Improving Design Quality

"A lot of attention is paid to manufacturing quality control," observes Gavin Finn, president and CEO, Prescient Technologies (Boston, MA). "But in engineering, where there are virtual product models, that quality philosophy is not established."According to Finn, the number of errors generally within models is such that "it's not even one sigma."Think about what this means.

"A lot of attention is paid to manufacturing quality control," observes Gavin Finn, president and CEO, Prescient Technologies (Boston, MA). "But in engineering, where there are virtual product models, that quality philosophy is not established."

According to Finn, the number of errors generally within models is such that "it's not even one sigma."

Think about what this means. In manufacturing, efforts are underway to achieve six sigma quality. That's the physical product. Six sigma means 3.4 defects per million opportunities. Yet in the engineering arena, where the digital model is made, a model that is used to make things like dies to produce the physical product, the quality of that model may be dubious, to put a good spin on it. One sigma means 690,000 defects per million opportunities.

3D design may drive product development time forward fairly quickly...until it needs to be translated into a form so that it is turned into tooling. Finn suggests that people downstream from engineering can spend 30 to 85% of their time "cleaning up" models. What's more, not all of the information is necessarily contained within the model: 50% of the engineering change orders or other information may be absent. Finn states, "Interoperability is one of the key issues relating to data quality. Companies recognize that their success depends on how effective the enterprise is at sharing accurate data. As companies begin addressing the issue of interoperability, they are realizing that, in many cases, they are moving incomplete or inaccurate information to downstream applications." In other words, while the proverbial "silos" may be eliminated, it doesn't do a heck of a lot of good if inaccurate or incomplete information gets in the way of getting products developed and produced. Additionally, while there has been the promise that the models can be reused for subsequent product derivations, Finn and his colleagues have discovered that many times people are starting from a clean screen, not reusing the original.

Because of these issues, Prescient has developed software tools for what Finn is calling "engineering design quality." The objective of these tools—the PrescientQA family—is to permit qualitative assessment of digital design data during the engineering design cycle. It should be noted that the Prescient software isn't used to do the engineering design but is coupled with the design systems—including CATIA, Pro/Engineer, and Unigraphics (and it runs on engineering platforms including Windows NT, IBM AIX, Sun Solaris, and HP-UX)—so that it can obtain information to measure the performance and output of the design functions.

The top level of the software suite is DriveQA. This is the manager's overview, one that provides summaries, reports, and analyses with a graphical interface that can be readily understood, such as a pie chart identifying the most common failures and a bar graph of the 10 most common failures. The simplicity allows quick decision making based on data, not opinion (which is a fundamental approach in quality improvement).

But all of the displayed information must come from somewhere, which is where some of the other packages within the family of tools come into use. There is DesignQA. This package allows best modeling and design practices to be captured. Then, during the design of a given product, it determines whether these practices are being followed or missed. To assist in the development of best design practices, there are 124 quality packages included in DesignQA. Instead of having best practices codified and filed in three-ring binders that collect dust on a bookshelf, this software allows them to be an integrated part of the on-going design and development process.

GeometryQA looks at geometric accuracy and helps assure that the number of design iterations are minimized by identifying any geometric problems in the design that could interfere with manufacturability. A feature that may not be readily produced in solid material or problems (e.g., gaps) will be highlighted so that the design engineer can reconsider that feature rather than passing it along only so that the manufacturing engineer has to work on it—and quite possibly send it back for modification.

Product data management (PDM) systems are becoming more common in engineering operations. So there is the CertifyQA package. Essentially, this software assures that the models released into the PDM system have a quality "stamp," that the models contain good, complete data.—GSV