Article Contents:

• Abstract
• Keywords
• Introduction
• Some of the challenges...
• The distributed manufacturing model...
• Sequencing events...
• Process clock conditions...
• Dynamic scheduling of FMS with integrated tool management...
• FMS dynamic scheduler specification summary...
• The WAIT procedure...
• The DELETE algorithm...
• The ASK algorithm...
• Buffering issues...
• What is interactive multimedia and why is it important in FMS?
• Integrated FMS tool management system...
• Conclusions and summary...
• Acknowledgements...
• References and bibliography...

Abstract

Go to top

We have to recognize that traditional, mass-produced, highly-controlled static and rigid design and manufacturing/ assembly methods cannot cope with the increased amount of information, knowledge and inferencing requirements of the customers of this rapidly changing world.

The life cycle of critical design and manufacturing technologies is increasingly contracting. As a consequence, we need to mass-customize, be pro-active, and customer driven at a very high quality and low cost. Furthermore we need to continuously innovate and improve. We have to accept paradigm shifts at a much faster rate than ever before, otherwise "somebody else will satisfy our customers" and consequently "we'll be out of business".

The challenge we face is to narrow the gap between the theory and the practice, and offer computable methods, tools and technologies, that will keep enterprises highly responsive to market changes and produce goods (and services) on demand at a high quality and relatively low cost in the forthcoming "Knowledge Age". This paper propagates the above, by putting some light on the challenges, as well as by offering generic algorithmic solutions to the problem of dynamic scheduling with integrated tool, fixture, robot hand management and multimedia support in distributed, lean and flexible manufacturing and design systems (including flexible assembly systems employing human operators as well as robots) that operate on a global basis. Our approach is that we start with the distributed macro level offering a global overview and then we "zoom in" to the networked factories, their FMSs, right into algorithms and rules of dynamic schedules, with multimedia and tool management - ending at the tip of the tools!

Keywords:

Go to top

Dynamic Scheduling of FMS, FMS/ FAS, Flexible Design, Manufacturing and Assembly Systems, Tool Management, Robot Hand Management, Engineering Multimedia, Distributed Enterprise, Logistics, Global Manufacturing

Introduction

Go to top

Concepts such as lean and flexible manufacturing, time to market, total quality management, local design, global manufacture, concurrent or simultaneous engineering, intelligent design and intelligent manufacturing systems are widely accepted. These methods and technologies require multiskilled and well - educated engineers and managers as well as lean, flexible, and feedback controlled manufacturing technologies, such as cellular manufacture, CIM, Concurrent Engineering and FMS run and producing goods based on customer orders using dynamic scheduling methods between facilities, plants and shops for all resources, including parts, machines, tools and fixtures.

Furthermore, it is important to recognize that we are living in an era when the customers, not the designers or salesmen, are the kings. Customers require increasingly better products at a lower cost. In other words, products require continuous improvement and change, therefore lean, agile and flexible design and manufacture, and the appropriate level of automation, must be provided throughout the life cycle of product development, as well as during its maintenance and finally its de-manufacturing.

The enterprise with a future is a continuously learning enterprise, that thrives on creating and employing new knowledge using its own resources as well as by networking and collaborating with others (i.e. the concept of "1+1=3"), often on a global basis, using technologies such as EDI (Electronic Data Interchange, including business as well as engineering data), the Internet and various Intranets.

Therefore the understanding and the processing methods of data, information and knowledge and systems are of utmost interest. This is why we have to focus on multimedia education too. In the education business the customers are the learners, i.e. students entering access courses, college and university courses, mature students who are prepared to study in the evenings at home, or at the University, or in Open Learning Centers, or most importantly on the job, next to their cell controllers and CAD computers. Furthermore, there are a large number of continuing education students and other professionals seeking new focused knowledge in this rapidly changing and extremely competitive world.

Before going further, let us spare a few words on terminology. In our discussions the term system refers to a specific configuration of objects. Then the term process refers to the pattern of behavior of an object such that the pattern is composed of a few primitive constructs. Therefore, a communicating process, such as for example a module of the dynamic FMS scheduler, is a subsystem of the environment, and a subsystem is a cluster of communicating processes.

Some of the challenges of our dynamically scheduled, highly parallel and distributed "Knowledge Age"

Go to top

Advances in computers have provided exciting opportunities in Lean and flexible design and manufacturing, rapid prototyping, Desktop Publishing and Desktop Design (CAD/CAM), but we are entering now into a new era of Knowledge Engineering at the desktop, covering all aspects of the enterprise.

Consider the fact that due to the advance of communications and automation technologies it is now common, that a product is designed in one part of the world, manufactured on an other part of the world (often another continent), assembled again somewhere else and used (with various minor customization changes) everywhere in the world. In other words we need to deal with a variety of combinations of local design and manufacture and global design and manufacture.

In order to attempt to illustrate this "global thinking" with simple models, please refer to Figure 1 to view the Data-flow Diagram (DFD) of a simplified small - to - medium size contract manufacturing company. (Note the lack of any serious design function in this model!). In this diagram ellipses mean processes, parallel lines are data stores, or data bases and square boxes are data sources, or data sinks, or sometimes both.

Figure 2 illustrates a more advanced company model. Note the various feedback loops between processes, including design and others, as well as the various internet/ intranet/ satellite links between various processes. Overall, this company could work well in cyberspace because it has the communications capabilities!

Please click on the image to view Figure 1 full size

Please click on the image to view Figure 2 full size

As an example for another "globally distributed" solution Figure 3 illustrates two companies sharing various functions and processes via the cyberspace using interactive multimedia based web-techniques. (Note, that we will discuss the importance of interactive multimedia further below). The core issue is, that in the forthcoming "Knowledge Age" multiple facilities and multiple firms must collaborate and simultaneously compete in a distributed, global environment. This figure attempts to illustrate with the aid of data flow diagrams, that there are several multimedia, Internet and Intranet EDI links possible to fuel design, marketing, flexible manufacturing and other collaborative, global logistic operations. Most importantly, one can design at one site and manufacture (and dynamically schedule) several thousand miles away at an other site using EDI facilities in such "distributed enterprises". This is feasible because the available IT technology, and it is cost effective because of the local logistic supply chains that the factories in question need to operate successfully.

