Discussion

Today's marketplace is characterized by increasing global competition, shrinking product lifetimes, and increasing product complexity. Industries need to be able to quickly develop new and modified products, and to manufacture products at the right quality, at competitive costs (including environmental-protection-related costs as well as the usual production costs). This makes the design task more challenging, as designers must acquire and process a wide variety of design information and still meet ever-tightening deadlines. To assist designers with this expanded role, manufacturability analysis systems will need to be improved to meet the following performance criteria:

• Scope. As manufacturing industries adopt newer processes and materials, and participate in more collaborative manufacturing with suppliers and customers, the scope of manufacturability analysis systems will need to be expanded to take into account a variety of manufacturing issues that they do not currently address.

• Accuracy. In the analyses produced by a manufacturability analysis system are not sound, this can result in considerable delays and/or financial losses. For example, Petroski [178] describes several cases in which design failures occurred because of errors made by software for analyzing design performance.

• Speed. Since design is an interactive process, speed is a critical factor in systems that enable designers to explore and experiment with alternative ideas during the design phase. Achieving interactivity requires an increasingly sophisticated allocation of computational resources in order to perform realistic design analyses and generate feedback in real time.

1. Ability to handle multiple processes. Many products are produced using a combination of different kinds of processes. For example, engine blocks are first cast, and then machined to final shape. Systems are being developed that handle more than one kind of manufacturing process [15,86,88]. However, manufacturability requirements for different processes are often in conflict. For example, a design shape that is easy to cast may pose problems when fixturing it for machining. It will be necessary to develop ways to handle such conflicts.

2. Alternative manufacturing plans. In many cases it is possible to manufacture a part using different manufacturing processes or combination of processes. Thus to accurately determine the manufacturability of a product, it may be necessary to consider alternative ways of manufacturing it. In certain cases, there might be a large number of alternatives, making it infeasible to consider all of them. In order to preserve computational efficiency in such cases, methods are needed to discard unpromising alternatives while still producing correct results. [25] provides an approach to this problem in the context of machined parts---but methods still need to be developed for other manufacturing domains.

3. Virtual enterprises and distributed manufacturing. Manufacturing industries are relying increasingly on distributed manufacturing enterprises organized around multi-enterprise partnerships. In such environments, manufacturability analysis cannot be done accurately without taking into account the capabilities of the various partners that one might potentially use in order to manufacture the product. Projects are underway to address this problem (e.g., [179]), but the work in this area is still largely in its early stages.

4. Process models and virtual manufacturing. A static knowledge base of manufacturing process capabilities may not be suitable for determining the manufacturability of a product in cases where the manufacturing processes are very complicated (such as near-net shape processes), or where the manufacturing technology is changing at a fast pace (such as composites processing). Projects such as [50,87] address this problem by analyzing manufacturability using data obtained from process models and manufacturing simulations. Some of the problems remaining to be solved include the development of better and up-to-date process models, and better integration of process models with manufacturability evaluation methods.

5. Manufacturability rating schemes. Fast decision-making regarding the manufacturability of proposed designs is becoming more important than ever. For helping designers and managers to make engineering and financial decisions, ratings of a qualitative or abstract nature will not be particularly useful---instead, the manufacturability ratings will need to reflect the cost and time needed to manufacture a proposed product, as done in [25]. We expect that future manufacturability rating schemes will not only represent production time and cost, but also provide detailed breakdowns of the time and cost of manufacturing various portions of the design. For such purposes, manufacturing-handbook data will not necessarily be accurate enough; instead, company-specific data (obtained, for example, via virtual [50,87] and physical [87,172] simulations) will be needed.

6. Accounting for design tolerances. Designers note dimensional and geometric tolerances on a design to specify the permissible variations from the nominal geometry that will be compatible with the design's functionality. Design tolerances are important aspect of the design and significantly affect manufacturability---but most existing systems have limited capabilities for analyzing the manufacturability of design tolerances. For example, most work on automated tolerance charting [176,177] focuses mainly on computing the optimum intermediate tolerances and has not been integrated with manufacturability analysis systems. In order to develop manufacturability analysis systems that are capable of handling problems posed by design tolerances, research in the area of estimating accuracy of parts made by different processes is essential.

7. Automatic generation of suggestions for redesign. For a manufacturability evaluation system to be effective, it is not always adequate to have the manufacturability rating of a component and a list of its production bottlenecks. Since designers often are not specialists in manufacturing process, they may not be able to rectify the problems identified by the manufacturability evaluation system. This is particularly true for cases where the part is manufactured by multiple manufacturing methods or is produced by a supplier. To address such problems, manufacturability analysis systems will need the ability to generate redesign suggestions.

   Most existing approaches for generating redesign suggestions [15,26,43] propose design changes on a piecemeal basis, (e.g., by suggesting changes to individual feature parameters)---but because of interactions among various portions of the design, sometimes it is not possible to improve the manufacturability of the design without proposing a judiciously chosen combination of modifications. Also, existing systems usually do not take into account how the proposed changes will affect the functionality of the design. This will require the systems to be integrated with some form of functionality representation scheme and manufacturing data base. Some work is being done to overcome both of these drawbacks [64], but it is still in the early stages.

8. Product life-cycle considerations. For more comprehensive analysis of the total cost of a product, other life-cycle cost considerations also have to be taken into account [180,181]. Recently there has been a proliferation of tools for critiquing various aspects of a design (performance, manufacturability, assembly, maintenance, etc.). As designers begin to use multiple critiquing tools, we anticipate problems in coordinating these tools. Since different critiquing tools are written to address different manufacturing objectives, the recommendations given by these tools will sometimes conflict with each other. Thus it will be necessary to develop ways to reconcile these conflicting objectives, so as to avoid giving the designer confusing and contradictory advice [182].

9. Making use of emerging technologies. Future manufacturability evaluation systems will need to make use of state-of-the-art developments in computer and information technology. It is conceivable that in future these systems will be available on-line for users world-wide. For achieving high accuracy at a fast response time the systems will be able to use computing capabilities at remote locations at a distributed manner.

In this survey, we have attempted to present a cross-section of the research community that has emerged to address the wide variety of problems faced when constructing automated manufacturability analysis systems. As evident in the above discussion, many important advances have been made. It is our belief that these successes demonstrate the huge potential impact that might be made by such systems.

However, there are a number of fundamental research challenges that need to be overcome in order to make automated design analysis tools realize their full potential. As evidenced by this survey, the current state-of-the-art contains many diverse, domain-specific systems. Each approach presents the community with a different aspect of the overall problem. Creating a truly interactive, multi-domain, multi-process system capable of satisfying the conflicting constraints posed by these domains and provide intelligent feedback and alternative suggestions to the designer. We are optimistic that the community is up to the challenges.