bearing lubricant. We
Control prototyped such an “intelligent” sensor utilizing
MEMS technology, which is able to detect moisture
ingress, particulate contamination, changes in
viscosity, oxidation levels and changes in acidity.
Integrated sensing, processing and communication
will allow a plethora of such sensors to communicate
with each other, and with controllers/processors for
real-time control and diagnostics. With integrated
Shop Floor sensors collecting immense amounts of raw data,
users need a system that seamlessly converts this raw
data into relevant information and presents that
Figure 1. Collaborative Manufacturing
information to controllers and the enterprise level to
prompt the most appropriate and timely response.
The next section provides a brief overview of the key
trends driving industrial automation today, and the
Figure 2 illustrates a layered view of an industrial
resulting challenges for the application of control
automation system. The real-time control layer
technologies.
comprising the factory control system such as
machine and robotic control, provides many
2. TECHNOLOGY TRENDS IN
challenging applications for automatic control
MANUFACTURING SYSTEMS
including modeling, identification, adaptive control,
discrete-event systems theory, and control system
The evolution of modern manufacturing systems is
design utilizing both linear and nonlinear control
driven by trends in five major technology clusters:
methods. The higher level, decision-making
Control, Communications, Computing/Electronics,
transactional layer is responsible for interfacing with
Software/Information, and Materials. The rapid
the business system for information such as a work
advancements in computing, communications and
order, and transforming the work order into a detailed
software technologies are driving a hitherto
schedule for the factory. Control technology
disconnected manufacturing enterprise to rapidly
applications at this layer include decision theory,
become connected with global business systems.
cognition and control, and autonomous systems.
Today’s mostly single-purpose manufacturing
systems that are designed to produce the same part or JD Edwards
Decision-Making; Transactional Control

component repetitively are migrating to adaptive and

Supply Chain Management
flexible manufacturing enterprises. This migration is
driven by the global trend towards mass- Enterprise Management
customization. Consider the food industry for Materials Mngmt.
Mngmt. Sales & Distribution Production Planning Quality Management

Plant Maintenance Human Resources Workflow Industry Solutions

example, where manufacturers would like to create
customized packages - one packaging for Wal-Mart, Manufacturing Management
another for Tesco - to give their products a different
Scheduling Genealogy WIP Quality UP/Down Time Efficiency
identity on retailers’ shelves. Another example is the
pharmaceutical industry where the trend is towards
manufacturing customized medications and dosage by Data & Visualization
HMI SCADA Trending Alarming SPC SoftLogix
requirements of each individual. This trend requires Appliances Open Platforms
Real-Time Control

flexible control of the manufacturing enterprise.

Machine Control Process Control
The current focus on energy efficiency and greater
Discrete Batch Process
productivity is driving the utilization of energy-
efficient electric motors and controllers to replace I/O Devices
eddy current, hydraulic, and pneumatic equipment.
Since electric motors account for about 60 percent of Figure 2. Layered Industrial Automation Architecture
the power demand in the U.S., a technology-driven
cost reduction opportunity lies in the development of In the following two sections, we will highlight the
smaller, more efficient motors and controllers. For applications of automatic control and autonomous
systems in industrial automation.

2
 integral control eliminates steady state error. These
3. AUTOMATIC CONTROL: A KEY gains may have to be de-tuned in order to obtain a
ENABLING TECHNOLOGY FOR INDUSTRIAL
AUTOMATION

