Copyright © 2002 IFAC

15th Triennial World Congress, Barcelona, Spain

LOW COST INTELLIGENT AUTOMATION IN MANUFACTURING

Heinz-H. Erbe

Technische Universität Berlin

Franklinstrasse 28/29
D-10587 Berlin, Germany

Abstract: To be competitive SMEs with small batch production or tool- and mould
making are forced to use numerical controlled manufacturing processes and
computer based shop floor control. But the technological support for these
enterprises should be shop floor oriented to foster and develop human skills in flat
hierarchies avoiding unnecessary dividing of work. With shop floor oriented
production support human skills and automation create synergetic effects. The
mastery of the manufacturing process is in the hands and brains of skilled workers.
Automation gives the necessary support to execute tasks and rationalize decisions.
This represents a way of low cost automation. To run the manufacturing process
effectively is not only a question of technology though an essential one. Together
with an adequate work-organization wherein human skills can be empowering
themselves it establishes the frame for a competitive manufacturing in SMEs. This
contribution deals with shop floor oriented technologies for manufacturing.
Copyright © 2002 IFAC

Keywords: machine tool controls, shop-floor oriented technologies, maintenance

support

1. INTRODUCTION problems for assembly stations and material flows

in production. It appears that short innovation
A recent study by G. Lay (2001) reveals that cycles define the limits of economical automation.
booming revenues for automation technology Within the last years so called shop floor oriented
vendors hide the fact that belief in automation in technologies got developed and achieved success at
the German investment goods industry is giving least but not only in small and medium sized
way to sober realism. More than a third of 1000 enterprises (SMEs). They are focused to agile
reviewed enterprises had already reduced their manufacturing, that means to use an intelligent
degree of production automation or plan to do so. automation combined with human skills and
The most important reason: The insufficient experiences at the workshop.
flexibility of highly automated systems. Some criteria of these technologies are to mention:
The combination of losses resulting from 1) with respect to the man-machine interfaces:
conversion, idle time and high technical - flexible machine tools with easy access to the
maintenance costs quickly negated the expected working area
economic benefits. Today at many locations highly -manual and numerical controllable machine
automated production facilities are making way for tools
systems with significantly lower degrees of - graphical-interactive programmable NC-
automation. Exaggerated automation was found just controls
as often at small enterprises as it was at larger - multimedia supported interpretation of the
companies. machine diagnosis respecting maintenance
Highly automated material flow systems in through operators,
assembly and highly automated processing 2) with respect to machine-machine interfaces:
machines are partly seen as poor investments. Two - PC´s at the shop floor with NC-programming
thirds of unsatisfied companies indicated that software for different controls but with
today's shrinking series sizes can no longer be identical structures and user surfaces as these
handled economically with these systems. machine tool controls and with program
Inadequate flexibility in capacities follows as the transfer between machine controls and PC in
second most popular reason for dissatisfaction. both directions,
Companies with innovative product ranges face 3) with respect to shop floor planning
special difficulties. They have considerable
 - electronic planning boards or shop floor machines of the first generation sometimes only
control systems usable by skilled workers for need a new control to put them to today's standards.
making decisions with respect to scheduling That is called "upgrading". Sometimes it is desired
tasks, shorten the through-put time and to save the conventional handling of machines
solving problems caused by still stand of despite its retrofitting, i. e. moving the tables and
machines, missing tools or jigs and fixtures, saddles mechanical with hand wheels in addition to
material or personnel, the numerical controlled servo drives. This
4) with respect to product and process quality facilitates the manufacturing of simple parts before
- devices for in-process measuring using the advantages of a CNC-control to
- for all skilled workers usable statistical process manufacture geometrical complex work pieces.
control and statistical experiments. Such controls of machine tools enabling the in-
Group- or teamwork supporting technology is an process switch from manual processing to a support
additional criteria. This means the technology of a numerical control is an intelligent enhancement
should be usable in a sporadic manner by all of the productivity of the shop floor. They are
skilled workers depending on their tasks just now. suitable for manufacturing small lots or single parts,
easy to handle manual and programmable with
2. NUMERICAL CONTROLS FOR MACHINE interactive graphic support at the human-machine
TOOLS interface. These controls are called "job-shop" -
Investing in a new or at least better technology is controls.
connected to decisions for machines with enhanced
productivity of the manufacturing process and also Productivity
to producing a better quality of the products. There
are to points to consider:
investing in new machines or
ls
retrofitting of existing machines. ntro C
Ma

