Contribution to Virtual Manufacturing Background Research, Phase II


By

Edward Lin, Ioannis Minis, Dana S. Nau and William C. Regli
Institute for Systems Research
University of Maryland
College Park, MD 20742

For

Lawrence Associates Inc.
5100 Springfield Pike, Suite 509
Dayton, Ohio 45431

Under Contract Number F33615-92-D-5812

Delivery Order Final Report for Period August 1995 through December 1995

Prepared for: Manufacturing Technology Directorate
Air Force Wright Laboratory
Air Force Systems Command
Wright-Patterson Air Force Base, Ohio 45433-6533

Executive Summary

This document presents the results of work performed by researchers at the University of Maryland as part of a subcontract to Lawrence Associates in support of the ManTech contract Manufacturing Technology Special Advanced Studies. This is a follow-up to our previous report entitled "Contribution to Virtual Manufacturing Background Research," which assessed research and applications relevant to key aspects of Virtual Manufacturing (VM) and presented an outlook for the future of VM technologies. The current report identifies areas in which VM may have a significant impact beyond the state of the art, and discusses prospects for the use of VM tools over the Internet.

Promising areas for VM. Our review of certain aspects of VM has revealed several areas in which VM may have a significant impact beyond the current state of the art. Some of the most promising areas include manufacturability analysis, validation and evaluation of process plans, partnering in agile enterprises, process design, and optimization of production plans and schedules. As specific examples, we discuss the IMACS tool for manufacturability analysis and the OSPAM for partnering in agile enterprises.

Prospects for VM tools on the Internet. Our experience in developing, enhancing, and maintaining the VM web site at http://www.isr.umd.edu/Labs/CIM/virtual.html suggests several promising areas for the development of VM tools that operate over the Internet. In the near term, the most promising areas for further development of VM tools on the Internet would appear to be application areas satisfying the following criteria:

In order to achieve efficient operation over the Internet, each user's interactions with the remote server while using the tool should not be very frequent, and should not require a very high bandwidth.

In order for it to be cost-effective to make a VM tool over the Internet rather than selling copies of the tool to users to run it at their local sites, it should be a tool that is useful to a large user group, but is relatively difficult or costly to get copies for direct use at user sites in comparison with how often any single user might want to use it.

Based on these criteria, we describe four potentially promising areas of application for VM over the Internet: CAD data translation; distributed manufacturing; production system design; and manufacturability analysis.

To provide a concrete example of a VM tool operating on the Internet, we have developed a tool for translating the sat files used in the ACIS solid modeler into the internationally accepted STEP format, and have included it in our VM web site. To give the reader a feel for the future potential of VM on the Internet, we present a scenario describing the possible operation (over the Internet) of a tool that automatically analyzes designs to formulate suggestions for how to improve their manufacturability.

1. Introduction and Background

This document presents the results of work performed by researchers at the University of Maryland as part of a subcontract to Lawrence Associates in support of the ManTech contract Manufacturing Technology Special Advanced Studies. This is a follow-up to our previous report entitled "Contribution to Virtual Manufacturing Background Research," which assessed research and applications relevant to key aspects of Virtual Manufacturing (VM) and presented an outlook for the future of VM technologies. The current report identifies areas in which VM may have a significant impact beyond the state of the art, and discusses prospects for the use of VM tools over the Internet.

This report is organized as follows:

Section 1. Introduction and Background
Section 2. Objectives
Section 3. Description of Effort
Section 4. Recommendations
Section 5. Conclusions

Appendices:
References

2. Objectives


The objectives of the contract were as follows:

1. To maintain and enhance the VM web site. This included keeping the site publicly available on the Internet, publicizing it to generate interest in it, and soliciting further information for inclusion in it.

2. To investigate the possibility of including software components in the web site, including both "canned" demonstrations of VM software, and actual working VM tools to be run in our server by visitors to the web site.

3. To prepare a prospectus based on the results of the above study, outlining possible ways of using the Internet to provide access to tools for Virtual Manufacturing.

The next section describes the steps we took to achieve those objectives.

3. Description of Effort

We have taken the following steps to maintain and enhance the web site:

To facilitate maintenance of the web site, we have moved it from its previous temporary location at the web account of one of the co-authors of this report, to a permanent location at http://www.isr.umd.edu/Labs/CIM/virtual.html.

We have reformatted and reorganized the pages at the web site, to make them easier to use and understand.

We have incorporated into the web site the technical reports "Virtual Manufacturing User Workshop" and "Virtual Manufacturing Technical Workshop," compiled and edited by Lawrence Associates Inc.

We have incorporated a copy of the article "Automated Manufacturability Analysis: A Survey," by Gupta, Das, Regli, and Nau. This article provides a detailed survey of techniques for analyzing the manufacturability of product designs.

We have incorporated "canned" demonstrations of two manufacturing tools developed at the University of Maryland: the IMACS tool for manufacturability analysis of machined parts (see Section 3.1), and the OSPAM decision support tool for partnering in agile manufacturing (see Section 3.2).

We have made arrangements to incorporate information on the virtual manufacturing work being done by NIST's Manufacturing Systems Engineering group (see Section 3.3).

We have developed a "forms" interface to enable researchers on virtual manufacturing projects to submit information for inclusion in our web site. A portion of the submission form is illustrated in Figure 1 on the next page.

We have sent a solicitation to the previous contributors to the web site, inviting them to use the forms interface to update their contributions.

To publicize the web site, we have announced it on several mailing lists and electronic newsletters, including the Process Planning mailing list at NIST, the Design Computing newsletter, and the sci.engr.computing Usenet newsgroup. In our announcement, we invited interested parties to submit information for possible inclusion in the web site.

We have incorporated a working manufacturing tool that can be run in our server by remote visitors to the web site. This tool, an ACIS-to-STEP translator, is described below in Section 3.4.

Some of these steps are discussed further below. Specifically, IMACS and OSPAM are overviewed in Section 3.1 and 3.2. Section 3.3 outlines an interesting example of VM research. Finally, Section 3.4 describes the ACIS-to-STEP translator.

Figure 1: A portion of our web site's interface for submitting information on VM research projects.

3.1. Interactive Manufacturability Analysis and Critiquing System (IMACS)

IMACS is a computer system that analyzes the manufacturability of proposed designs for machined parts, in order to help designers produce designs that are easier to manufacture. Starting from a CAD model of the design (including the design geometry and tolerances), IMACS systematically generates and evaluates alternative operation plans for the proposed design. From this analysis, IMACS provides feedback to the user about how much time will be needed to machine the part, including a breakdown showing which design elements are especially problematic. If the design cannot be produced by machining, then IMACS tells the user which portions are unmachinable.

The techniques used in IMACS have the potential to reduce the need for redesign (resulting in reduced lead time and product cost), and help to speed up the evaluation of new product designs in order to decide how or whether to manufacture them. Such a capability will be especially useful in flexible manufacturing systems, which need to respond quickly to changing demands and opportunities in the marketplace. IMACS is described further in the VM web site and in Appendix A.