In this section, we will focus on the core technology

that enables real-time control on the factory floor -
Automatic Control. In theory, a workable control
solution can be found for practically any application
in industrial automation. Researchers have
successfully modeled, simulated and demonstrated the
efficacy of numerous control methods for Figure 3. Acceptable PID performance after de-tuning
mechatronics, robotics, machine control and process of gains
control applications. However, the actual
implementations of automatic control in stable response; however, a given system in most
manufacturing plants reveal a preponderance of the cases, can be tuned to operate well under numerical
classical PID control. Why hasn’t industry embraced control. For example, the plot on the left hand side of
the numerous advances in modern control and Fig. 3 illustrates the response of a physical system
implemented them in its processes? There are two where the bandwidth was too high for unknown (un-
key factors: (1) ROI (better, cheaper, and faster), and modeled) system resonances. The plot on the right
(2) ease of application. To be utilized broadly, a new hand side shows how the system can be stabilized by
technology must demonstrate tangible benefits, should reducing the bandwidth, yielding a compromised but
be easier to implement and maintain, add more stable response. This is a well understood technology
functionality, or substantially improve performance and readily accepted by the industry. Operators
and efficiency. Many times, a promising approach is trained in PID control can develop the intuition and
not pursued due to poor usability during operation and understanding to manually tune the gains for control
troubleshooting in an industrial environment. The PID loops. As manufacturing systems become more
due to its simplicity, scores well on this front. complex with a growing number of multivariable
systems, we believe there is significant potential for
As new, advanced manufacturing processes continue the application of alternative control technologies
to evolve and the existing processes are pushed to over the classical PID technique.
their limits, the need to evolve the control
architectures past the ubiquitous PID control into a 3.2 Intelligent Control: Applications in Industrial
new control arena is becoming a necessity. We Automation
believe that advances in information and
communications technologies resulting in greater Let us briefly consider the application of “intelligent”
availability of connected computing resources, will control schemes to factory automation. Neural
provide the impetus to the greater adoption of Networks, Fuzzy Logic, Genetic Algorithms, and
automatic control methods beyond the classical PID. Expert Systems, each have their benefits and
challenges for implementation. For example, certain
3.1 Why PID is Still the Mainstay In Industrial types of Neural Networks require training data to
Applications calculate the weights. In many cases, it is impractical
to operate an industrial process over its entire
An industrial manufacturing plant can be divided into operating range to collect this data. Fuzzy Logic
four types of applications: extraction, processing, systems require an expert who knows the process to
finishing, and assembly, with manufacturing be able to transfer this knowledge into the rules and
processes broadly in three categories: continuous, input/output data sets, which is not a straightforward
batch, and discrete. Raw materials enter the plant and task. We note that fuzzy logic has become a popular
finished goods exit. Control of the systems in- control scheme for cement manufacturing because of
between range from the well known Single-Input its superior performance over other techniques. This
Single-Output (SISO) PID controllers to the more is a direct result of the commitment of significant
challenging Multi-Input Multi-Output (MIMO) effort between the cement plant experts and fuzzy
modern control techniques. logic control designers. However, a general purpose
fuzzy logic engine to control complex multivariable
PID controllers continue to be the work-horse of the systems does not exist today.
industry today. Although no real system is truly
linear, many systems can be approximated by a linear Although systems based on intelligent control can be
system. Also, a system may be segmented into made to work (as in the cement application), the
piecewise linear subsystems. PID controllers question still remains on whether the new system is
adequately handle 3 main problems: proportional gain significantly better than an existing solution, is easier
drives the controller based on the system error, to install, test/verify, and maintain, and is it worth the
derivative control dampens out disturbances, and change.

3
3.3 Modern Control Technologies For Industrial performance for a given roll is also a challenge.
Automation Synchronizing the system to obtain perfect
registration of four colors on a sheet of paper is a
Recent advances in estimation and modeling, coupled difficult control problem. These motors are connected
with ubiquitous computing and networked systems, through a gearbox to long shafts with flexible
are paving the way for future applications of modern, couplers. The shafts are connected to large inertial
automatic control techniques in industrial automation. rolls that have the paper wrapped around them. The
Demanding applications with increasing requirements gearbox, shafts, and couplers introduce lost motion
for precise and stable control can benefit from the and resonance in the system to make control a
application of model-based control methods such as: difficult objective. It is not uncommon to have inertia
ratios between the load and motor of up to 50:1.
• LQR theory, which calculates the optimal Current approaches include inertia adaptation or
gain matrix K such that the state-feedback acceleration feedback to increase the electronic inertia
law u = -Kx minimizes some cost function J of the system, thus effectively making the system
= Integral{x'Qx + u'Ru + 2*x'Nu}dt subject stiffer to disturbances at high frequencies. This is
to the state dynamics xdot = Ax + Bu. similar to adding velocity feedback to dampen
oscillations at the mid frequencies. There is
• LQG theory, which is similar to LQR with significant potential for a model-based control
the main difference being that noise is technique to improve the control performance of this
introduced into the calculations. application.