hop co d CN
nu a

For machine tools or manufacturing cells or s r

Job da
an
lM

systems it is not always necessary to replace them St

ach
inin
g

1 10 Lot Size 100

Fig. 2. Benefits in productivity with job-shop
controls

The knowledge and experience of skilled workers

can be challenged and a dividing of work in
programming and operating to different personnel
is unnecessary, saving costs while avoiding
organizational effort and is more flexible. Figure 1
shows machine tools with job-shop controls and
Figure 2 its productivity respective the lot size. Job-
shop controls of Siemens, Hurco et al
Fig. 1. Machine tools with job-shop controls
(www.ad.siemens.de/jobshop;www.autocontech.co
m), are using PC-operation systems like Windows
completely. In many enterprises one can find
NT and can therefore be integrated in an enterprise
conventional machines in a very good state but
network and connected to the internet. These allows
meanwhile not suitable to produce parts of high
for a flexible manufacturing, because not only an
quality in an adequate time. Conventional machine
easy programming at the machines but also
tools like lathes or conventional mills have usually
archiving of programs and loading of programs
a machine bed of a good quality and stiffness. They
from other places is possible. Moreover the
should not be thrown on the scrap-heap. These
connection to tool data management systems allows
machines could be equipped at least with linear
for a quick search and ordering the right tool at the
rules and digital read-outs (DRO) to improve the
workplace. These open factory systems are meant
manufactured quality with respect to required
to expand the functionality of traditional numerical
tolerances. The next improvement could be the
control (CNC) technology by leveraging the
refitting with a numerical control (NC-program
interoperability and familiarity of a PC. Incremental
interpreter, PLC, servo-drives as actuators for each
functionality such as ease of operation, custom
controllable axis, a measurement system as
interfaces, networking, and diagnostics are all tasks
sensors), while the drive of the main spindle can be
ideally suited for machines combining CNC and PC
controlled using a frequency converter.
technologies. Careful selection of a PC-enhanced
Retrofitting packages are available on the market
CNC can simplify use, expand functionality, reduce
(www.giddings.com). Numerical controlled
maintenance, and raise productivity - and can do all Virtual tools are available using tool management
this while lowering the total cost of ownership. systems provided by tool makers via the internet.
The introduction of the PC to the shop floor has the An accurate simulation of jigs and fixtures is also
potential and do already change the manufacturing. possible using data of fixture suppliers also via the
The CNC and PC should be integrated, not just internet. The workers can document the first
connected. This mean that the CNC and PC are
installed in the same unit, but integration is possible
in remotely connected units as well because what's
critical is the degree of logical integration. The
CNC and PC must be able to exchange extensive
amounts of data at rates consistent with the inter-
workings of the CNC, and this data must also
permit real-time adjustments to the machining
process. This tight integration is seen often in
adaptive control applications, diagnostics, and
custom user interfaces. Improved operator
interfaces, central tracking and analysis of
statistical process control (SPC) data, and access to
the machine from the internet are all common
applications appearing on open systems CNCs.
Figure 3 shows a machine shop solution of GE
Fanuc Industrial Systems (www.gefanuc.com). Fig. 4. Virtual machining (CGTech/Vericut)
As it was concluded within the study of G. Lay