3.2. Optimal Selection of Partners in Agile Manufacturing (OSPAM)

The OSPAM project focuses on the design for distributed manufacture of electronic products, such as microwave modules and hybrid micro-assemblies. Specifically, we are developing methods for
critiquing the design of such products with respect to the production capabilities of potential partners,
alerting the designer to elements of the design that correspond to costly operations, and
selecting a good combination of partners that contribute to the manufacture of the design.

Thus, OSPAM is a design advisor that integrates partnering (including make or buy) decisions in the design phase, and facilitates design improvements with respect to partner capabilities. The project is described further in the VM Web site and in Appendix B.

3.3. NIST Effort on Virtual Manufacturing

An interesting example of ongoing VM work is the system being developed by the Manufacturing Systems Engineering group of the National Institute of Standards and Technology (NIST). The goal of this project is to define the information exchange needs for different VM tools in an integrated VM environment. To this end, the NIST researchers are integrating commercial software to support major steps of the product development process. Software integrated to-date include ProEngineers CAD, the PART generative process planner, and Denebs simulation tools. Using these environment, a product development team can design a product, derive its process plan and simulate its production on the shop floor. The production portion includes a detailed simulation of the 3-axes milling process, as well as simulation of the material flow within the shop. A detailed description of this project will be available shortly at the NIST web site http://elib.cme.nist.gov/msid/trojs/metk/homepage.htm which will be accessible from our VM site.

3.4. ACIS to STEP Translator

We developed a CAD model translator and included it in the VM Web site in order to demonstrate the feasibility of making generic VM tools available to designers and planers over the Internet. We chose this example for two reasons. First, CAD model translators from application-specific formats to a neutral one (such as STEP) are necessary for product data exchange between heterogeneous manufacturing applications. Secondly, it is possible with current technology to provide such services over the Internet, since the information exchange between sites involves simple file transfers.

ACIS is a powerful solid modeling kernel used in many CAD systems, such as AutoCad and Microstation. It uses typical entities for the boundary representation of a solid, including body, lump, shell, face, edge, loop, vertex, surface, direction and point . For data archival it stores this information in files of a special format (ACIS sat files). STEP is the international standard for product data exchange (ISO 10303), and defines generic as well as application specific information models. Examples of the former include the geometry and topology models [1], while examples of the latter include application protocols for printed circuit assemblies or NC process plans for machined parts.

The STEP entities for geometry and topology are similar to the ones used by ACIS for shape representation; however, certain basic differences exist in the contents and format of these entities as well as in the structure of the two information models. As a part of the project on Optimal Selection of Partners in Agile Manufacturing (OSPAM) [2] (see also 3.2 and Appendix B) we have studied these differences, determined ways to construct one information model given the data of the other, and have built ACIS to STEP and STEP to ACIS translators for boundary representations. In doing so, we have used the STEP standard to define the schema of an Object Oriented database (OODB), in which the STEP information models are stored. The ACIS to STEP translator accepts an ACIS sat file, performs the necessary information transformation and stores the results in the STEP OODB; the reverse is performed by the STEP to ACIS translator. Reference [3] describes in detail both translators.

For this VM project, the ACIS to STEP translator has been extended to provide STEP exchange files following the ISO 10303-21 standard [4]. In addition, the translator can be accessed from the VM Web site. This VM tool accepts an ACIS sat file that users have uploaded to the ftp site of the Institute for Systems Research and returns a STEP exchange file to that remote site as well as to the user via email. Figure 2 shows the functionality of the extended translator. A user from a remote site may call the translator by clicking the ACIS to STEP File Translator from the VM Web home page or by opening the URL
http://www.isr.umd.edu/Labs/CIM/vm/vmxlator.html

In response, the VM site provides a description of the translator and also provides a link to the form that can be used to invoke the translator. The remote user provides the requested information and presses the OK button. The translator routine reads the ACIS sat file, performs the necessary data transformations and populates the STEP OODB database. Subsequently, a file generator routine queries the OODB for the appropriate STEP data and builds a STEP file using the STEP exchange structure format. The resulting STEP file is sent back to users Web browser and mailed to the user. Appendix C provides a manual for the operation of the ACIS to STEP translator.



Figure 2: Functional representation of the ACIS to STEP translator.


It is emphasized that the scope of this translator is limited to shapes with planar, cylindrical and conical surfaces only. The correctness of the STEP output file has been tested and validated using NIST tools.

Our experience with the ACIS to STEP translator has shown that manufacturing utilities which require information exchange in a batch mode are ideally suited for placement on the Web. The operation of such software is not impeded by the bandwidth problems of the Internet, since it does not require real-time user interaction. Furthermore, in this case the data transfer consisted of the exchange of ASCII files, facilitating the interaction between heterogeneous operation systems and hardware.

4. Recommendations

In Section 4.1 below, we discuss areas in which VM may have a significant impact beyond the state of the art. In Section 4.2 we discuss prospects for the use of VM tools over the Internet. In Section 4.3 we present a scenario for the possible operation of a specific VM tool over the Internet: a tool that automatically analyzes designs to formulate suggestions for how to improve their manufacturability.

4.1. Examples of Promising VM Contributions

Our review of certain aspects of virtual manufacturing has revealed several areas in which VM may have a significant impact beyond the current state of the art. As mentioned in our phase 1 report [5], most virtual manufacturing applications require a robust information infrastructure that comprises rich information models for products, processes and production systems. Based on this infrastructure, decision support systems may be developed to predict or simulate key manufacturing activities in the computer. With this in mind, we discuss below a few examples that illustrate potential contributions of VM.

Example 1: Validation and evaluation of process plans for machined parts

Validation of process plans is typically performed by manufacturing a prototype, and therefore it is a long and costly procedure. On the other hand, plan-based manufacturability evaluation (for example, the IMACS system described in Section 3.1 and Appendix A) determines the difficulty of manufacturing a certain part in terms of the required manufacturing time and cost. In the case of machined parts, this is accomplished by estimating the manufacturing time of each operation in the process plan. This time estimate is based on handbook values of the machining parameters (speed, feed rate and depth of cut), the parts material, and the shape and volume of the corresponding machining feature.

Virtual manufacturing may play a significant role in two areas:

VM may validate the process plan in the computer, thus eliminating the need for costly prototypes.

Having a valid process plan, VM may estimate the manufacturing times and costs accurately. Furthermore VM may estimate other important measures, such as part quality.

VM-based process plan validation includes two levels of sophistication: First the tool path specified in the process plan may be verified through simulation. In this case a solid model of the work piece and a model of the machine tool are used to simulate the machining process. The cutting tool follows the prescribed path removing material from the work piece. Thus, infeasible tool trajectories (such as those interfering with portions of the machine tool, fixtures or the work piece) are easily detected and corrected. The second level of sophistication determines whether the cutting parameters specified in the plan are appropriate or even feasible. For example, high depth of cuts may lead to machine tool chatter and thus damage the work piece, the cutting tool or even the machine tool itself. In addition, high feed rates may lead to unacceptable surface roughness with respect to the designers specifications.

