NSF Workshop on Logic Control for Manufacturing Systems

University of Michigan, Ann Arbor

June 26-27, 2000

Sponsored by NSF and ERC/RMS

Presentation abstracts

(Workshop Schedule)

Industrial Challenges in Logic Control

What Level of Control is Required for Machine Tools?

Bryan Graham, Lamb Technicon

Automated machine tools require sophistical logic to provide untended operation, and to insure adequate levels of machine and personnel safety. Customer expectations for sophistication are also increasing, creating machine control systems that are exceptionally complex. This presentation will focus on the various functional reqiurements for software and logic in future machine tools.

General Motors Powertrain Future Controls Software Requirements

Sushil Birla and Jerry Yen, General Motors

General Motors Powertrain (GMPT) has been the leader in implementing open, modular architecture controller (OMAC) technologies in its manufacturing applications since 1986. The interest in OMAC has been greatly expanded since 1994 because of the advancement of personal computer technologies and the publishing of the OMAC whitepaper by the U.S. automotive companies stating the requirements of OMAC technologies in automotive applications.

The purpose of this presentation is to introduce the GMPT future control software requirements. Having technologies and products that satisfy these requirements will accelerate the realization of the OMAC concept. An overview of GMPT and other end user's manufacturing requirements is presented first. These requirements are then translated into control system requirements in general and control software requirements in specific. Technical elements that are necessary to meet the control software requirements are presented next. Finally, technical topics that require further research are suggested.

International Standards for Open, Distributed Automation

James Christensen, Rockwell Automation

The IEC 61131-3 standard for programmable controller languages, adopted by the International Electrotechnical Commission in 1993, introduced modern software engineering concepts into the programming of industrial control and automation systems. A major advance in this standard was the introduction of programming library elements for the encapsulation and reuse of data (in data types), procedures (in functions), and both (in function block objects). In addition, IEC 61131-3 defined a unified suite of programming languages for such encapsulation and reuse.

The IEC 61499 project is expected to produce at least two Publicly Available Specifications (PASs) this year for a two-year trial use period. This specification extends the application of IEC 61131-3 concepts to the domain of distributed industrial process automation and control. Key to this extension is the replacement of the concept of a "program" by the concept of a "distributed application" represented by the flows of data and events in an extended function block diagram. Additional key elements include the introduction of event-driven state machines for control of algorithm execution, and service interface function blocks which can provide interfaces to any operating system-provided services defined as specified in the Open System Interconnection (OSI) service model. Software portability is supported by XML DTDs (Document Type Definitions) for reusable software library elements.

If time permits, a demonstration will be given of prototype IEC 61499-based software tools for the modeling and simulation of manufacturing process control and automation.

For more information, see http://www.holobloc.com/papers/opendist/

VALID Tool Set: Validation of Software Applications

Bertil Brandin, Siemens Research Labs

The scope and complexity of software applications has grown enormously in recent years, especially in applications for which component coordination is critical. As a consequence of this growth in scope and complexity, and due to possible legal consequences and image loss linked with faulty products, software development requirements have become more stringent and software validation increasingly important.

When comparing software development methods and development requirements for coordination applications, we realize that a productivity crisis is about to happen, which cannot be managed alone with traditional validation techniques (i.e. reviews, simulation and testing), but urgently requires validation technology innovation. Coordination applications are typically component-based, event-driven, and characterized by a clear separation between the tasks to be coordinated and their coordination. These particular characteristics, not encountered in general software applications, can be effectively exploited for validation purposes, and constitute - together with expressive, clear and verifiable application descriptions, and powerful verification algorithms - the basis of the validation technology innovation proposed by the VALID tool set.

The talk will first focus on current trends and issues in the development of validated coordination applications and the related customer requirements. Subsequently, the VALID tool set will be introduced. This tool set supports the design of software applications by means of a well-defined methodology and guidelines, it can be used stand-alone or in conjunction with other standard development environments and frameworks, and integrates modeling, verification, simulation, as well as the generation of executable code, test cases and documentation. The main benefits to software developers are:

  • Improved software quality with provably correct coordination of software components,
  • Shorter development times through faster incremental design and reduced integration and testing times (lead customers have reported 8-time speedups for successive design refinements, and 20% reductions in overall development times).
Systems control theory and formal verification theory constitute the basis of the automated verification algorithms used in the VALID tool set. They allow users to prove whether given behavioral properties of interest are satisfied, or otherwise provide counter-examples illustrating why such properties are not satisfied. These algorithms remain transparent to the user who is able to readily access them at the push of a button via very user-friendly interfaces.