(2001) mentioned in the introduction high assembly of fixtures in a machining center in a data
automation of production processes could be base at a PC or the machine control, using a digital
inflexible regarding the quickly changing demands camera. This allows for saving the knowledge and
of the customers. Future factories have to be experience for future use and shortens the setup
dynamically reconfigurable to quickly support time when manufacturing new parts.
customer requirements. To achieve this human A very important tool to reduce the cost of
skills and ingenuity at the shop floor is unavoidable manufacturing is the Design of Experiments
and have to be assisted. (DOE). This is connected with the names of
An important tool is "virtual machining". An Taguchi and Shainin (Bhote, 1991). Automatic
Statistical Process Control (SPC) used in-process
only identifies deviations from the demanded
quality of the manufactured parts. But which
machine parameters have to be changed is
sometimes a fiddly question. DOE offers a
systematic search to find the most important
parameters and their interrelation. The experiments
can be executed through the skilled workers at the
shop floor, thereby shortening the time to an again
stable manufacturing process. Figure 5 shows a
graphic representation identifying the most
important machine parameter.
Validación de Calidad CbRm
Bueno1 BmRb
560
Bueno2
AmRb DmRb
BbRm
520
CmRb
DbRm
480
Fig. 3. Machine shop of GEFanuc R=Resto
Malo1
b=bueno
Malo2 m=malo
accurate simulation of a manufacturing process in a 440 CbRm=Red X
AbRm BbRm=Pink X
machine tool or machining center by high-end
Prueba de A B C D
visualization got developed (www.CGTech.com) Parámetros de Proceso
Significancia
running on PCs (Figure 4). Skilled workers can use
Fig. 5. Variable/Parameter search of DOE
this tool to plan the complex manufacturing for
optimizing in short time. Furthermore a change of
Production to market demands requires the high
the production process due to customer demands
availability of machines and this is last but not least
can be quickly simulated to find the optimal way.
a problem of maintenance.
 3. MAINTENANCE due to the damages of equipment done by the
maintenance action itself.
According to Morel (2001) maintenance should An implementing of condition-based and predictive
keep a systems facilities functioning in order to maintenance substitutes scheduled maintenance to
contribute to the enterprise target. Maintenance decrease expenses and to improve the global
should fulfil the right CRAMP parameters (Cost, performances of the automation system. Studying
Reliability, Availability, Maintainability, and the systems "degradation" to anticipate its failures
Productivity) for any automation system. offers an image of the future situation of the
Maintenance strategies and operations should be system.
considered as a complete process aiming to Therefore the implementation of these processes
maintain the production resources while inter- within the framework of a predictive versus pro-
working with other shop-floor and business active maintenance strategy involves some a priori
processes (automation, planning, quality, knowledge on the system and on its components
management, financial, etc) to carry out the global with a use of this knowledge to represent the views
enterprise goal. Engineering of such a maintenance on the future directions of this system.
system needs a holistic approach for integrating
views and evaluations, not only of the systems Control System ∧ Process System ⊃ System Goals
themselves, but also for their mutual interactions PERFORMANCE
CONTROL PROCESS
and their interactions with the environment. r(k) u(k)
Two approaches for the maintenance of an u(k)=γ (r,x,k) SYSTEM ? J [x(0, …, x(K), u(0), …, u(K)