Research has already focused on tool path verification using VM. For example, investigators at the National Institute of Standards and Technology have used the PART generative process planner and Denebs simulation software to develop a tool verification program for three-axes milling [6] (see also Section 3.1). To the best of our knowledge, however, no VM work to-date has examined the appropriateness of machining parameters in a process plan of a certain design; this despite the proliferation of analytical models describing the physics of most machining processes. Such models may be used in conjunction with the solid models of the work piece and the machine tool to construct a virtual machining process. In virtual machining the analytical models will represent the physics of the cutting process, while the solid models will represent the process geometry. Thus, given a part design, the machine tool to be used and the parts process plan, the virtual machining process can be run to not only verify the tool path, but also to predict whether the selected machining parameters will lead to undesirable results (chatter, poor surface roughness, excessive dimensional variability, and excessive tool wear).

Beyond plan verification, virtual machining may be used to estimate accurately the merit of a process plan, and, based on this evaluation, determine appropriate process conditions to improve (and even optimize) the plan. With virtual machining, the fidelity of the machining time and cost estimates is expected to improve. In addition, modeling the process physics will allow to predict the quality of the machined part, which cannot be determined easily and reliably without producing several physical prototypes. This information is invaluable to both the designer and the process planner. The former may use it to modify the design in order to improve its manufacturability. The latter may use it to tune the machining parameters and improve, or even optimize, the plan. All this can be done early at the design/planning cycle without resorting to costly prototypes.

Example 2: Process design

The integration of physical and visual simulation models to construct virtual manufacturing processes may be extended beyond the mature area of machining. Typical candidates include net shape processes, such as casting, forging and injection molding. For example, being able to predict and simulate the transition from a billet to a finished product during forging, or the entire injection molding cycle from die fill, to part solidification to ejection, is invaluable to the part designer, the die designer, and the process planner.

As already mentioned in Example 1 above, models that describe the process physics are critical for the construction of dependable virtual processes. However, although process modeling is a mature subject, it has yet to play a significant role in concurrent engineering. One of the major reasons for this is the lack of unified ways to deal with process models. This appears to be a major roadblock for virtual manufacturing; especially in assessing the feasibility of producing a design with a certain set of manufacturing processes and evaluating the ease of manufacture with these processes. Thus, novel representations are necessary to capture the physical models of various processes (i.e. analytical, statistical and simulation-based models) and to provide unified interaction mechanisms with the virtual manufacturing environment. Having developed such representations, it is possible to construct virtual processes and use them to:
Determine the feasibility of a process and a given process plan to yield the desired product characteristics.
Evaluate process performance with respect to processing time, cost and quality of the manufactured product.
Support process design, i.e. the tuning of critical process parameters to optimize process performance.

Example 3: Optimization of production plans and schedules

Discrete event simulation has traditionally been considered a powerful tool for production applications. Numerous special languages and software packages have been developed to simulate the production activities of a manufacturing shop. However, applications of simulation have been limited to validating designs of production systems and to comparing production strategies. There has not been extensive industrial application in manufacturing operations, including production planning and scheduling. There are two main reasons for this:

Simulation is suitable for production lines, in which all products follow similar sequences of operations. It does not deal effectively with job shops, which produce small batches of numerous product types with diverse production routings. In these cases the effort to build a simulation model is almost prohibitive due to the large volume of the necessary production data.
Simulation is suitable for push-type systems, in which the shop orders are launched upon the receipt of the customer orders. Most simulation packages are ineffective for just-in-time production, which is increasingly popular in current practice.

However, if these problems are addressed effectively, simulation (and virtual manufacturing in general) has the potential to yield major contributions to both operations management, and product and process design. In operations management, simulation may support production planning and scheduling, especially when combined with powerful optimization tools. For example, simulation combined with perturbation analysis may be used to optimize batch sizes for minimizing inventory and set-up costs. In this case, perturbation analysis will be used to evaluate the necessary cost gradients efficiently from a small number of simulation runs. Similar approaches may be used to determine optimal threshold values for the Work -in-Process at bottleneck workstations, beyond which no work orders are released to these bottlenecks. As another example, simulation may be combined with simulated annealing or genetic algorithms to optimize the schedule of bottleneck workstations.

In product/process design simulation may play a significant role in integrating product design and process planning with production planning and scheduling. Given a design and a process plan (or a set of alternative plans), various production scenaria may be simulated to determine the impact of the new product on the operations of the shop. By doing so, the designer will be able to determine early in the design stage the effect of certain design decisions in production planning and scheduling. Furthermore, the process planner will be able to determine alternative plans that are appropriate for certain states of the shop. It should be emphasized, however, that product design and process planning are time-independent (static) activities, while production planning and scheduling are dynamic activities. This complicates their integration and presents a challenging topic for further research.

In order for virtual manufacturing to have major impact in these areas, discrete event simulation should be integrated to the existing information systems of a manufacturing company. For example, production simulations should be built with minimal user input using typical MRP II data, such as Bills-of-Materials, Workcenters, Production Routings, Customer Demand and Shop Floor Control data. In addition, both the CAD and the process planning systems should be integrated in this practical VM environment in order to enable the use of simulation for trade-offs in design and planning.

4.2. Virtual Manufacturing over the Internet

4.2.1. Background

The Internet has recently become a major information resource provider for industry, and its demand keeps growing. The World Wide Web (WWW) plays an especially important role in providing information service on the Internet. Most current WWW applications are processed in batch mode and thus do not require intensive user interactions. Application areas include:

Product and service suppliers information: examples are Thomas Register of American Manufactures for manufactured products [7] and Internet Shopping Network for computer goods, home and office supplies [8];

Quotation and bidding: Examples include a product information and on-line quotation system from Power Computing [9], and a real-time auction tool from Onsale [10];

Tracking System, for example, UPS's package tracking system/service information [11];

Software Product Demos, for example, TAE Plus demonstration Program (a GUI development tool) [12] and the NIST Tool Kit (Data Probe and STEP class library release) [13];

Education (training and tutorial, e.g., computer-based learning systems which aid students with their understanding of the complex phenomena underlying engineering domains from the INTERACT project [14].

Not all of the services provided on the Internet are free. Several billing mechanisms are being deployed: membership, pay by charge card, and pay by e-cash. In the first case, each member of the service will be given an account number (users name) and a password in order to use the service. In the second case, users may supply their charge card number in an electronic order form on the Web to order product or services. Finally, E-cash is an electronic currency which is being developed by the Amsterdam-based company Digicash and will be offered by Mark Twain Bancshares [15]. In order to use pay-by-e-cash method, both buyers and sellers must have accounts in Mark Twain Bancshares. A buyer electronically withdraws the so-called coins, specially encoded symbol strings, from his/her account and transmits them to the seller to buy products or services.