• H-infinity controllers, which calculate the Although Electronic line shaft (ELS) systems have
controller F(s) and the parameterization K(s) been around for awhile, there exist significant
using the loop-shifting formulae. Given any opportunities for improvements. An ELS system is
stable infinite norm U(s) less than or equal to commonly used in many continuous web applications
one, F(s) is formed by wrapping feedback such as paper converting and printing. Through the
U(s) around K(s). application of modern control methods, the
opportunity exists to dramatically improve the
• H-2 control, which has the same operation as performance of web line systems in the following
H-infinity but with a 2-norm instead of the areas: dynamic synchronization of motors, electronic
infinity-norm. “camming,” and tension control. Closing the
acceleration feedback loop greatly aids the
• The mu-synthesis technique, where the performance of an ELS system. However, additional
objective is to find a stabilizing controller F analysis of the acceleration loop and its interaction
and diagonal scaling matrix D such that the with the current loop is required. This may be
infinite norm constraint is less than one. i.e. extended to the analysis of the jerk and other control
mu-synthesis uses H-infinity to automate the loops as well.
finding of D-F through multiple iterations.
The above is just a narrow sample of the large number
• Linear Matrix Inequalities (LMI), a design of challenging opportunities for modern, automatic
tool to help perform control design and control methods in industrial automation. We expect
system identification. A LMI is any to see greater application of such methods in the
constraint of the form A(x) = A0 + x1A1 + future, which will supplant the current dominance of
x2A2 + … +xnAn < 0. LMIs can be applied the PID.
to all of the previous techniques.
4. AUTONOMOUS CONTROL
• General model-based control methodology
called Model-Predictive control (MPC), Most factory automation systems feature tightly-
coupled automation components, with fixed pre-
wherein dynamic optimization problem is
solved on-line at every instance of control configured automation. While this works well in
execution, is another potential technology many instances, it is vulnerable to single-point of
that has applications in industrial automation. failure, is difficult to expand as the plant’s needs
change, and makes behavior prediction or dynamic
There are numerous examples of challenging control response to unanticipated events in the system
applications in Industrial Automation for the difficult to achieve.
application of the above methods. We will briefly
outline two applications: position registration in Today’s complex, multi-cell systems challenge
printing presses, and electronic line shaft (ELS). control engineers for coordinating the operations of
many small controllers. The proper operation of these
A major control challenge in modern printing presses linked controllers depends on a fixed sequence of
is the synchronization of one motor/drive system to preplanned operations. The current trend toward
the next on a web line. The individual motor/drive consumer-directed, customized manufacturing will
drive a higher frequency of changes and eventually, a

4
highly flexible architecture for manufacturing systems. of failure due to battle damage or other means, (2)
In a flexible manufacturing system, the underlying automatic system reconfiguration for achievement of
control system needs to tackle the numerous the cooling objective, and (3) elimination of a single
changeovers in the mainline configuration, tool point of failure in the system. A highly distributed
allocation, material distribution, process steps, and intelligent shipboard automation system with a large
product quality. In this section, we will focus on the number of autonomous agents was designed for the
higher-level control and decision making in an control of the chilled water system (Fig. 4). Each
industrial control system. agent is programmed with operation-specific
intelligence. The intelligent agents carry out flexible
Autonomous Control systems that can distribute negotiation utilizing a capability-based language.
intelligence to eliminate any single point of failure
and make future expansion much easier are currently In the chilled water system there are 3 subsystems:
being investigated and prototyped by a large number plants, mains and services. There is one plant per zone
of researchers. In such a system, software agents are (2 zones), and there are 14 vital and 2 non-vital
typically used to negotiate and dynamically allocate services. The water-cooling plants are modeled as
tasks to the system’s resources. These autonomous single agents. The main circulation piping is
software agents hold the promise to reduce system partitioned among ‘T’ pipe sections, which are also
downtime by reconfiguring the system in real-time in
the event of a failure. Such actions improve the
productivity of the manufacturing plant and enable the
life of the capital equipment to be extended by
dynamically planning repairs, enhancements and
upgrades in accordance with the production demand.
LPAC
LPAC V003

