Conclusions

Creating families of parts requires significantly more effort than creating a single part. A single part can be constructed to known dimensions, even a parametric part which can be modified later by changing the dimensions and locations in the part. The major difference between creating a single parametric part and a master part for a family of parts is that the regeneration of a single part is done through user interaction with the CAD program. Operations can be performed on the model after it has been regenerated. In a master part, the final member parts are generally created in their final form without user interaction. These parts have to be much more robust than their single-use parametric counterparts.

The extra effort put into creating a family of parts results in a substantial overall savings of effort. Creating a robust model will also often lead to a smaller, more efficient model, saving storage space by producing smaller member parts. Even if the family is not used in a library of standard parts but only for a single project, the savings are significant. More detail can be included in the master part while requiring only a fraction of the effort that including the same detail in every member part would require. Thus, family of parts generally results in smaller, more efficient, and higher quality models created in a fraction of the time that building individual members of the family from scratch would require.

Creating efficient, robust, families of parts is not a simple task. This methodology has made is significantly easier for new users to create families of parts. Many potential problems exist that are not readily apparent to the first-time user. This document, which defines many of these potential problems, will help the new user to avoid these problems. In addition, the methodology contains several suggestions for increasing the quality of the master model while decreasing the time required to build the parts. An example of such a suggestion (drawing up a detailed method including all variables and equations before starting to actually build the model) was developed for use in the final two test cases. As a result, the final models were improved considerably over the original proposed models.

The requirement that the families be easy to access led to the creation of an efficient search engine. Instead of searching through hundreds of entries in a catalog to find the proper part then having to match that part to a CAD model, the two operations are combined in a computerized search that returns not only the name of the CAD model, but the vendor part number if available, and even generates the CAD model of the part if necessary.

The time required to find and generate a member part has been reduced to a few of minutes, and the designer never has to leave the workstation to find a catalog. Previously if a designer wanted to create one of these parts, it could take hours to find all the necessary information about a part, determine a method of modeling it, and finally model it. The individual part models are often less efficient and detailed than the family parts. The amount of time necessary per part to create a more efficient and detailed part is significantly higher for single parts. In families of parts, the effort required to create a more efficient and detailed part is roughly the same as in single parts, but the benefits are multiplied by the number of member parts in the family.

In a production environment where CAD plays a large role, the issue of disk space used by the parts is also a major concern. Parts are routinely moved from disk to tape when they are not being used. The storage space for the parts must expand to fill the ever growing need for more storage. The family of parts methodology alleviates this problem. Member parts that aren't needed can simply be deleted, but when a designer needs one, it can be regenerated in a matter of minutes. This project produced a library of almost fourteen hundred parts that takes up only 3 and a half megabytes of disk space when none of the parts are being used. The member parts will never have to be moved from disk to tape for archival storage. Also, if a particular family isn't needed anymore, the master part and table can be moved to tape, and the member parts deleted without losing any information.

The family of parts methodology makes creating a family of parts simple enough that anyone with a basic understanding of parametric modeling can create a usable family of parts. During the survey, over 30 groups of parts were suggested as candidates for this process. Many of them were eliminated from the list of possible families of parts due only to the small number of parts that exist in the family. These parts can still be created using this methodology, but making them into families of parts would not be worth the extra effort. Instead, they should be made from parametric models into regular library parts. Of the remaining parts, all but one of them could be made into families of parts with this method. The final group of parts, rack-mount panels, would have to be split up into smaller families by the number of mounting holes in the panel. A generic model that would cover all possible panels could not be made, as the panels have a variable number of mounting holes that occur at standard locations, but that are not evenly spaced. CADDS5 can only create a variable number of holes at evenly spaced locations.

The Family of Parts software in CADDS5 serves as a starting point for the creation of a usable library of standard parts. However, it has a poor user interface and has no system for part management and database administration. This thesis has made up for several of these shortcomings, and has created the core of a working library that can be easily used by all of the designers without requiring detailed knowledge of the details behind the implementation. For those who want to expand the library, it has provided useful information that will help them create high-quality parts that will work well with this system.

Last Modified: Wed Aug 28 14:41:29 EDT 1996