4.2.2. Transmitting VM Information over the Internet

Several languages are being developed specifically for data exchange over the Internet. The best known example is the HTML language used for constructing the documents read by web browsers such as Mosaic or Netscape. Documents written in HTML include embedded commands that change the formatting of the text or specify remote locations from which further information can or should be retrieved.

HTML documents may be useful for sending certain kinds of Virtual Manufacturing information over the Internet, by transmitting design and manufacturing data to a web browser from a program running at a remote location. Examples include STEP files, CAD models, and text data such as the results of a design or manufacturing analysis or the information found in a manufacturer's catalog. However, the utility of this approach may be limited by the available bandwidth. For example, if one is trying to view a CAD image in a web browser, it may take several minutes to transmit and interpret the image. This makes it impossible to do interactive operations (such as rotating the CAD model) in real time.

For certain kinds of VM data, it may be possible to alleviate the bandwidth problem by transmitting the data not as bitmapped images, but instead using the Virtual Reality Modeling Language (VRML). VRML is a language for describing multi-participant interactive 3D simulations. Its initial intended application is the creation of virtual worlds network via the Internet and hyperlinked over the World Wide Web. All aspects of virtual world display and interaction can be defined using VRML. It is the intention of SGI and the other designers of VRML that it become the standard language for interactive 3D simulation within the World Wide Web.

Another way to address the bandwidth problem is by transmitting computer code specifying programs to be executed at the local site. In principle, this could be any code that could run on the local machinebut in practice, it is more likely to be code written in a language specifically developed for that purpose, such as HotJava or Telescript. Programming languages and environments have been developed specifically designed for the exchange of executable software components (often in an interpreted language) over a network. On the receiving end of these exchanges, a viewer or interpreter executes the program and enables it to interact with the user. In spite of the enormous security issues at stake, languages such as Sun Microsystems' Java (see http://java.sun.com) and General Magic's Telescript (see http://www.genmagic.com/Telescript/index.html) provide generic tools to extend the reach of software over a computer network. With these tools, one can compose interactive transmittable programs that contain documents in a variety of media, or control for physical devices (e.g. machine tools).

HotJava is a World-Wide Web browser built using Sun's Java, an object-oriented programming language. The essential difference between the HotJava and other browsers is that it allows users to dynamically interact with the browser without overloading the network traffic. In order to do this, HotJava adds a new type of HTML tag: APP, for "applet." An applet is a program written in the Java language to be run within HotJava. Developers use Java to write applets, compile them, and install them on the server. HotJava can dynamically link the Java code from the host and execute it on the local machine. Currently, the HotJava browser is still in the alpha release. It is available on Solaris (2.3 or higher) and Windows NT/95. More information on Java and HotJava can be found at http://www.sun.com and in the Usenet newsgroups comp.lang.java and alt.www.hotjava.

Intriguingly related to these tools are many current application development environments in commercial software systems. These tools are already being used by applications builders to construct modular ``add on'' software packages to be coupled with other tools (by analogy, one might thing of these tools as application-specific forms of Java). In the computer-aided design market examples include Pro/Engineer's Pro/Develop, EDS/Unigraphics' GRIP and the MicroStation Development Language (MDL) from Bentley Systems. Programming environments such as these can be used to create network-savvy software modules to be shared, executed, or purchased over the network from within a single CAD environment.

4.2.3. Example Application Areas for VM Development on the Internet

In the near term, the most promising potential areas for development of VM tools on the Internet would appear to be application areas satisfying the following criteria:

In order to achieve efficient operation over the Internet, each user's interactions with the remote server while using the tool should not be very frequent, and should not require a very high bandwidth.

In order for it to be cost-effective to make a VM tool over the Internet rather than selling copies of the tool to users to run it at their local sites, it should be a tool that is useful to a large user group, but is relatively difficult or costly to get copies for direct use at user sites in comparison with how often any single user might want to use it.

With these criteria in mind, we now provide three brief examples of potentially promising areas of application for VM over the Internet: CAD data translation, distributed manufacturing, and production system design. Following these examples, Section 4.3 provides a detailed scenario for the use of VM techniques over the Internet in a specific application area.

Example 1: CAD data translation

The ACIS to STEP translator developed as a part of this project (see section 3.4) is a clear example of the simplest type of VM applications that can be used over the Internet. This application meets several of the criteria mentioned above:

it requires simple file transfers between the client and server sites;
it is computationally intensive at the server site only;
it is very useful to a large user group.

We envision that such services may be provided by Internet brokers. The service described in Section 3.4 is highly applicable to users who occasionally exchange product files in the STEP format. However, users that frequently use STEP files for data exchange will tend to purchase one or more of the commercial STEP translators and use them at their site. It is noted that the development of such translators is a straightforward task of limited research interest.

Example 2: Manufacturing resource models for distributed manufacturing

A highly distributed VM application is the evaluation of a product design with respect to the production capabilities of potential manufacturers, and the subsequent selection of the most appropriate partners for a virtual enterprise to manufacture the candidate product. We have been working on this problem for over two years and we are currently completing the OSPAM software system which provides decision support for this application (see Section 3.2 and Appendix B).

Although most of the design critiquing and partner selection tasks may be performed at a single site, it is clear that the data describing the capabilities and historical performance of a firm should be owned and maintained by that firm in a manufacturing resource model that is accessible over the Internet. Tools such as OSPAM should be able to access and query these information models in order to compare the design requirements against the firms capabilities. This suggests that the manufacturing resource data are not included in simple HTML documents, but rather they are stored in databases that can be queried by remote applications. Mediator or agent technology may be appropriate to enable the integration of these heterogeneous databases and the applications that use them.


Figure 3: Manufacturing facility design process and information.

Example 3: Design of production systems

The design of production systems is a mature subject, certain aspects of which have been thoroughly researched [16]. Figure 3 (taken from [17]) shows the major steps of the manufacturing facility design process and the data needed by this process. Numerous methods have been developed for certain steps of Figure 3, such as cell formation and resource layout. Other design steps have not been adequately addressed, including the design of the material handling network, and the integration of the entire shop design process. Virtual manufacturing tools may be used to evaluate and validate the systems designed by these methods. Beyond discrete event simulation (which can determine important operational system measures such cumulative material handling time and distance), animation and virtual reality may be used to illustrate the entire shop in operation as experienced by the shop staff.

Several software packages have been developed for production system design, including PDS of the University of Maryland [18,19] which addresses many of the steps of Figure 3. Although such packages have a wide applicability, manufacturing shop design or redesign is not a frequent activity and, thus, the software are not used on a regular basis by most manufacturers. Thus, this application is an appropriate candidate for use over the Internet. It is also noted that most of the information exchange between the client and server sites is limited to file exchanges as Figure 3 implies. For example, the inputs of the design software include typical MRP II data, such as work centers, production routings and forecasted product demand, all of which can be represented as text. Much of the output may also be described in a textual form. It is noted, however, that the exchange of evaluation and validation data will be probably limited to the values of performance indicators rather than animation or virtual reality images.