Please click on the image to view Figure 3 full size

This means several things ([1] to [4]). Most importantly, we need to:

* Improve the production efficiency

* Improve the responsiveness to market needs and changes

* Reduce cost substantially

* Shorten product life cycles, in particular in the automotive, semiconductor, electronics, and other high technology industries

Naturally, the gap between theory and practice must be reduced to solve the above listed problems. Furthermore, one must address the dynamic and interdependent nature of modern production facilities that are needed to bridge this gap, as well as that in the new "Knowledge Age" multiple facilities and multiple firms must collaborate (and compete) simultaneously (Figure 3 ).

The distributed model of dynamically scheduled FMSs operating on a global basis

Go to top

In order to appreciate the importance of the distributed model of dynamically scheduled FMSs operating on a global basis ([5] to [12]), let us briefly summarize the differences between sequential versus distributed processes in such systems with the view, that by understanding the model we can generalize it and expand it to help us to understand our distributed enterprise model, that includes not just flexible manufacturing, but also shared/ distributed/ concurrent engineering, marketing and other functions too.

By definition, a sequential process is considered to be a spatially localized event structure that is described by a pattern of events separated by regular or irregular time intervals. Consequently the events in such processes or systems are totally ordered.

On the contrary, in a distributed system the processes are partially ordered and spatially separated in such a way that the transmission time, T, that is required to transmit a message between process P(i) and process P(k), is not zero.

Based on the earlier works of Ranky, Lamport, Anger, Chandy, Mistra, Tanik and Chang ([1 ], [2 ], [29], [30], [31], [34] to [37]) let us discuss a new, extended algorithm to illustrate the operation and messaging processes in our distributed model of dynamically scheduled lean and flexible design and manufacturing systems operating on a global basis. In order to simplify the understanding of our model, let us assume that all processes that are subject to our analysis, actually occur within the same time frame. (Note, that this simplification does not effect the generic nature of the algorithm, but simplifies its understanding at this stage).

As examples, this algorithm works well in a distributed FMS control environment, in which many cells, machine tools, robots, AGVs, etc. are dynamically re-scheduled (Figure 4), or in cases where multiple CAD/CAM workstations concurrently communicate with each other (see Figure 2 and Figure 3 again). In general it works well in a distributed system in which each process or activity is a model of a subsystem of a larger system. (In this context consider a dynamically scheduled workstation, a cell, an FMS system, a CIM factory, or globally networked CIM factories as various models of subsystems that are part of a larger system).

Please click on the image to view Figure 4 full size

The logical time mechanism as well as the time stamping technique used in our algorithm can be used for dynamic resource allocation for overall FMS dynamic scheduling. In most cases, due to the just-in-time need (versus e.g. capacity maximization) of such systems (as FMSs), the priority allocation scheme is on a first come- first serve basis.

This means that in this priority allocation scheme the resources that the FMS has in terms of setup, processing, tooling, clamping/ fixturing, material handling, part and tool changing, etc. is to be dynamically allocated and re-allocated (meaning de-allocated if the capacity is limited) at instances, represented by events of the time span analyzed. (Note, that as an example, the shortest possible time span is typically what it takes to change a pallet for a milling center, or load/ unload a part for a turning center by means of a robot. From a dynamic scheduling point of view this is considered to be the shortest possible time span available for such analysis or dynamic re-scheduling of the FMS. In practice, this is approximately 5 to 7 seconds for milling centers and around 3 to 5 seconds for turning centers, meaning that very fast algorithms are needed.

Assuming that the planning horizon allows the luxury to look at a longer time span, then the sum of the shortest processing time with allowances for setup, part and tool transportation times is considered to be the shortest possible time span available for such analysis (or dynamic re-scheduling).

As examples for the above, consider the following algorithms for calculating such critical times:

Procedure A

Begin

Sum:=0;

For I:=1 to Task_No do

Sum:=Sum+Completion[I];

MFT:=Sum/Task_No

End;

End (* Proc A *);

 

Procedure B

SumA:=0;

SumB:=0;

For I:=1 to Task_No do

Begin

Writeln (Print, trunc(Task[I]:10, y[I]:10:2,

Completion[I]:10:2, Trunc(Z[I]):8, X[I]:10:2);

SumA:=SumA+Completion[I]*Z[I];

SumB:=SumB+Z[I];

End;

WMFT:=SumA/SumB;

End (* Proc B*);

It is obvious of the above, that as the most important design criteria of the dynamic scheduler for distributed FMS systems with integrated tool management is to manage the available FMS resources, determined at the earliest time stamp (T + d) from among real-time requests of various concurrent events. Furthermore, this means that a process P(i) must be able to wait for a request with the earliest time stamp (T(i) + d(i)).

In practical terms, in real-world dynamic FMS scheduling, a cell or workstation, or an other resource is not allocated by the dynamic scheduler to take the next task until there is at least one message in every queue to the effect that the request with the earliest time stamp can be selected and at the same time the process can update its clock.

In order to describe the algorithm, let us define the following conditions and rules:

1. If a and b are events in the same process P, and event a precedes event b, then a "implies" b. (Please note, that in order to make sure that all readers of this text see the same characters over the web, even with older (i.e. pre-html 4.0) browsers, instead of the usual "right arrow character" we use the mathematically correct term "implies" in our description).

2. If event b is the recipient of the same message sent by another event (k) as well as event a, then a "implies" b

3. In all above conditions, the relation "implies" on the same set of events in a distributed system is the smallest relation that satisfies each condition. We also assume, that "implies" is not reflexive.

4. Furthermore, the relation "implies" is transitive, meaning, that if a "implies" b and if b "implies" c, then a "implies" c

5. Also note, that events in our model are partially ordered, meaning, that if neither a "implies" b nor b "implies" a, then a and b are two concurrent, or simultaneous events.