automation system can be considered: y(k)

?
an a posteriori approach based on an
exploitation of the system data coming from Fig. 6. A closed-loop system with system
experience and the knowledge practices of the real performance optimization rather than control
operation. It allows for decision making of performance optimization (Morel, 2001)
maintenance strategies as a feedback in the system
design phase or as an optimization on line. Again Morel et al (2001) stress the consideration of the
according to Morel et al one of methods most used performance of the complete system, that is
to improve the maintenance plans for automation interesting the owner respecting the cost, rather
systems is RCM (Reliability Centered than the control performance only. I.e. a
Maintenance). RCM is a procedure for determining compromise between cost of maintenance and cost
maintenance strategies based on reliability of stillstand of the automation system (Figure 6).
techniques. The methodology takes into account the Due to increased demands on performance of
prime objectives of a maintenance program: manufacturing systems (quality, reliability and
- minimize Costs. safety) spending for maintenance is likely to grow.
- meet Safety and Environmental Goals. Therefore an agility for the maintenance of the next
- meet Operational Goals. generation of information- intensive manufacturing
The effects of redundancy, cost of spare parts, costs systems requires to broaden the mindset of the well-
for maintenance personnel, equipment aging and established system-oriented approaches in favor of
repair times must be taken into account along with more intelligent technologies and way of thinking.
many other parameters. The RCM process may be After this more theoretical considerations some
used to develop a living strategy with the operational solutions of maintenance are discussed
automation system model being updated when new in what follows.
data is available or design changes take place. The concept of Total Productive Maintenance
A good source of failure information about an (TPM) integrates maintenance into production and
automation system is the experience of operators on therefore enlarge the tasks of the productive
site. Translating this information into future workers at the shop floor. To master these
probability of failures can be performed with the additional tasks the operators of manufacturing
Bayesian reliability analysis (Morel, 2001). systems need a deep understanding of the structure
an a priori approach based on the and support for their decision making respecting
knowledge to controlling system variables active and proactive maintenance. Necessary for
regularly, for (a) optimizing the intervals between these tasks are well prepared interpretations of the
repairing, (b) reducing the number and the cost of diagnosis information. With this support and their
unexpected stillstands, (c) anticipating the failings collected experiences when operating the
of the system. Some studies on the efficiency of manufacturing system they are enabled to locate the
the management in maintenance showed that a huge problem and to decide if it is possible to solve the
part of maintenance costs results from useless or problem by themselves cooperating in a group or a
badly made repairs. For example, sometimes specialist have to be consulted (for example the
scheduled maintenance is blind because the machine builder).This is possible now via Internet,
equipment, which is changed, could be in a perfect because most of the modern controls of machine
operating condition. This not only increases the tools use operation systems like Windows NT with
costs of production, but also decreases the lifetime
a connection to Internet, and most of the machine be used to point to approaching inadmissible
builders offer on-line service, not only for system states for indicating the limit of the space of
maintenance problems but also to give advise to acceptable performance. This is possible with the
solve manufacturing problems. help of a simulation tool that is not yet developed.
With this support and their experience when It should be stressed here again: the MAS serves as
running the system total productive maintenance a support system, leaving the strategic decisions to
will be effective for enhancing the availability of workers. To achieve this, software designers are
manufacturing systems. challenged to participate the experienced workers in
Often expert systems are in use to instruct the designing such systems.
operators to detect and to remove faults or repair
parts of machines and manufacturing devices. But, 4. SHOP FLOOR CONTROL
if even experienced operators are sticking long time
on advises of these systems their competence is Shop floor control is called the link between the
fading away. They loose their understanding what planning and administrative section of an enterprise
happens and what to do. and the actual manufacturing process. In SME´s
An alternative got developed by and John & Marzi with usually few staff to perform planning and
(2001). Operators use the information provided setting up of tasks, shop floor control becomes the
through a multi agent support together with their information backbone to the entire production
experiences what to do to achieve again the normal process.
operating state of the manufacturing system (Figure What SME´s need are low cost control means not
7). Two software only to avoid more or less complicated and
expensive technology with respect to investment
Multimedia Surface and maintenance but the work-organization should
Information of the Information of the be effective.
“vertical” agent “horizontal” agent Of course the shop floor needs software support to
enhance the flexibility and productivity. But
support is to be stressed, not determination what to
Exchange of
Information via do by automatic decision making.
KQML In small batch or single production (moulds, tools,
Diagnostic Strategies of the Diagnostic Strategies of the
spare parts) one needs devices for a dynamic
Agent Agent planning. Checking all solutions to the problems
arising at the shop floor, while taking into account
all relevant restrictions, short term scheduling
outside the shop floor either by manual or
C/C++ - API for time automatic means is not possible.
JAVA-API JAVA-API
critical functions
It is without sense to schedule the manufacturing
processes exactly for weeks ahead. One needs
devices which are capable of calculating time
Agent specific databases corridors. The shop floor can do the fine-planning
with respect to actual circumstances much better
Fig. 7. Architecture of the software agent system then the central planning. Human experience