4.3. Scenario: Automated Redesign for Improving Manufacturability

4.3.1. Motivation

Redesign and re-planning of machined components may be needed if the production conditions are different from those that were anticipated during the component design phase. This can occur, for example, in the production of spare parts for aging equipment. In this case, the production quantities are much lower than when the original equipment was built, and the manufacturer of the spare parts may have different capabilities (e.g., due to new technology) than the manufacturer of the original equipment.

If the production conditions are different from those anticipated during the design phase, then the original design and the original process plan may be inappropriate for several reasons:

When the original equipment was built, it may have been cost-effective to produce special-purpose cutting tools and fixturing devices (and sometimes even machine tools!) in order to produce the equipment in high production quantities. For the production of spare parts in small quantities, such special-purpose hardware will no longer be cost-effective, and thus changes may be needed in both the design and the process plan in order to accommodate standard cutting tools, fixturing, and machine tools.

When new technology (e.g., multi-axis CNC machine tools) becomes available in an existing manufacturing environment this may make it possible to produce products at reduced production cost and lead time, and improved quality. However, in the case of products for which designs and process plans already exist, modifications may be needed to the designs and process plans to take advantage of the new technology.

We believe it is possible to develop a VM tool operating across the Internet, that systematically generates redesign suggestions to improve the manufacturability of product designs while still satisfying the designs' functional requirements. Below, we develop a scenario describing how such a tool might work for products that are manufactured using a variety of discrete manufacturing operations, including assembly, metal-cutting, joining, and forming. Such operations are used in manufacturing a wide variety of products, including sheet metal parts, machined parts, and electro-mechanical assemblies.

4.3.2. Previous Approaches

Redesigning a product usually consists of two steps: (1) identifying ``redesign clues'' (information about what attributes of the design need improvement and why), and modifying these design attributes in order to synthesize an improved design. Existing approaches to this task can be classified as direct and indirect approaches, as described below.

In direct systems [20,21,22] rules are used to identify infeasible design attributes from direct inspection of the design description. These infeasible design attributes are then modified using predefined rules to create improved designs. Due to interactions among machining operations, it can be very difficult to determine the manufacturability of a design directly from the design description---and thus the applicability of direct systems is rather limited.

Indirect systems [23,24,25] proceed by generating a detailed manufacturing plan, and modifying various portions of the plan in order to reduce its cost. Once this has been done designs that correspond to these modified plans are presented to the user as possible redesigns. Although these systems have wider applicability than direct systems, they have several limitations:

There may be many possible alternative plans for manufacturing the product, and it is not clear which of these plans to use as a basis for generating redesign suggestions. Selecting the most promising plan for the initial design may not necessarily produce the best redesign suggestions.

If the initial design is not manufacturable, then there will be no plan for the design, and thus no clear way to generate redesign suggestions.

Since most existing indirect systems do not take into account the design's functionality, this makes it difficult to ensure that the proposed changes will not violate functionality requirements.

4.3.3. Our Suggested Approach

We believe the limitations of the direct and indirect approaches to product redesign can be significantly reduced by developing an approach that incorporates some of the best aspects of both of them. Like the direct approaches, it will generate local modifications that satisfy design constraints arising from functional requirements of the design---but like the indirect approaches, it will assess the merit of these modifications by generating and evaluating the manufacturing plans that will be needed to produce them. The basic steps of the approach we envision are as follows:

(a) an example part P1; (b) design constraints for P1; (c) a modified version of P1.

Figure 4: A machined part P1 , some of its design constraints in an assembly, and a modified version of P1 that can be machined in fewer setups.


1. Obtain the initial design, as a file submitted via the Internet. This design will consist of a CAD model in a standard format(e.g., a STEP file or an ACIS sat file), plus a set of design constraints arising from the intended functionality of the design. To attempt a complete approach for representing and reasoning about design functionality is a very complex taskbut we do not believe it will be necessary to represent the functional requirements in a detailed manner. Instead, we note that the functional requirements give rise to various constraints on what kinds of modifications to the product design might be permissible---and the designer can be asked to attach such constraints as annotations to the CAD model, in a simple constraint language. For example, Figure 4 shows some of the constraints that a designer might specify on a part that is intended to mate with other parts in an assembly.


Figure 5: Some of the machining operations for the part P1. Often there is more than one machining operation capable of creating the same portion of the design.

2. Preprocess the initial design . Generate a set of manufacturing features F for the design (see Figure 5). Our work on the automated extraction of machining features from CAD designs [26,27] suggests one possible approach for doing this. Along with this feature set, identify relationships among the features that will dictate precedence constraints among the operations. Each feature represents some portion of the design that may be created using a single manufacturing operation. Since the features may overlap, the feature set F may include a number of alternative ways to manufacture various portions of the design; and thus an operation plan for the design will correspond to a subset of the features in F .

3. Analyze manufacturability of the initial design . Generate alternative operation plans [28,29] for manufacturing the design, as alternative subsets of the feature set F . Evaluate them by considering tradeoffs among criteria such as cost, time, and quality, to find the plan that best satisfies whatever combination of criteria the user specifies. This plan will establish a baseline for evaluating possible design modifications below---and if this plan is satisfactory to the user, then the user may elect to stop here without requesting redesign suggestions at all.


Figure 6: Generating a side-milling feature as an alternative to an end-milling feature.

4. Generate local modifications . Use both direct and indirect techniques to generate alternatives to various features in the feature set F . One possible way to do this is by means of feature modification operators [30]. For example, Figure 6 shows a modification that produces a slot-milling feature as an alternative to an end-milling feature. Identify and discard any alternatives that violate the constraints specified by the designer in Step 1. For example, if there were a constraint saying that the rounded ends of the end-milling feature in Figure 6 were important, then the slot-milling feature would be discarded. Augment the feature set to include the new features.

5. Synthesize manufacturing descriptions . From the augmented feature set F , generate combinations of the features that are sufficient to constitute complete designs. Each such combination of features represents a possible alternative modified version of the original design. Identify and discard any alternative designs that violate the constraints specified by the designer in Step 1. For the others, generate and evaluate manufacturing plans using the approach described in Step 3. Use the results of this analysis to rank the alternatives using the same optimization criteria as before.

6. Provide feedback. The best designs found represent possible ways to modify the design to improve its manufacturability. Send CAD models of these designs over the Internet to the designer, as alternative suggestions for redesign.

4.3.4. Anticipated Impact

The approach described above will be applicable to products that are manufactured using several manufacturing operations, including assembly, metal-cutting, joining, and forming. As shown in Table 1, such operations are used to manufacture a wide spectrum of products, including sheet metal parts, machined parts, and electro-mechanical assemblies.

Table 1: Discrete manufacturing operations for various types of products.

Assembly Cutting Joining Forming
Sheet-metal parts
Machined parts
Electro-mechanical assemblies

Such a system, if implemented, would directly support the goals of agile manufacturing, i.e. the ability to respond rapidly to the market demand for cost-effective, customized products of high quality. Specifically we envision the following benefits:

The system will provide the means to speed up the redesign process. This is especially important in the current defense and commercial business environment, in which the designer is forced to design a large number of custom variations of a product, each of which is produced in low quantity.