Sequencing events

Go to top

Next, we have to turn our attention to the process of sequencing events in our distributed model. First, this will be defined by a set of axioms as follows:

* As the events of the system unfold in real-time, the logical system clock mechanism (Clock ) should operate without interruption.

* Let Event = U Event(i) be a set of events in our distributed model.

* Our system clock, Clock: Event "implies" N is defined (where N = natural numbers) as the set of logical clocks for each event a in process Pi as well as for each process Pi , Clock(i) = Event "implies" N by letting

Clock(a) = Clock(i)(a)

* If a and b are concurrent events (meaning that a E Event(i) (read as: event a is an element of set Event(i)) and that

b E Event(j) ) then it is possible for the event Clock to have

Clock(a) = Clock(b).

Furthermore to the above listed criteria, the mapping clock Clock must meet the following system clock condition:

For any events, a and b, if a "implies" b then Clock(a) < Clock(b)

The system clock condition is equivalent to the following two conditions imposed on the process clock:

[Condition_1]: In a process P(i) if a "implies" b then Clock(i)(a) < Clock(i)(b)

[Condition_2]: If event b is the recipient of the message sent by process P(j) and if event a is the sender of a message by process P(i), then Clock(i)(a) < Clock(j)(b)

Process clock conditions

Go to top

In order to meet the two process clock conditions, as defined above, each process must implement its own logical clock as follows:

Since there are two consecutive events in a process marking the start and the end of an information processing activity, at the occurrence of each event, the logical clock must be incremented by an integer number that is proportional to the real elapsed time. This means:

[Rule_1]: Each process P(i) increments Clock(i) by an integer number that is proportional to the real elapsed time between the current event and the immediately preceding event.

 

[Rule_2]: If event a is the sender of a message by process P(i )then that message must be time stamped with the value

T = Clock(i)(a).

If event b is the recipient of a message sent by process P(j) then the process P(j) must increment its logical clock Clock(j) to the greater of the value proportional to the elapsed time between the current event and the immediately preceding event the value of T + d where T is the time stamp and d > 1 is defined with reference to the system time.

It must be noted, that in order to assure, that the concurrency remains asynchronous, [Rule_1] assures that [Condition_1] is satisfied and that at the same time, by means of logical clocks hides the negligible differences between the current values of the real-time clocks.

Dynamic scheduling algorithm of FMSs with integrated tool management

Go to top

The most important operation control activities in FMS/ FAS identify three levels at which simulation and optimization is required, prior to, or during FMS/FAS part manufacturing. The three levels are as follows:

1. The factory level, or business level handled by the business system of CIM.

2. The FMS off-line level representing scheduling, simulation and optimization activities prior to loading a batch or a single component on the FMS, (handled sometimes by the CAM system, sometimes by the FMS part programming computer).

3. The real-time controlled level handled by the FMS/FAS operation control system, a dynamic scheduler with integrated tool management and multimedia support, representing a situation where the parts are already physically as well as logically in the real-time controlled environment.

Due to its complexity, a truly integrated approach is required in designing a production rule base to provide the job description for the FMS dynamic scheduler. This is because the dynamic system relies heavily on the knowledge base as represented by the rule base, and an overly restrictive rule base will lead to inefficient, at times even wrong, decisions. In other words, such a structure should represent all the multi-level interactions and their possible precedence rules that relate to the manufacturing process planning and processing decisions in an FMS. This turns out to be a difficult task ([13] to [22]). To underline this again, please refer again to Figure 2 , reminding you of the fact that we are working with a networked enterprise, Figure 3 illustrating again that enterprises can be globally linked and distributed, and Figure 4 showing a relatively complex FMS with all its crucial process capabilities that should be reachable globally).

Rules are nothing but representations of constraints. Various types of constraints were studied. It was also realized at an early stage of the development that a user friendly, ionized, multi-windowed interactive multimedia with a graphics interface, with videos and photographs must support the process of building up the production rule base. (Note that further research is currently being carried out by the author in to representing FMS process planning and job control knowledge.)

It should be underlined, that the application of multimedia at this level is extremely beneficial in terms of part program preparation, teaching/ training operators on setting up parts, fixtures, tools, machines, for troubleshooting, for regular maintenance, at CNC level programming and other tasks (Figure 3).

In order to get acquainted with the system design concepts of our validated and tested FMS dynamic scheduler let us discuss a simplified, commented example that illustrates the way parts get processed using various resources of the FMS, including the option of dynamic re-routing without any disruptions to the system. (Please note, that one of the most important benefits of a "well designed" FMS over a transfer line is this, often misinterpreted, dynamic re-routing capability).

---

Let us suggest that as you read the algorithm below, simultaneously open Figure 4 in the browser in a separate window by choosing "New Window with this Link..." by keeping the mouse button pressed down immediately after you have clicked on the hyperlink.

---

BEGIN FMS_Part_program/ID Code of this program: A121297

(* Note that this is the top level of the FMS part program. This program guides all major pre-programmed activities of the FMS dynamic scheduler with its integrated tool management system and interactive multimedia interfaces (where applicable). It is basically a structure, rule oriented job description on what the FMS should do with a certain part number in the FMS. (i.e. Note, that not just one machine, or workstation, or cell, but the whole system is programmed this way!)

It describes all pallet level, i.e. setup and operation level programs, including pointers to CNC code, in the part program database, pointers to the tooling and clamping/ fixturing needs, in the tool management database, probing and assembly robot hand needs and others, including exactly what is needed, at which workstation, or cell, and the data on when is the resource needed - note, that time is calculated here relative to the completion time of the current process on the designated, programmed resource. The real-time system clock follows the time stamping rules described earlier *)

 

BEGIN Pallet_program/Code: No.1, Pallet_code: P1; Part_code: Part ABCxyz

(* Note that part/ pallet setup data is stored at this level. i.e. the workpiece, or part and the pallet and clamping/ fixturing devices are specified for the dynamic FMS scheduler. Furthermore note, that as a general principle these ID numbers act as pointers to the actual data - in this case part and fixturing/ pallet data - stored in the appropriate part, fixturing/ pallet databases of the FMS system.