(Marzi & John, 2001) regarding solutions, changing parameters and
interdependencies is the very basis of shop floor
agents are cooperating together and with the decisions and needs to be supported rather than
operators. A "horizontal agent" are looking for replaced. Figure 8 shows a network linking all
faults along the same level of a component-tree relevant modules to be used by skilled workers
regarding the functional dependencies between the at the shop floor. Apart from necessary devices as
components. The component-tree represents a
hierarchical order beginning with the machine itself
until the smallest component. A "vertical agent"
looks for faults in a component and its sub-
components. Both agents use implemented specific
strategies based on information of the machine
builder and experiences of former problem solving.
Obtained experiences when running and
maintaining the agent system can be fed to the
knowledge base (enhancing or altering the
component-tree) thereby enhancing step by step the
performance of the system. Also the diagnostic
strategies can be enlarged or changed by the
operators to make the system more effective and Fig. 8. Modules of a Shopfloor Network
adaptable to new challenges. The support could also
tool-setting and others an electronic planning board can occur unnoticed and will be perceived only
is integrated. The screen of the board is available at when the quality decreases, times of delivery are
all CNC-controls to be used at least as an exceeded or through put time increases. A decision
information on tasks to be done at certain support has to intercept such movements of the
workplaces to a certain schedule. As all skilled normal work performance against the limit of
workers as a group are responsible for the acceptable performance. Attempts have been made
manufacturing process they have beyond the access to formulate these problems as Dynamic
to information the task of fine-planning of orders Scheduling, using machine learning methods or
they got with frame data from the management. extensions of CSP discussed also by Dimopoulos &
Therefore they use the electronic planning board at Zalzala (2000). But all these attempts are not
the CNC-controls or alternatively at a PC besides. satisfactory respecting computation time and
Using an electronic planning board at the PC screen probably cannot take into account all possible
based on a database the decisions of the workers uncertainties.
with respect to shorten the through-put time or to Therefore a scheduling support with soft constraints
schedule delivering can be first simulated to developed by Fauser (www.fauser-ag.com), called
consider the access of tasks to available resources. JobDispo, results in an electronic planning board
So disturbances at the shop floor (breakdown of for the shop floor used by the group of operators as
machine tools, not just available tools, just missing mentioned above.
personnel) can be ruled out.
Sawaragi et al (2000) introduced an interface agent 6. OUTLOOK AND CONCLUSIONS
as a associate for human operators, preventing them
from a flood of data for decision making and on the A low cost intelligent automation is based on a
other hand avoiding computer based instructions for quick information providing and communication
taking action. According to Sawaragi et. al. an between customers and suppliers as well as inside
interface agent is a semi-intelligent software that the enterprise. Despite of partly expensive software
can learn by continuously "looking over the the complete costs of the manufacturing have to be
shoulder " of the operators as they are performing considered. The reconfiguration and the
actions. The agent has to coexist with the operators maintenance of automation systems can be very
so that it can evolve by itself as the operators time consuming whereas the direct costs of the
proficient level improves, but in a way to stimulate manufacturing itself are decreasing. Human skill
the operators creativeness, rather than to replace and experience can reduce idle time when
them by itself after some time. To be a human supported intelligently with suitable automated
collaborative an interface agent has to make equipment like machines with job-shop controls,
decisions analogous to humans. Sawaragi proposes electronic planning boards at their workplaces and a
a shift from classical normative decision making to support for fault detection and proactive
a naturalistic decision making concentrating on the maintenance.
proficient operators situation assessment ability
looking at a situation and quickly interpreting it, REFERENCES
using their highly organized base of knowledge
recognizing and appropriately classify a situation. Bhote, K.R. (1988). World Class Quality, AMA,
Again we are focusing on a support for human New York.
operators with software agents for job-shop Dimopoulos, C. and Zalzala, A.M.S. (2000). Recent
scheduling. A manufacturing system is a net of Developments in Evolutionary Computation for
different machine tools and handling devices. It Manufacturing Optimization. IEEE Transactions
processes information and material. It have to be on Evolutionary Computation, Vol. 4, No 2
operated to achieve an optimum with respect to Lay, G. (2000). Sackgasse Hochautomatisierung?.
material and time resources regarding the tasks to Mitteilungen Fraunhofer Institut Sysremtechnik
be done. This represents the normal work practice und Innovationsforschung, 22.
and may be defined as normal job-shop scheduling. Marzi, R. and John, P. (2001). Design and
Many attempts have been done to solve this Evaluation of a Competence promoting Decision
combinatorial problem, like CSP (Constraint Support System. Preprints of the 8th IFAC Symp.
Satisfaction Problem). An overview and discussion On Analysis, Design and Evaluation of Human-
give Dimopoulos & Zalzala, 2000). But static Machine Systems, Kassel, Germany
scheduling problems are not mapping the real world Morel, G. et al (2001). Maintenance Holistic
practice. Additional express orders or just not Framework for optimizing the Cost/Availability
available resources causes a fluctuation around this Compromise of MS. In: Proceedings of the 6th
normal work practice or performance of the IFAC Symp. On Cost Oriented Automation,
manufacturing system. Inappropriate planning and Elsevier Science Ltd., Oxford
control caused by time and costs pressure, tight Sawaragi, T. and Ogura, T. (2000). Concept
resources on personal, or deficiencies by illness, all sharing between Human and Interface Agent
unexpected issues are moving the system to the under time criticality. In: Advances in
limit of the space of acceptable performance. This Networked Enterprises, Boston, pp 269-278.