The system would ensure DFM. It will provide such suggestions automatically, and will synthesize design variations with improved cost, lead time, and quality. This will help designers determine what changes to a design will best improve its manufacturability.

The functionality of the system is directly applicable to the redesign of highly manufacturable spare parts for out-of-production equipment. For such parts the design changes will take advantage of new manufacturing technology introduced since the original equipment was built; in addition, the changes will adapt the design of a spare part for low quantity, batch production.

The design support provided by the system will reduce the need for costly changes after the product has been launched to production. Preventing such changes will also eliminate the associated production delays.

The suggestions provided by the system during redesign will illustrate to the designer configurations that are favorable for manufacturability. Thus, the automated redesign system will also act as a DFM education tool for the designer.

5. Conclusions

Our review of certain aspects of virtual manufacturing has revealed several areas in which VM may have a significant impact beyond the current state of the art. As examples of these, we have discussed the IMACS tool for manufacturability analysis, and the OSPAM for partnering in agile enterprises. We have also described some other promising areas in which VM tools can potentially be developed:
validation and evaluation of process plans for machined parts;
process design;
optimization of production plans and schedules.

Furthermore, our experience in developing and maintaining the VM web site suggests that there are several promising areas for the development of VM tools that operate over the Internet. In the near term, the most promising potential areas for further development of VM tools on the Internet would appear to be application areas satisfying the following criteria:

In order to achieve efficient operation over the Internet, each user's interactions with the remote server should not be very frequent, and should not require a very high bandwidth.

In order for it to be cost-effective to make a VM tool over the Internet rather than selling copies of the tool to users to run it at their local sites, it should be a tool that is useful to a large user group, but is relatively difficult or costly to get copies for direct use at user sites in comparison with how often any single user might want to use it.

To provide a concrete example of such a VM tool, we have developed a tool for translating the sat files used in the ACIS solid modeler into the internationally accepted STEP format, and have included it in the VM web site. Furthermore, in our report we have described four other potentially promising areas of application for VM over the Internet:
CAD data translation;
distributed manufacturing;
production system design;
manufacturability analysis.
For the fourth area, we have presented a detailed scenario, describing the possible operation (over the Internet) of a tool that automatically analyzes designs to formulate suggestions for how to improve their manufacturability.

Appendix A: IMACS, A System for Computer-Aided Manufacturability Analysis

Introduction

The ability to quickly introduce new quality products is a decisive factor in capturing market share. Because of pressing demands to reduce lead time, analyzing the manufacturability of the proposed design has become an important step in the design stage. In a typical CAD environment, the designer creates a design using solid-modeling software, and uses analysis software to examine different aspects of the proposed design's functionality. As shown below, the IMACS project is extending the design loop to incorporate a manufacturability analysis system that can be used once the geometry and/or tolerances have been specified. This will help in creating designs that not only satisfy the functional requirements but are also easy to manufacture.

We assume that the proposed design is available as a solid model, along with the tolerance and surface finish information as attributes of various faces of the solid model. We assume we have information about the available machining operations, including the process capabilities, dimensional constraints, etc. As shown on the next page, our approach is to generate alternative interpretations of the part as collections of machining features, map these interpretations into operation plans, and evaluate the manufacturability of each operation plans. The ultimate goal of the IMACS project is to provide tools for manufacturability analysis as part of the CAD systems used by designers. We believe our work will help designers design products that are easier to manufacture. This will reduce the need for redesign, resulting in reduced lead time and product cost. In addition, it will help to speed up the evaluation of new product designs in order to decide how or whether to manufacture them. Such a capability will be especially useful in flexible manufacturing systems, which need to respond quickly to changing demands and opportunities in the marketplace.

Manufacturability Analysis

Given a computerized representation of the design (i.e. a solid model) and a set of manufacturing resources, the automated manufacturability analysis problem can be defined as follows:

Determine whether or not the design attributes (e.g., shape, dimensions, tolerances, surface finishes) can be achieved.

If the design is found to be manufacturable, determine a manufacturability rating, to reflect the ease (or difficulty) with which the design can be manufactured.

If the design is not manufacturable, then identify the design attributes that pose manufacturability problems.

In general, a design's manufacturability is a measure of the effort required to manufacture the part according to the design specifications. Our approach to measuring manufacturability is to estimate the manufacturing time and cost. Since all manufacturing operations have measurable time and cost, these can be used as an underlying basis to form a suitable manufacturability rating. Ratings based on time and cost can easily be combined into a overall rating. Moreover, they present a realistic view of the difficulty in manufacturing a proposed design and can be used to aid management in making make-or-buy decisions.

Modeling Machining Operations with Features

In a machining operation, a cutting tool is swept along a trajectory, and material is removed by the motion of the tool relative to the current workpiece. The volume resulting from a machining operation is called a machining feature. A machining feature corresponds to a single machining operation made on one machine setup. Each machining feature has a single approach direction (or orientation) for the tool. In IMACS, features are parameterized solids that correspond to various types of machining operations on a 3-axis machining center, including the ones shown below [31]:
side-milling feature: 
face-milling feature: 
end-milling feature: 
drilling feature: 

Approach

One of the fundamental objectives of IMACS was to develop a methodology for systematically generating and evaluating alternative operation plans for machined parts. This involves representing the design as a collection of machining features such as those shown above. To get these features from the CAD model, IMACS uses the feature recognition subsystem [32] described at
http://www.cs.umd.edu/projects/cim/feature_rec.html
Given this feature-based representation of the design, there may be, in general, several alternative representations of the design as different collections of machinable features, corresponding to different ways to machine the part. As shown in the figure on the previous page, the basic idea is to generate alternative interpretations of the part as collections of machinable features, map these interpretations into operation plans, and evaluate the manufacturability of each operation plan. More specifically, our approach involves the following steps [33,34]:

1. Build the set of all potential machining features by identifying various features which can be used to create the part from the stock. Each of these features represents a different possible machining operation which can be used to create various surfaces of the part.

2. Repeat following steps until every promising feature-based model (FBM) has been examined:

A. Generate a promising FBM from the feature set. An FBM is a set of machining features that contains no redundant features and is sufficient to create the part. We consider an FBM unpromising if it is not expected to result in any operation plans better than the ones which has already been examined.

B. Do the following steps repeatedly, until every promising operation plan resulting from the particular FBM has been examined:

i. Generate a promising operation plan for the FBM. This operation plan represents a partially ordered set of machining operations. We consider an operation plan to be unpromising if it violates any common machining practices.

ii. Estimate the achievable machining accuracy of the operation plan. If the operation plan cannot produce the required design tolerances and surface finishes, then discard it and go to Step 1.

iii. Estimate the production time and cost associated with operation plan.
3. If no promising operation plans were found, then exit with failure. Otherwise exit with success, returning the operation plan that represents the best tradeoff among quality, cost, and time.

We now illustrate this analysis on two different designs.

