Improving Model Quality for 3-D CADImproving Model Quality for 3-D CAD

As CAD models continue to take on a broader, more significant role in the development of new products, it is becoming imperative that these files flow smoothly into downstream applications. Industries across the board are realizing significant savings of time, money, and quality improvement through the expanded use of these models. As a result, model quality has never been more important. Models that contain hidden errors can cause significant time delays, loss of design intent, inferior product quality, and additional effort, wiping out the time savings they are meant to provide.

IMM recently spoke with Carl Izurieta, customer liaison for International TechneGroup Incorporated (ITI--Milford, OH), to explore solutions to model quality problems. Izurieta works with organizations to implement model quality programs and capture interoperability requirements for new product development.
IMM: What are some of the causes for error in 3-D solid models, and how much time do you estimate these errors waste?
Izurieta: Model quality problems are rooted in a variety of contributing factors that range from bugs in the CAD modeling system to data translation software errors. Also, modeling techniques that do not anticipate the needs of downstream shape-based applications can create anomalies. These human-error types of problems arise when CAD designers don't fully understand the geometric requirements of downstream applications or have no efficient way to validate their CAD models against those requirements to identify the potential problems.

We met recently with manufacturers to determine how model problems impact product development, and we found that up to 70 percent of the hours spent during FEA and mold filling simulation, for example, were wasted correcting geometry problems. Rapid prototyping, NC tool path generation, and product data exchange functions were likewise requiring 50, 20, and 20 percent of the users' time respectively to rework the geometry problems.

Regardless of cost or the vendor who developed it, every CAD system in use today is susceptible to geometric and topological anomalies when creating complex 3-D surfaces and solids. Often these problems don't surface until well downstream of design--during analysis, rapid prototyping, data exchange, NC programming, or other manufacturing applications.

Most of the time, these problems present more than just minor inconveniences for those who receive them. CAD model problems can bring product development processes to a grinding halt. Sometimes the corrupt model is shipped back to the designer to be fixed. A more common scenario, though, has the downstream user forced to make corrections or reconstruct the design. Obviously, this practice can have a detrimental impact on design intent along with the obvious ramifications associated with lost time and cost overruns.
IMM: Can you categorize model quality problems and offer some concrete examples of the problems?
Izurieta: Model quality problems can be generally categorized into three areas: structure, accuracy, and realism. Structural problems include loop orientation inconsistencies, missing geometry, and self-intersecting geometry among others. Structural errors violate the solid modeling application's own rules for what constitutes a correct model, and can also cause modeling programs to crash without warning. These types of errors can cause programs for finite-element mesh generation, NC toolpath generation, and intersystem translation software to behave somewhat unpredictably.

An example of a structural error is a face with an edge that isn't shared by another face. In a manifold solid volume, such edges shouldn't exist because they cannot physically be manufactured. They occur because some solid modeling programs allow nonmanifold topology as a midpoint to creating complete volumes.

Accuracy requirements place limits on gaps between geometric entities such as vertices, edges, and faces that are adjacent. They can also limit the minimum sizes of trimmed entities such as edges, faces, and regions. Nontrivial gaps occur because intersections of curves between nonplanar surfaces are approximated in most solid modelers. Approximations are used when the precise intersection between two geometric entities (faces, curves of intersection, vertices where intersection curves meet) is too complex to compute exactly. Solid modelers use different tolerances to compute the maximum deviation allowed between topological entities.
IMM: Then what happens?
Izurieta: If the deviations between entities are too large, toolpath and finite-element mesh generation programs can fail. They can uncover gaps in geometry that are too small to be seen in shaded or hidden-line images of a model.

The translation between programs can also fail if the maximum allowable tolerances between surfaces and edges in the exporting program are larger than those of the importing program.

All CAD modeling systems must balance the accuracy, or precision, of models with the amount of geometric information required to define them. Extremely precise models require complex and large data structures to define them. In general, the smaller the gaps, the smaller the edges and faces may become in complex models.

Finally, realism errors render a part unmanufacturable due to physical limitations. Realism errors include transition cracks and sliver faces. Transition cracks in solid models, like physical cracks in engineering materials, are nearly invisible gaps between features of a model. Like physical cracks, they may not extend completely through the object. Slivers are small, elongated faces that are generated by the system to patch between larger surfaces in a model.

Additional restrictions on the realism of model features are added by many concurrent engineering applications such as finite-element analysis, NC toolpath generation, and rapid prototyping. For example, these tools are very sensitive to unrealistic features such as sliver faces, minute edges, and very acute angles between edges at a vertex.
IMM: Obviously, we can't expect CAD systems to change overnight in these areas. So what can designers do to correct the errors before they become obstacles in the manufacturing process?
Izurieta: Regardless of the reasons why they exist, the fact remains that if model quality problems aren't effectively resolved, downstream processes simply can't work. The solution is to implement a model quality program, which allows designers to better identify and resolve the source of these problems through a combination of improved modeling techniques, better software bug reports, and real-world user requirements for current research in this field.

Likewise, the downstream software user should implement such a program as well. This allows the recipient of a CAD file to quickly analyze the model and locate problem areas before production begins. The file can then be returned to the designer with errors highlighted for quick turnaround. The downstream user may choose to make the modifications depending on the severity of the errors.
IMM: What tools are available for those interested in starting a model quality program?
Izurieta: Currently, there are several model quality software packages that allow integrity to be easily analyzed and verified during construction, eliminating the need to rework or rebuild a model further down the road. Some packages, such as our CAD/ IQ, also let users present their CAD system provider with detailed reports pertaining to bugs in their software. Once identified, many of these problems can be resolved by the designer in the early stages of development at a point where changes can be incorporated quickly and at less cost with full knowledge of the part's design intent.

Other problems, such as software bugs or translation errors, will require longer-term efforts by CAD/ CAM/CAE vendors and researchers. But while these bugs or errors were communicated through user phone calls and hotline reports in the past, a comprehensive model quality control program can provide more detailed and valuable insight into identifying these problems.

It is important that these tools do more than just check models against "rules." They should provide the CAD user with the power and flexibility to analyze the model for conformance to a wide variety of applications and specific CAD/CAM/CAE system requirements. This lets the designer anticipate system restrictions and ensure that models created will flow seamlessly into all downstream applications. In short, this allows unrestricted interoperability to be designed into the model.

Implementing a model quality testing program can yield breakthrough levels of improvement. Validating CAD files prior to release significantly reduces model rework time. ITI estimates that model rework time can be cut by 50 percent in downstream FEA, product data exchange, and NC applications. That number jumps to savings of up to 80 percent for rapid prototyping.