This enables all the checks the dynamic FMS scheduler must perform before issuing the EXECUTE (real-time) instruction to the FMS hardware, using the current time - as per the stamped clock time *)

 

EXECUTE_operation/ Code: No1, Tool_file P1_No.1_T01;

(* Note that tooling data for each operation is stored at this level in a tool file, obtained from the CNC part program, or a high level CAM programming package. This enables the FMS dynamic scheduler to compare the actual contents of the CNC tool magazine with the desired, or programmed need - given here.

If there is a discrepancy an automated multimedia REPORT is generated to fix the problem as soon as possible. If it is not possible, e.g. because the desired target machine/ cell is busy or broken down, the next alternative is chosen - see <Condition true/ false below *)

 

EXECUTE _operation/ Code: No2, Tool_file P1_No.1_T02;

IF <Condition true> THEN

(* Note that a condition, true/ false, analysis can mean resource availability check, or a condition relating to priority change, or others.

A resource availability check typically includes the following: Is the machine up and running? Is it available/ or busy? Are the appropriate tools in its magazine? Is the final checksum OK to load a palletized part and execute the program? Do we need to perform secondary optimization at cell level, and re-schedule the cell input queue ? , and others *)

BEGIN

EXECUTE _operation/ Code: No3, Tool_file P1_No.1_T08;

EXECUTE _operation/ Code: No4, Tool_file P1_No.1_T06;

(* More alternative operations can be described here...Note, that alternative operation specifications at this level represent built-in redundancy for alternative choices that the FMS dynamic scheduler can take in order to complete the job on due date or earlier.

This is an excellent technique to increase the decision space and "give more room" and often time for the real-time dynamic scheduling system to complete the desired task. This feature will in most cases increase the productivity of the FMS. It must be underlined though, that the issue is to complete jobs on due date, not to fully utilize the capacity of the FMS, and by doing so downgrade it to a transfer line... as it happens in some cases in industry... *)

END

 

ELSE

BEGIN

EXECUTE _operation/ Code: NoA3, Tool_file P1_No.2_T04;

EXECUTE _operation/ Code: NoA4, Tool_file P1_No.7_T12;

END;

END Pallet_program/Code: No.1, Pallet_code:P1;

 

BEGIN Pallet_program/Code: No.2, Pallet_code:P2;

(* Note that this could be another setup, or even an alternative setup for the same part. Setup data is stored at this level. i.e. pallet and fixture. We do not detail it, just its structure *)

 

EXECUTE _operation/ Code: No1, Tool_file P2_No.2_T01;

(* Note that tooling data for each operation is stored at this level in a tool file, obtained from the CNC part program, or a high level CAM programming package, such as GNC, etc. *)

CASE OF <Condition_1 true> THEN

(* Note that a condition, true/false, analysis can mean resource availability check, or a condition relating to priority change, or others. Also note, that we have introduced a new way of describing conditions with the Case...Of statement *)

BEGIN

EXECUTE _operation/ Code: No2, Tool_file P2_No.2_T012;

EXECUTE _operation/ Code: No3, Tool_file P2_No.2_T09;

END

CASE OF <Condition_2 true> THEN

(* Note that a condition, true/false, analysis can mean resource availability check, or a condition relating to priority change, or others *)

BEGIN

EXECUTE _operation/ Code: NoA3, Tool_file P2_No.2_T04;

EXECUTE _operation/ Code: NoA4, Tool_file P2_No.2_T12;

(* More alternative operations can be described here...*)

END;

(* More alternative operations can be described here...*)

END Pallet_program/Code: No.2, Pallet_code:P2;

(* More pallet programs can be described here before inspection *)

 

BEGIN Pallet_program/Code: No.3 Inspection, MUST go onto Machine ID#: CMM1

(* Note, that this command "MUST go onto Machine ID#..." will force the FMS dynamic scheduler to select the designated machine. In this case the Co-ordinate Measuring Machine. By invoking this command, the dynamic scheduler in collaboration with the FMS tool management system will check the number and type of probes needed/ available at Machine ID#: CMM1 (actually it will check the logical probe ID#s in its automated probe changer).

This will make sure, that by the time the part gets to this machine, all conditions are fulfilled. If for any reason at current time this is not the case, the FMS dynamic scheduler will have to call upon the tool management systems reporting routines and instruct the tool manager on duty, or automatically, to load up the CMM with the necessary tools/ probes.

Note, that the actual probe IDs are listed in the actual CMM program headers below!

Note that setup data is stored at this level. i.e. pallet and fixture. Inspection is carried on this part and this pallet. Pallets can be loaded automatically onto Co-ordinate Measuring Machines *)

 

EXECUTE _operation/ Code: No1, Tool/ Probe_file P1_No.1_T012;

(* Note that probing/ tooling data for each inspection/ test operation is stored at this level in a tool file, obtained from the CMM part *)

 

FEEDBACK & CORRECT

(* The FEEDBACK & CORRECT instruction is a very powerful command, since it will instruct the FMS dynamic scheduler that upon an out of tolerance measurement it should check - in the given order - on the responsible surface of the part, then the operation ID#, then the tool ID#, then the tool's actual set correction data (e.g. length and radius), identify the Machine/ workstation ID#, Cell ID#, and even report on exactly where the responsible tool is at current time - using the reporting function below.