Example: analyzing two alternative designs for a socket

Design #1:


Design #2:


Analysis of Design 1:
Top pocket cannot be completely machined using end-milling because there is no corner radius.
Concentricity can only be achieved by drilling the holes in same setup, but this violates the L/D ratio limit for the smaller hole.

Analysis of Design 2:
Machinable by drilling and end-milling operations.
The best plan requires 13 operations in 3 different setups.
Total time required to machine the socket: 31.13 minutes.
Since Design 2 is achievable, IMACS's output includes an optimal operation plan for the design. As shown below, this plan includes three setups:
Setup 1:


Setup 2:
  

Setup 3:
  


IMACS also gives estimated machining time needed for each step in the operation plan. For example, here are the estimated machining times for two of the faces:
3.29 minutes:


2.54 minutes:



Conclusions

We anticipate that the results of our work will be useful in providing a way to speed up the evaluation of new product designs in order to decide how or whether to manufacture them. Such a capability will be especially useful in flexible manufacturing systems, which need to respond quickly to changing demands and opportunities in the marketplace. Some of the benefits of our approach include:

Since we consider various alternative ways of machining the part, this allows us to consider how well each one balances the need for a quality product against the need for efficient manufacturing. This gives more accurate results than if we considered only one way to machine the part.

The system operates on-line. Thus it helps in identifying potential manufacturing problems early in the design stage.

Our approach is based on theoretical foundations which enable us to make rigorous statements about its soundness, completeness, efficiency, and robustness.

Future plans and work in progress include incorporating an interface to computer-aided fixturability analysis, and extending IMACS to automatically formulate suggestions for how to redesign products to improve their manufacturability.

Acknowledgements. The work on IMACS has been supported in part by National Science Foundation Grants NSFD EEC 94-02384, IRI-9306580, DDM-9201779, and by the University of Maryland General Research Board. Software grants were provided by Spatial Technologies, Ithaca Software, EDS/Unigraphics, and Bentley Systems, Inc. General Electric Corporation, through their Forgivable Loan program, has provided additional support to William Regli.

Personnel. Faculty and students involved in various aspects of the IMACS project have included Diganta Das, Alex Elinson, Satyandra K. Gupta, Ioannis Minis, Dana S. Nau, William C. Regli, and Guangming Zhang. The project leader is Dana Nau.

Appendix B: Optimal Selection of Partners in Agile Manufacturing (OSPAM)

This appendix presents the contents of the VM Web site that are relevant to OSPAM. The OSPAM project focuses on design evaluation and partner selection in distributed manufacturing. It is being conducted at the Institute for Systems Research of the University of Maryland in cooperation with the State University of New York at Buffalo, Westinghouse ESG, Martin Marietta, Orlando and The National Institute of Standards and Technology. The work is funded by the U.S. Army TACOM.

The focus of this project is the design for distributed manufacture of electronic products, such as microwave modules and hybrid assemblies. We are developing methods for
critiquing the design of such products with respect to the production capabilities of potential partners,
alerting the designer to elements of the design that correspond to costly operations, and
selecting a good combination of partners that contribute to the manufacture of the design.
Thus, we are developing a design advisor that integrates partnering (including make or buy) decisions in the design phase, and facilitates design improvements with respect to partner capabilities.

The research contributions of this project include the following:

Information infrastructure :
Developed a STEP-based product information model for Microwave Modules, and implemented of this model in an Object Oriented database.
Developed Object Oriented Group Technology (OOGT) codes for design retrieval and variant design critiquing.
Developed a manufacturing resource model to represent the process capabilities and historical performance of manufacturing firms.

Decision support :
Automated generation of a products OOGT from its STEP information model.
Methods for variant design critiquing.
Methods for high-level process planning in distributed manufacturing.
Methods for plan-based design critiquing in distributed manufacturing.
Methods for multi-objective partner selection in distributed manufacturing.

The project is described further in the VM Web site. A functional diagram navigates the user through the OSPAM architecture. Hypertext within the blocks representing the major modules of the system allows the user to access more detailed descriptions of the modules.

Appendix C: ACIS-TO-STEP Translator Users Manual

This appendix provides step by step instructions on using the ACIS to STEP translator. This process consists of two steps. First, you are required to upload your ACIS sat file to the "ftp.isr.umd.edu" in the incoming/cimlab subdirectory. Then, you need to fill in the required information in the provided form and run the translator.

Upload ACIS sat files:

You are required to anonymous ftp to "ftp.isr.umd.edu" and put the ACIS sat file to the incoming/cimlab subdirectory.

In the following, we illustrate how to upload "product.sat" in your current directory to the incoming/cimlab subdirectory of the "ftp.isr.umd.edu".

$ftp ftp.isr.umd.edu
Name(ftp.isr.umd.edu:"your name"):anonymous
Password: your email address here
ftp>binary
ftp>put product.sat /incoming/cimlab/product.sat
ftp>bye
Goodbye.

Run the translator:

Fill in your name, email address and the name of the .sat file in the form. Then, click on the submit button. The STEP file will be displayed on your web browser once it is processed. This file will also be sent to you by email.

Appendix D: Brief Biographies of the Authors

D.1. Edward Lin

Edward Yi-Tzer Lin is a research engineer in the Institute for Systems Research at the University of Maryland. His research interests are in the areas of production/logistics/material handling systems, computer-integrated-manufacturing systems, factory automation, an object oriented technology. Dr. Lin received his undergraduate degree in mechanical engineering and an M.S. in automatic control engineering, both from the Feng Chia University, Taiwan, and has another M.S. in operation research (1989) and a Ph.D. in Industrial and Systems Engineering (1994) from the Georgia Institute of Technology.

D.2. Ioannis Minis

Ioannis Minis is an assistant professor at the University of Maryland, in the Department of Mechanical and the Institute for Systems Research. His research interests are in the areas of production systems, distributed manufacturing, and machining dynamics and control. Dr. Minis received his undergraduate degree in mechanical engineering from the National Technical University of Athens, Greece (1982), his M.S. in mechanical engineering from Clarkson University (1983) and his Ph.D. in mechanical engineering from the University of Maryland. Dr. Minis is a recipient of the 1993 Earl E. Walker Outstanding Young Manufacturing Engineer Award of the Society of Manufacturing Engineers. He also received the best paper award in the area of Engineering Database Management: ``Use of PDES in Group Technology Applications for Electronics,'' at the 1992 ASME International Conference on Computers in Engineering.

D.3. Dana Nau

Dana S. Nau is a Professor at the University of Maryland, in the Department of Computer Science and the Institute for Systems Research. He is the leader of ISR's Virtual Factories project, and the co-leader of ISR Systems Integration research thrust. His research interests include AI planning and searching, and computer-integrated design and manufacturing. He received a B.S. in applied mathematics in 1974 from the University of Missouri---Rolla, and an A.M. (in 1976) and Ph.D. (in 1979) in Computer Science from Duke University, where he was a NSF graduate fellow. He has had summer and/or sabbatical appointments at IBM Research, NIST, the University of Rochester, and General Motors Research Laboratories. He has received an NSF Presidential Young Investigator award (1984-89), the ISR Outstanding Systems Engineering Faculty award (1993-94), a best-paper award at the ASME 1994 Computers in Engineering Conference, and various other awards. He has more than 150 technical publications; copies of recent papers are available at http://www.cs.umd.edu/~nau.