4.1 Autonomous Agents

V002
V002 DHYR 22 V003
DHYR
V103
V103
V121
V121
IC/
IC/ CHW
CHW
CIC
CIC
GYRO 2
GYRO 2 PLANT 11
PLANT
EQPT 22
EQPT

HVAC
HVAC
CLS 2
CLS 2
SONAR
SONAR V127
V127
EQPT
EQPT STBD
STBD

Autonomous agents are software modules, usually

V601
V601 ARRAY
ARRAY
V122
V122
SLQ-32
SLQ-32
400 HZ
400 HZ
CNVTR 1
CNVTR 1

associated with a physical entity such as a pump,

V603
V603
PORT
PORT
ARRAY
ARRAY
C&D ELEC
C&D ELEC V231
V231
EQPT
EQPT V004
V004
CIC
CIC

valve, motor, pipe, or sensor that has local goal-

EQPT 22
EQPT
V001
V001
HVAC
HVAC
CLS 1
CLS 1 CHW
CHW
IC/ LPAC 400 HZ
HZ PLANT 22
PLANT

directed behavior and control (autonomy). In addition,

IC/ LPAC 400
GYRO 11
GYRO DHYR 11
DHYR CNVTR 22
CNVTR
V230
V230

V234

an agent possesses cooperative, collaborative

V234
V204
V204

capabilities to work in conjunction with other agents V006

V006 V005
V005

to satisfy a more critical, overarching goal. Clusters Figure 4. Agents for Chilled Water System
of agents dynamically form and negotiate amongst
themselves on how each will control a part of the ag ents. There are standalone valves in the main
overall system to achieve new and important circulation loop for the supply and return lines
capabilities. This mode of operation is highly re presented by valve agents. The total system has 68
distributed - there is no centralized control - and no agents. Each agent is associated with capabilities and
central point of failure. The analogy for such control each capability is associated with a specific set of
methods comes from biological systems where beings operations. The negotiation among agents uses local
cooperate to capture food more efficiently, carry planning and negotiated planning (i.e., cooperation),
heavy objects, or form structures. An international and the agents use their local ‘world observations’ to
standards organization, Foundation for Intelligent determine their actions which are translated into
Physical Agents (FIPA), is focused on establishing a execution steps. The agents are loosely coupled, and
common framework and language for disparate agents when dynamic interactions occur during the decision-
to communicate and collaborate in problem solving making process, logical relationships are temporarily
and control. There are several development and established. The agents make internal calls to the
simulation tools available today to support efficient operations offered by their local capabilities, and in
system development of such systems. negotiated planning, agents discover each other’s
capabilities.
To illustrate the application of autonomous agents, we
now provide a summary of an application to The architecture is organized according to the
shipboard automation. following characteristics:
1. Autonomy: Each agent makes its own
4.2 Control of Chilled Water Distribution decisions and is responsible for carrying out
its decisions towa rd successful completion.
On a certain class of Navy ships, a chilled water 2. Cooperation: Agents combine their
system is utilized to cool missile launchers following capabilities into collaboration groups
the launch of a missile. Chilled water is routed (clusters) to adapt and respond to divers e
through a network of pipes and valves that have events and goals.
significant redundancy. The Navy identified this 3. Communication: Agents share a common
system as requiring a distributed, fault-tolerant control language to enable inter-operation.
system to allow, (1) continued operation in the event

5
 4. Fault tolerance: Agents possess the capability
to detect equipment failures and to isolate Marik, V., P. Vrba, M. Fletcher (2004): Agent-Based
failures. Simulation: MAST Case Study, BASYS 04, Vienna,
Austria.
The shipboard autonomous system prototype was
successfully tested in the laboratory, and also Maturana, F., R. Staron, and K. Hall (2004):Real
validated on an actual chilled water distribution Time Collaborative Intelligent Agent Solutions, IEEE
system. The prototype demonstrated the International Conference On Systems, Man and
accomplishment of the three primary objectives Cybernetics, The Hague, The Netherlands.
relating to fault-tolerance, continuous performance,
and elimination of a single point of failure. A brief Staron, R, F.P. Maturana, P. Tichy, and P. Slechta
summary of the lessons learned from this work are, (1) (2004): Use of an Agent Type Library for the Design
simulation in-the-loop was required throughout the and Implementation of Highly Flexible Control
agent implementation cycle to properly design and Systems. 8th World Multiconference on Systemics,
architect the system, (2) although the system is highly Cybernetics, and Informatics, SCI2004, Orlando, FL.
distributed, some degree of centralization is required
to ensure system stability, sub-optimal performance, Vasko, D., F. P. Maturana, A. Bowles, and S.
and reduce the number of agents and complexity, (3) Vandenberg (2000): Autonomous Cooperative
implementation can be on COTS controllers and no Systems for Factory Control. PRIMA 2000, Australia.
special hardware is needed, and (4) interoperability
standards and communication interfaces can be based
on standard protocols such as FIPA, Ethernet/IP, and
CIP™.

As stated earlier, the primary benefit of this

architecture is flexibility. Intelligent agents are easy to
scale, and hold great promise for reducing system
downtime. Such systems will be useful in an
automation system for flexibility, and continuous
operation in the event of failures.

5. CONCLUSION

Since the industrial revolution, control technology has

driven the evolution of the modern manufacturing
enterprise. Systems theory and control design are
ubiquitous in industrial automation systems, from the
control of individual electric motors to the control of
complex processes. Continued evolution of industrial
automation offers a wide range of opportunities for
the application of most modern control methods, as
well as decision theory and autonomous systems.

ACKNOWLEDGEMENT

I would like to acknowledge the significant support

from Dr. Ram Pai and Dr. Peter Schmidt of Rockwell
Automation towards the development of the content
of this paper.

REFERENCES

Discenzo, F., F. Maturana, R. Staron, P. Tichy, P.

Slechta, V. Marik (2004): Prognostics and Control
Integration with Dynamic Reconfigurable Agents,
IASME Transactions, Issue 2, Volume 1, April 2004.

Marik, V, M. Fletcher, and M. Pechoucek, (2001)

Holons and Agents: Recent Developments and Mutual
Impacts. International Workshop on Industrial
Applications of Holonic and Multi-Agent Systems,
HoloMAS.