This will enable automated tool length correction at CNC level, automated sister tool selection at CNC tool magazine level and automated error messaging at FMS system control and tool management levels - using multimedia - to all interested parties. Note, that the detailed data flow and control system of this automated, self correcting feature was a very novel concept in the early days of FMS and even today it is a rarely found option. Furthermore it underlines the importance of self correcting, intelligent machines integrated into a system such as an FMS *)

 

REPORT

(* This command automatically creates a multimedia report, stores it in the FMS System History Database and sends it to all parties that need to act immediately, as well as to those who must be kept informed only *)

 

EXECUTE _operation/ Code: No2, Tool/ Probe_file P1_No.2_T002;

(* Note that probing/ tooling data for each inspection/ test operation is stored at this level in a tool file, obtained from the CMM part *)

 

FEEDBACK & CORRECT

(* More FEEDBACK & CORRECT activity can be generated here - if requested - in principle, as described above *)

 

REPORT

(* More multimedia REPORT generation can be activated here - if requested - in principle, as described above *)

 

EXECUTE _operation/ Code: No3, Tool/ Probe_file P1_No.3_T006;

(* Note that probing/ tooling data for each inspection/ test operation is stored at this level in a tool file, obtained from the CMM part *)

 

FEEDBACK & CORRECT

(* More FEEDBACK & CORRECT activity can be generated here - if requested - in principle, as described above *)

 

REPORT

(* More multimedia REPORT generation can be activated here - if requested - in principle, as described above *)

END Pallet_program/Code: No.1, Pallet_code:P1;

(* Further alternative pallet programs can be described here...*)

 

(* Automated / manual assembly instructions could follow here in principle in a similar structure to the above *)

END FMS_Part_program

The virtually random part scheduling aspects of the system

Go to top

What can be easily recognized of the above structure, is that to be truly flexible, the FMS/ FAS system should be capable of making real-time control decisions that include dynamic scheduling, variable part routing and fast secondary optimization (meaning the reordering of parts at input queues of cells).

A system such as this which allows virtually random part scheduling can demonstrate the true flexibility of the FMS/ FAS concept. Within this overall goal, the secondary objectives are:

1. To show the general applicability of production rule bases and also how this new concept can be extended to other aspects of CIM.

2. Develop the system as a component in the overall control structure of an integrated CIM system. Hence, the system should be modular and integratable with other systems.

There are two distinct aspects to the development of such a system:

1. The interactive multimedia interface to build a production rule base. Note that the real expert on the shop floor is often not a scheduling specialist. Nor is he expected to be familiar with how to use a rule based system. Hence this component of the system should comprise a very friendly graphic interface which shows the rule structures as they are built up. This sub-system could generate a file e.g. in a simple structured multimedia format, which can be understood both by shop floor managers and by engineers.

These files, constituting the production rule base, will be accessed by the dynamic scheduling program, which will choose from different candidates depending on the real-time status of the FMS and the processing needs sent in the form of orders by the business processing system of CIM.

2. The dynamic FMS scheduling program. This program accesses the rule base and chooses from the alternative production routes offered, and makes real-time control decisions based on fast secondary optimization well as tool/ fixture/ machine (resource) availability.

The overall specifications of the FMS dynamic scheduler can be summarized as follows:

Go to top

At any instant of time during a given period of time, such as a shift:

1. The system should be capable of interpreting and evaluating the production situation by reading FMS system status data via the distributed FMS data processing network, from the cell controllers in real-time. (It should also display this status in a readily understandable way, using multimedia, if required).

2. It should read and evaluate the current dynamic (transient) input data which may include breakdowns (e.g. tool breakage at a cell, or the breakdown of the entire cell, or the material handling system, etc.).

It should cope with job priority changes, job deletions and with the rearrangement of part input queues and buffer stores. (This data should be used together with the predetermined off-line operation control input data in order to generate a task for any required instant of time).

3. It should access the variable route part program data base (i.e. the production rule base) along with other manufacturing data to schedule the various jobs from the job list on different cells using a simple built-in secondary optimization routine. (If required, the dynamic schedule, showing variable routes for this instant of time should be displayed).

4. In case of an error in the operation of the FMS or in this system, and if either of these is irrecoverable, appropriate messages must be generated. A self-learning, expert diagnostic module should be incorporated which could learn from such failures and suggest different ways of changing or updating the rule base.

5. Relevant statistical information and schedule evaluation criteria based on the operations during the particular shift should also be generated as part of this system.

6. The interactive multimedia user interface should be such that even a non-technical shop-floor worker should be able to supervise/operate the real-time operation control system.

7. The system should be modular in such a way that at each sub-system level of implementation any of the sub-system modules should consist of two layers, these being:

* The top level layer, which is FMS architecture independent, i.e., allows programming at a generic level that can work in virtually any truly flexible FMS architecture.

* The installation specific layer itself should be such that the FMS could be expanded and modified without affecting the rest of the system, i.e., by localizing the implementation specific details here in this layer, the rest of the module is hidden from these details.

Furthermore the dynamic FMS scheduling system with integrated tool management should incorporate procedures such as DELETE workpiece from any production route and/or the buffer store, WAIT with the production of a particular workpiece, (because the system is not capable of accepting further components for the time being), ASK the status of a component, REARRANGE any production route in real-time because of a break down or maintenance, and others.

The WAIT procedure

Go to top

The WAIT procedure for example could be implemented using this algorithm:

if the production route can be loaded with further components (in other words if the queue is not full yet) and/or if there is a free place in the buffer stores,

then the part should be placed logically and physically as soon as possible into the production route,

else the part should wait in the automated warehouse, or, if already delivered to the system, in a free buffer location, close to the cell.

The DELETE algorithm

Go to top

As a further example, let us show the way the DELETE algorithm could be designed:

if the processing of the workpiece in any of the possible production routes is deleted,

then remove the workpiece logically from the production queue and physically from the buffer store without interrupting the continuous production of the system.

 

The ASK algorithm

Go to top

Finally, the ASK algorithm could be as follows:

if the dispatcher and/or subsystem sends a request for workpiece status,

then list those FMS cells and selected routes which the component has passed through and also those which are planned to be met in the current schedule. Upon request provide accurate, actual due date (time) and resource related information.

It is clear from these algorithms that, for each possible manufacturing route, two lists are required: a waiting list of workpieces to be manufactured on each selected route, gained as a result of the frequent simulation of the system, and the current status description of the availability of the system, gained using real-time status information.