Finally, application examples from medical engineering, robotics, automotive body electronics and GUI applications will be presented.

DES Formalism: Finite State Machines

From Programmable Logic Control (PLC) to Discrete Event Systems (DES)

Christos Cassandras, Boston University

The use of Programmable Logic Control (PLC) in manufacturing systems is akin to the use of assembly language in computer technology: Its relatively primitive nature makes the task of designing and implementing sophisticated controllers hard and inflexible. Yet, it is not difficult to see that basic control functions in a manufacturing environment involve discrete events that simply cause transitions in the "state" of the factory. The modeling framework of Discrete Event Systems (DES) is a natural setting that not only describes the event-driven dynamics of the factory, but also provides systematic means for analysis and synthesis of control mechanisms. This presentation will first provide motivating examples supporting the need for migrating from PLC to higher-level languages and control software and then introduce the modeling foundations of DES. Finite State Automata (State Machines), Timed State Automata, and Petri Nets will be discussed and illustrated with examples from the manufacturing system domain. Computer simulation will be seen to emerge as a systematic state trajectory generator of a DES. The basic supervisory control problem will be introduced in the context of Finite State Machines. Issues of verification and performance will be discussed, including a brief introduction to sensitivity analysis methods that are particularly useful in timed DES models.

Implementation Of A Hybrid PC/PLC Architecture For Manufacturing System Control

Beno Benhabib, University of Toronto

Current manufacturing strategies adopted by many industrial companies necessitate the use of automated production systems, which can be reconfigured and reprogrammed with great efficiency. Such systems are expected to provide manufacturers with a rapid response capability, namely, be flexible in order to cope with inflexibility in customer demands in terms of product variations and delivery times. In this context, the utilization of Flexible Manufacturing Workcells (FMCs) has been advocated for the fabrication of lower-volume, but wider-variety, products.

An FMC is defined herein as a system comprising automatic processing machines, typically serviced by robotic material-handling devices, working under the control of a supervisor. An FMC is, thus, supposed to have two levels of flexibility: At the device level, flexibility for part production or part transfer is achieved via re-configuration and re-programming. At the system level, a supervisory (workcell) controller is synthesized to monitor and control the workcell using commercial controllers (such as, industrial PLCs).

In the academic literature, FMCs have been modeled as Discrete-Event Systems (DESs) utilizing a variety of control theories, most notably by Petri Nets (PNs) and Ramadge-Wonham Automata theory. The literature has also advocated the use of PCs for the direct control of FMC, utilizing formal control theories and, thus, providing an open hardware architecture. In our work, however, we advocate the use of a hybrid PC/PLC supervisory-control system that is based on the strengths of each device: namely, PCs provide users with computational and human/machine-interface flexibility in order to efficiently communicate with PLCs that in turn allow efficient communication with cell devices and their control.

Our talk will present a novel and generic PC/PLC-based software/hardware architecture for the control of FMCs. The proposed implementation methodology is based on the utilization of any one of the available formal DES control theories in conjunction with state-of-the-art industrial programmable-logic controllers (PLCs).

The methodology has been verified to be a viable technique through its actual implementation in our laboratory using a PLC-controlled industrial robotic-workcell testbed. The specific control theory used is a combination of Extended Moore Automata and Ramadge-Wonham Automata, which has been developed by our research group.

The modular control software architecture has been developed for MS-Windows environments (running on one PC that is interfaced to the PLCs) and allows the use of different formal control theories as well as different commercial PLC hardware. The effective graphical user interface provides a transparent programming environment, where users are not expected to have a full knowledge of the formal control theory used.

Logic Control for Flexible Manufacturing Systems

Placid Ferriera, University of Illinois

In this work, we decompose manufacturing control into a three-level hierarchy: device/machine control, system co-ordination or logic control and performance control or scheduling. We address the problem of logical control (or supervisory control) of flexible manufacturing systems. Within this framework, we see the logic (supervisory) control module as an autonomous, configurable entity given the problem of managing a set of finite capacity resources. Jobs, when they enter the system, may request exclusive use of the capacity of these resources in any sequence during their processing. The supervisory control problem is to manage the execution of the jobs and the allocation of resources under the constraints that a) Jobs move according to their specified routing through the system's resources (stations) b) Resource constraints such as buffer capacity be respected at all times and c) the overall system maintains itself free of deadlocks (non-blocking) so that all jobs in the system can always reach completion. While simple in concept, such problems rapidly become intractable as the size of the system (number of jobs, number of resources or stations, and route length of the jobs) increases. The real-time nature of supervisory control exacerbates this problem further. Given the high cost of system failures and downtime and the size of systems in typical manufacturing situations, it is necessary to provide scalable, yet provably correct solutions to this problem.