D.4. William Regli

William Regli holds an appointment as a Computer Scientist in the Manufacturing Engineering Laboratory of the National Institute of Standards and Technology (NIST). He is currently working in the Factory Automation Systems Division's Manufacturing Information Dissemination Technologies group, exploring the use of network information tools to support design, manufacturing, and process planning. This work is under the Systems Integration of Manufacturing Applications (SIMA) thrust of the High Performance Computing and Communications (HPCC) initiative. From 1996-1998 he will be a National Research Council Postdoctoral Fellow with NIST.

In addition, Mr. Regli is a Research Assistant at the University of Maryland in the Department of Computer Science and the Institute for Systems Research, where he is currently completing the Ph.D. in Computer Science. He received a B.S. in Mathematics from Saint Joseph's University in Philadelphia in 1989. His dissertation research involves solid modeling and its applications in computer-integrated manufacturing. He is a member Sigma Xi, ACM, AAAS, AAAI, and is the recipient of the Institute for Systems Research's 1995 George Harhalakis Outstanding Systems Engineering Graduate Student Award.

References


1. International Standards Organization, "Integrated Resources: Geometric and Topological Representation," ISO 10303-42.

2. Ioannis Minis and Jeffrey W. Herrmann, "Design Evaluation and Partner Selection for Agile Manufacturing," submitted to IIE transactions.

3. Vijay Ramachandran, 1995, "Information Models for Agile Manufacturing," Master's Thesis, University of Maryland at College Park.

4. International Standards Organization, "Product Data Representation and Exchange: Implementation Methods - Clear text encoding," ISO 10303-21.

5. Edward Lin, Ioannis Minis, Dana S. Nau, and William C. Regli, "Contribution to Virtual Manufacturing Background Research," Technical Report, The University of Maryland, 1995. Also available at http://www.isr.umd.edu/Labs/CIM/virtual.html.

6. NIST Virtual Manufacturing Web Site, http://elib.cme.nist.gov/msid/trojs/metk/homepage.html.

7. Thomas Register of American Manufactures, http://www.thomasregister.com.

8. Internet Shopping Network, http://www4.internet.net.

9. Power Computing, http://www.powercc.com.

10. Auction on the Web, http://www.onsale.com.

11. UPS package tracking system/service information, http://www.ups.com.

12. TAE Plus Demonstration Programs, http://www.cen.com.

13. NIST Tool Kit, http://www.hike.te.chiba-u.ac.jp/ikeda/documentation/STEP/nist-tk.html.

14. INTERACT Project, http://www.eng.cam.ac.uk/interact/interact.html.

15. E-Cash, http://www.chron.com/content/chronicle/business/95/10/23/internet.html.

16. R. L. Francis, L.F. McGinnis, and J. A. White, 1992, Facility Layout and Location: An Analytical Approach , 2nd edition, Prentice Hall.

17. G. Ioannou, 1995, Integrated Manufacturing Facility Design , Ph.D. Dissertation, University of Maryland at College Park.

18. L.F. McGinnis, 1991, "Computer-Aided Facilities Design Revisited: A Prototype Design Workstation for AGV Systems," Progress in Material Handling and Logistics, Vol. 2., pp. 67-93, John A. White and Ira W. Pence (eds.), Springer-Verlag.

19. PDS User's Manual (Layout Software), Computer Integrated Manufacturing Laboratory, University of Maryland at College Park.

20. Kosuke Ishii. Modeling of concurrent engineering design. In Andrew Kusiak, editor, Concurrent Engineering: Automation, Tools and Techniques , pages 19-39. John Wiley & Sons, Inc., 1993.

21. M. Jakiela and P. Papalambros. Design and implementation of a prototype intelligent CAD system. ASME Journal of Mechanisms, Transmission, and Automation in Design , 111(2), June 1989.

22. David W. Rosen, John R. Dixon, Corrado Poli, and Xin Dong. Features and algorithms for tooling cost evaluation in injection molding and die casting. In Proceedings of the ASME International Computers in Engineering Conference , pages 1-8. ASME, 1992.

23. C. C. Hayes, S. Desa, and P. K. Wright. Using process planning knowledge to make design suggestions concurrently. In N. H. Chao and S. C. Y. Lu, editors, Concurrent Product and Process Design , ASME Winter Annual Meeting, pages 87-92. ASME, 1989.

24. C. C. Hayes and H. C. Sun. Plan-based generation of design suggestions. In R. Gadh, editor, Concurrent Product Design , ASME Winter Annual Meeting, pages 59-69, New York, 1994. Design Engineering Division, ASME.

25. Wynne Hsu, C. S. {George Lee, and S. F. Su. Feedback approach to design for assembly by evaluation of assembly plan. Computer Aided Design , 25(7):395-410, July 1993.

26. S. K. Gupta, T. R. Kramer, D. S. Nau, W. C. Regli, and G. Zhang. Building MRSEV models for CAM applications. Advances in Engineering Software , 20(2/3):121-139, 1994.

27. W. C. Regli, Satyandra K. Gupta, and D. S. Nau. Extracting alternative machining features: An algorithmic approach. Research in Engineering Design , to appear 1995.

28. S. K. Gupta, D. S. Nau, W. C. Regli, and G. Zhang. A methodology for systematic generation and evaluation of alternative operation plans. In Jami Shah, Martti Mantyla, and Dana Nau, editors, Advances in Feature Based Manufacturing , pages 161-184. Elsevier/North Holland, 1994.

29. S. K. Gupta and D. S. Nau. A systematic approach for analyzing the manufacturability of machined parts. Computer Aided Design , 27(5):343-342, 1995.

30. D. Das, S. K. Gupta, and D. Nau. Reducing setup cost by automated generation of redesign suggestions. In Kosuke Ishii, ed., Proc. ASME Computers in Engineering Conference , pages 159 170. Best-paper award winner.

31. S. Gupta, W. Regli, and D. Nau. "Manufacturing feature instances: Which ones torecognize?" ACM Solid Modeling Conference , 1995.

32. W. C. Regli, S. K. Gupta, and D. S. Nau. "Extracting Alternative Machining Features: An Algorithmic Approach." Research in Engineering Design , to appear.

33. S. K. Gupta and D. S. Nau. A systematic approach for analyzing the manufacturability of machined parts. Computer Aided Design 27:5 (1995), 343-342.

34. S. K. Gupta, D. S. Nau, W. C. Regli, and G. Zhang. A Methodology for Systematic Generation and Evaluation of Alternative Operation Plans. In Jami Shah, Martti Mantyla, and Dana Nau, editors, Advances in Feature Based Manufacturing . Elsevier/North Holland, 1994.