Buffering issues

Go to top

Normally components waiting to be manufactured via the same route in the FMS should be taken one-by-one from the front of each queue (FCFS, first come first serve) unless otherwise programmed, until there are no components in the queue, but the rearrangement of components in queues should also be allowed using different scheduling rules. This feature is important if a job must be processed in the shortest throughput time, or in "panic situations" when the system needs to be rescheduled or recover quickly.

Having allowed for real-time alternatives and for re-scheduling logically and/or physically existing queues, the program must keep track of the number of parts with which a production route can be loaded and must also know the number of workparts already accepted for production for each route.

Most FMSs have some part buffering capability (see Figure 5 again). This may not be for scheduling reasons, but for technological, i.e. process planning reasons (e.g. the part must cool down somewhere before an accurate inspection procedure is performed). Some level of buffering is useful and necessary because of reliability reasons. (The actual number of buffer store locations should be established on the basis of simulation and experience.)

Cells often have some buffers too (e.g. imagine a chain type pallet magazine of a milling cell. Note that the ideal situation is when there is one AGV input and one output docking station for each cell). The reason for this is that by providing a part in the input queue of the cell just before the currently processed part is finished at the particular cell, the cell is kept running at its highest efficiency level, since time is only "wasted" for part changing, typically lasting only 5-7 seconds. The other important point to note is that well-designed part buffers offer a direct access pickup/load facility, making the rescheduling process in the queues short, simple and dynamic.

What is Interactive Multimedia and why is it important in FMS?

Go to top

Interactive multimedia combines and integrates text, graphics, animation, video and sound. It enables learners to extend and enhance their skills and knowledge working at a time, pace and place to suit them as individuals and/or teams and should have a range of choices about the way they might be supported and assessed.

In other words:

* The student has a choice and the freedom to learn

* He, or she is supported by the multimedia based learning materials and technology

* The tutors are creating an effective, enjoyable learning environment and infrastructure

Interactive multimedia is essential for all operators, engineers, managers, etc. to be able to integrate with, operate within and partially control the typically highly computerized and increasingly complex systems we all have to work with. As an example, an FMS is a highly automated, distributed and feed-back controlled system of data, information and physical processors in which decisions have to be taken real-time.

This is only possible if all information processors (including the human resources of such systems) are "well informed", meaning that they have the exact information at the exact time, format and mode they need it to take responsible decisions within given time constraints ([27], and [32] to [33]). Figure 5 illustrates one of the many cell operator support, educational, training, quality control, production control, maintenance and other application opportunities of multimedia in FMS).

The integrated, multimedia supported FMS/FAS tool management system

Go to top

The design of an FMS/FAS (Flexible Manufacturing/Assembly System) tool management system incorporates a vast amount of analysis and system development work. It must be done by a team of process and industrial, manufacturing systems and design engineers as well as data processing staff, headed by an experienced team leader who understands not only the data management problems but also the operation control and process planning aspects of FMS/FAS systems, within the CIM environment.

When designing the FMS/FAS tool management system one should consider the following steps:

1. Collect all current and possible future user and system requirements.

2. Analyze the system (i.e. the data processing and the FMS/ FAS hardware and software constraints).

3. Design an appropriate data structure and data base for describing tools (and/ or robot hands, probes, sensory-based inspection and assembly tools, etc.).

4. Specify and design programs, and query routines and dialogues that are capable of accessing this data base as well as communicating with the real-time production planning and control system of the FMS/ FAS.

Probably the most important question to be answered before starting to design an FMS tool management system and a data base is "Who" is going to use the data, "When" and "For what" purposes in the particular system ?

Tooling data in FMS, are typically going to be used by several sub-systems as well as by human beings. These are as follows:

* The production planning sub-system.

* Process control.

* Part programming.

* Tool preset and tool maintenance.

* Tool assembly (manual or robotized).

* Stock control and material storage.

For example, the production planning system has to be informed in real-time about the availability of tools in stock, as well as about the current contents of the tool magazines of the machine tools (in the case of Flexible Assembly Systems the robot hands in the End-of-Arm-Tool magazines) otherwise it will not be able to generate a proper production schedule.

It must be noted that the real-time aspect is important because tools are changed in the magazines of machines, (or cells) not only because they wear, but also because different part programs may need different sets of tools. (The actual tool changing operation is done in most cases by manipulators or by robots. The tool magazine loading/ unloading procedure is performed mostly by human operators, sometimes by robots or special purpose mechanisms, such as a tool shuttle).

Both the process control and the production planning systems have to update any changes and act in real-time, otherwise the operation of the system can be disrupted.

From the FMS/FAS tooling and tool management points of view one must emphasize the links between the CAD system, in which the parts are designed (using design for manufacturing principles), and the CAM system, where the FMS part programs are written. Typically, an FMS part programmer analyses the CAD output (i.e. the design drawings of the component to be manufactured on the FMS), the fixturing, the different setup (i.e. work mounting) tasks, as well as the necessary operations, their alternatives, the required tools and finally a precedence list of the resources (i.e. the possible candidates of processing stations, or cells, or machines). (Note that the actual selection of these candidates should be done real-time by the dynamic scheduler).

The real-time data bases and software systems are also important, since they provide the reports and status information that are needed for the smooth operation of the FMS. (In particular, its dynamic scheduler and other subsystems such as maintenance should be emphasized here.) The tooling system related tasks are divided into two levels, the first being TOOL PREPARATION, which is typically an off-line operated service/support activity (and area, or station), and the second being REAL-TIME TOOLING activities (usually situated physically inside the real-time controlled FMS).

We now discuss both of them in more detail.

As explained earlier, the output of the CAM system is a production rule base. This is the knowledge the FMS needs to produce each part. In this production rule base, amongst others, tools are assigned to each operation. The tool codes are selected by the FMS process planner, or automatically assigned by a process planning system, and are obtained from the tool data base. On the basis of the requested tools a list is sent via the network to the Tool Preparation Facility, or Station, where the actual tools are prepared (i.e. assembled and preset) and stored in an appropriate way, such that the material handling system of the FMS - typically AGVs, or a special purpose tool shuttle can pick them up.