In this talk we will discuss the automated generation of supervisors or logic controllers for flexible manufacturing systems. In general, the problem rapidly becomes intractable. We will discuss, conditions and typical manufacturing system topologies for which the problem (of generating and running a maximally-permissive supervisors) remains tractable and easy to implement. Schemes under which such controllers can be hierarchically decomposed (into cell, system and plant controllers) while maintaining their non deadlocking properties will also be discussed. Finally, some our current research in shared (human-computer) supervisory control will be discussed.

DES Formalism: Petri Nets

Modeling, Control, and Performance Analysis of Manufacturing Systems using Petri Nets

Alan Desrochers, Rensselaer Polytechnic Institute

Petri nets are an effective tool for modeling, control, and performance analysis of automated manufacturing systems. They can handle problems that cannot be modeled by queueing theory, and they avoid the trial and error approach of simulation.

The modeling problem is characterized by concurrent and asynchronous events. Petri nets are well suited for modeling manufacturing systems because they capture the precedence relations and interactions among these events. In addition, a strong mathematical foundation exists for the analysis of system properties such as deadlock, conflict, and boundedness.

The Petri net model can also be used as a real-time controller for a manufacturing system. The flow of tokens through the net establishes the sequence of events to carry out a specific manufacturing task.

Petri nets are also a very valuable performance analysis tool. When time is added to the firing of the transitions, it becomes possible to calculate such measures as throughput, production rates, average machine utilization, and the probability that a machine will be blocked or starved.

Applications to a machining workstation will be presented along with numerous examples from transfer lines and production networks.

Petri Nets and Discrete-event Control Methods for Logic Controller Design

MengChu Zhou, New Jersey Institute of Technology

For the past decades programmable logic controllers (PLC's) using relay ladder logic (RLL) programming have been the workhorse for controlling event-driven industrial automated systems. RLL proved flexible compared with hard-wired RLL control implementation due to its feature of software implementation. As automated systems are more complex, they become more difficult to understand and maintain. It takes tremendous effort to accommodate specification changes that become frequent to meet today's flexible and agile automation needs. Several methods emerge to overcome the shortcomings of RLL. Petri nets (PN's), initially proposed as a modeling tool, have been developed as such a method. This presentation overviews a number of discrete-event control design approaches including
  • Direct Implementation of RLL,
  • Instrument Society of America Logic Diagrams (ISA standard S5.2-1976),
  • Timing/sequence diagrams,
  • State diagrams,
  • Real-time Petri Net models, and
  • Grafcet/Sequential Function Charts (IEC 1131-3).
This presentation focuses on Petri nets and RLL. It presents an industrial scale system to compare them so that the advantages of Petri nets like approaches are fully recognized. The criteria are the understandability that relates to the ability to evaluate the programmed logic, verify its correctness, and maintain the control system, and flexibility that relates to the easy modification of logic when the specification changes. The presented results support that PN like advanced discrete event control design methods are better than RLL in terms of understandability and flexibility of a resulting control design.

Modular Logic Controllers for Reconfigurable Machining Systems

Dawn Tilbury, University of Michigan

The automotive industry demands large quantities of parts to be produced rapidly and with high quality. To accomplish this goal, high-volume transfer lines are used instead of more flexible CNC machines for the machining operations. In these systems, several machines are linked together by a dedicated material handling system (the transfer bar) to provide the desired quantity and quality of cylinder heads, engine blocks, etc. A discrete event supervisory system, called a logic controller, coordinates the parallel and synchronized operation of these machines. The logic controller is typically implemented on a PLC. In current practice, the logic controller is programmed in a low-level language by an experienced control engineer; long testing and debugging cycles are commonplace.

In this talk, we describe a new formal representation for a logic controller of a high-volume transfer line using Petri nets. The starting point for the control design is a timing bar chart, created by the mechanical designers of the machining system. The timing bar chart can be converted to a Petri net representing the logic control for the normal operation cycle. This formal representation allows the logic controller to be verified as live, safe, and reversible before it is implemented on the hardware. The incorporation of other control modes (manual mode, failure recovery, etc.) into the logic control framework will also be discussed. The Petri net structure is easily converted to Sequential Function Charts, a standard PLC programming language, for ease of implementation. The logic controller is constructed in a modular fashion, following the modularity of the transfer line hardware, enabling the transfer line control system to be more easily modified or reconfigured. This talk represents joint work with Pramod Khargonekar and Euisu Park.

7/13/00 dmt