The tool preparation station also deals with other activities, among which the most important are as follows:

* Tool service and maintenance.

* Tool assembly to orders (as it is necessary to replace worn tools).

* Tool preset, tool inspection and adjustment

* Real-time tool pickup and tool transportation organized to serve the needs of the real-time FMS.

The tool preparation station receives its orders, initially originated by the CAD data processing system via the FMS network and technically specified by the CAM system in the form of a production rule base. Order data arriving at the tool preparation station includes:

* Part orders (consisting of part codes and quantities). Note that this is a very important data set for the real-time FMSdynamic scheduler too.

* Notification of when the parts are physically available for FMS processing, representing a due date for tool preparation.

* A priority order (note that this can change because of some real-time changes in the system, thus this station must be able to cope with this task too).

* The portion of the production rule base, describing the requirements regarding tool preparation.

The tool preparation station keeps in touch with the real-time FMS system, as well as with the rest of the system, by feeding back important tooling system related data relating to:

* Stock reviews (regarding tools).

* FMS status report (regarding tools).

* Part priority status reports (in case dynamic changes must be performed in the FMS which have an effect on tooling needs and tool preparation due dates).

The Real-time part of the FMS tool management system must deal with the following tasks:

* It must handle the application of tools for a variety of processes as defined in the production rule base and assigned real-time to the FMS/FAS resources by the dynamic scheduler.

* It must provide data to control the transportation of tools and tool magazines within the FMS.

* It must provide information to perform and supervise tool changes and tool magazine changes at all levels.

* It must be notified of tool inspection results (e.g. in the case that it finds a worn out tool as a result of an inspection procedure it must generate a command that would instruct the tool magazine update system to change the tool in question in the appropriate tool magazine).

* It must provide information in the case of emergency, e.g. if a tool breaks.

* It must provide the necessary interfaces and data to perform diagnostic/recovery operations, preferably using diagnostic expert systems.

Finally, let us underline an important feedback loop starting at the real-time system, and ending at the tool preparation station, which contains the real-time tool status, wear and part priority information. This data is often useful to those people and/or system software systems that deal with the generation of the production rule base. It is also a very useful data set for FMS designers, since a lot of data which would previously have been lost is going to be saved in this way.

The tool preset station in the Tool Preparation unit must be able to inform the process control system about tool preset and offset data, preferably via a digital tool preset unit linked directly to the data processing network of the FMS. Note that the final adjustments, i.e. the x, (y) z tool length correction, have to be performed at the tool preset station or at the machine, if the machine is capable of performing this operation. Note that length and diameter values are often set at the cell prior to the first operation using that tool, as well as between operations using the real-time operated tool at the machine, in order to check tool adjustments and tool wear.([33])

Regarding the rest of the tooling data it must be mentioned that when writing FMS part programs, one must know the actual sizes of tools and their characteristics and behavior in different conditions. Tool geometry data is also used when checking for tool collision by graphics simulation in the CAD/CAM environment.

Conclusions and Summary

Go to top

We are moving into the "Knowledge Age" when competitive advantage cannot be maintained without continued and sustained commitment to innovation, research and continuous creation and application of new knowledge.

In terms of corporate investment, the financial stake involved in turning good quality, flexible, Open Access Education and research into competitive advantage is enormous. As many examples show in Europe, United States, Japan, Asia and elsewhere, entire nations' future and wealth (both intellectual and material) depend on the quality of education provided to learners. This means that the wrong education, research and development strategy for product design, manufacture and management can easily spell disaster.

To summarize, one should realize that FMS scheduling cannot be studied, analyzed and designed in isolation from the distributed, lean and flexible CIM environment, because it simply would not work in practice. In other words, it must be part of an overall production control and management strategy based on fundamentally JIT oriented, "pull" versus "push type" production control methods, capable of considering all affected resources (i.e. machines, tools, fixtures, human operators, etc.) and changing dynamically according to real-time requests.

Acknowledgments

Go to top

I hereby would like to express my thanks to NJIT, New Jersey Institute of Technology, The University of East London (UK), the Ford Motor Company, Hitachi Seiki, FESTO, Denford Machine Tools, Rolls-Royce Motor Cars, Lucent Technologies, Aircruisers, NSF, DARPA, General Motors, Hughes, The US National Guard, The New Jersey Technology Council and the United States Department of Commerce for their continuous support of my research, industrial and educational projects.

References and Bibliography

Go to top

[1] Ranky P.G., The Design and Operation of FMS (IFS (Publications) Ltd. and North Holland), p. 348, 1982.

[2] Ranky P.G., Computer Integrated Manufacturing (Prentice-Hall International) p. 528, 1985.

[3] Ranky P.G., A Program Prospectus for the Simulation, Design and Implementation of Flexible Assembly and Inspection Cells, International FMS Conference at The University of Michigan, Ann Arbor, 1986.

[4] Ranky P.G. and P H Francis: Design and Solid Model Simulation of Generic Assembly Cells and Systems, Japan-USA Symposium on Flexible Automation, 1986, Osaka, Japan.

[5] Ranky P.G., End of Arm Tool Management System for Robotized Assembly Lines, Robots West Conference, SME, Long Beach, CA, 1986.

[6] Ranky, P.G.: An Operation Control Strategy of FMS, International Conference on Productivity Research, ICPR'87, The University of Miami, 1987.

[7] Ranky P.G.: Real-Time Quality Control Feedback Loops in CIM Environment, Automated Inspection and Product Control Conference organised by IFS Conferences and IITRI, Chicago, 1987.

[8] Ranky, P. G: Libreria Software FMS (An FMS Software Library), PIXEL (Computer Graphics, CAD/CAM, image processing), Vol 5, No 1, 1984, pp 49-54 (in Italian).

[9] Ranky P.G.: FMS in CIM (Flexible Manufacturing Systems in Computer Integrated Manufacturing), ROBOTICA, Cambridge University Press, 1985, Vol 3, pp. 205-214

[10] S Y Nof, M M Barash and J J Solberg: Operational control of item flow in versatile manufacturing systems. International Journal of Production Research, 17 479-489., 1979

[11] J Pasquier: High level simulation of flexible card assembly lines. Research Report RC-10881, IBM Thomas J. Watson Research Center, Yorktown Heights, New York, USA.

[12] M L Pinedo et al: Sequencing in a flexible assembly line with blocking to minimize cycle time. Proceedings of the Second ORSA/TIMS Conference on Flexible Manufacturing Systems: Operations Research Models and Applications, (Amsterdam: Elsevier Science Publishers B.V.) pp. 499-508. 1984

[13] R V Rachamadugu et al: Real-time scheduling of an automated manufacturing center. Proceedings of the Conference on Real-Time Optimisation in Automated Manufacturing Facilities, National Bureau of Standards, Gaithersburg, Maryland, USA, pp. 293-316. 1986

[14] T J Schriber and K E Stecke: Machine utilizations and production rates achieved by using balanced aggregate FMS production ratios in a simulated setting. Proceedings of the Second ORSA/TIMS Conference on Flexible Manufacturing Systems: Operations Research Models and Applications, (Amsterdam: Elsevier Science Publishers B.V.) pp. 405-416. 1986

[15] An AI Approach to Schedule Generation in a Flexible Manufacturing System, Sheldon Shen and Yih-Long Chang, The Ohio State University, USA. Proceedings of the Second ORSA/TIMS Conference on Flexible Manufacturing Systems: Operations Research Models and Applications, edited by K.E. Stecke and R. Suri, Elsevier Science Publishers B.V., Amsterdam, 1986 - Printed in The Netherlands.

[16] Y Kuratani et al: A pre-computational report on job shop simulation research. J. Oper. Res. Soc. Japan 2, 145-183 (1960).

[17] E LeGrande: The development of a factory simulation system using actual operating data. Management Tech. III(1), 1-19 (1963).

[18] A Ashour:Sequencing Theory. Lecture Notes in Economics and Mathematical Systems. Springer-Verlag, Berlin (1972).

[19] S Eilon et al: Job shop scheduling with due date. Int. J. Production Res. 6(1), 1-13 (1967).

[20] S E Elmaghraby: The machine sequencing problem-review and extensions. Naval Res. Logistics Quart. 15(2), 205-232 (1968).

[21] J Erschler et al: Periodic loading of flexible manufacturing systems. In: Advances in Production Management Systems (Elsevier Science Publishers IFIP) 1984

[22] Ranky P.G.: Dynamic Simulation of Flexible Manufacturing Systems (FMS), Applied Mechanics Reviews, Special FMS issue, ASME (The American Society of Manufacturing Engineers), 1986.

[23] Ranky P.G.:Tool management tasks in flexible manufacturing systems (FMS). NAMRC XIV (North American Manufacturing Research Conference), Minnesota, USA, 1986.

[24] Ranky P.G.: End of arm tool management system for robotized assembly lines. Robots West Conference, Long Beach, California, USA, 1986

[25] Suri, rajan, Desiraju, Ramakrishna: Performance analysis of Flexible Manufacturing Systems with a Single Discrete Material - Handling Device, The International Journal of Flexible manufacturing Systems 9 (1977) p. 223-249, 1997

[26] Caprihan, Rahul and Wadhwa, Sughash: The Impact of Routing Flexibility on the Performance of an FMS - A simulation study, The International Journal of Flexible manufacturing Systems 9 (1977) p. p. 273-298, 1997

[27] Ranky, P G: An Introduction to Total Quality and the ISO90001 Standard An Interactive Multimedia Presentation on CD-ROM with off-line Internet support by CIMware (IEE and IMechE Approved Professional Developer). Multimedia design & Programming by P G Ranky and M F Ranky, http://www.cimware@cimwareukandusa.com

[28] Ranky, P G: Computer Network Architectures for World Class CIM Systems, CIMware, ISBN 1-872631-01-0, 231 p. http://www.cimwareukandusa.com

[29] Ranky, P G: Manufacturing Database Management and Knowledge Based Expert Systems, CIMware Ltd., ISBN 1 872631-03-7, 240 p http://www.cimwareukandusa.com

[30] Ranky, P G: Flexible Manufacturing Cells and Systems in CIM, CIMware Ltd., ISBN 1-872631 02-9, 264 p. http://www.cimwareukandusa.com

[31] Ranky, P G: Concurrent /Simultaneous Engineering Methods, Tools and Case Studies, CIMware Ltd., ISBN1-972631-04-5, 264p. http://www.cimwareukandusa.com

[32] Ranky, P G: An Introduction to Concurrent/ Simultaneous Engineering, An Interactive Multimedia Presentation on CD-ROM with off-line Internet support (650 Mbytes), published by CIMware, Multimedia design & programming by P G Ranky and M F Ranky, http://www.cimware@cimwareukandusa.com

[33] Ranky, P G: An Introduction to Flexible Manufacturing, Automation and Assembly, An Interactive Multimedia Presentation on CD-ROM with off-line Internet support (650 Mbytes), published by CIMware. Multimedia design & programming by P G Ranky and M F Ranky, http://www.cimware@cimwareukandusa.com

[34] Anger, F D: On Lamport's Interprocessor Communication model, ACM Transactions on Programming Languages and Systems, 11(3): 404 - 417p.

[35] Lamport, L: Time. Clocks and the Ordering of Events in a Distributed System, Communications of ACM, 21(7) 558-565 p.

[36] Chandi, K and Mistra J.: Asynchronous Distributed Simulation via a Sequence of Parallel Computations, Communications of ACM, 24(4) 198-406 p.

[37] Tanik M M and Chan E S: Fundamentals of Computing for Software Engineers, Ginny's, 1995, p. 251.

Go to top