Industrial parametric CAD model efficiency: how complexity and quality characteristics influence recommended modelling strategies

This study concerns the development of parametric CAD modelling strategies to improve efficiency, based on measured complexity and quality characteristics. This is especially valuable to industry where Parametric CAD is the primary development tool in the engineering design process. Nonetheless, when developing parts, the use of a history-based model tree and numerous interdependent references is prone to error and misuse. This leads to suboptimal quality and efficiency, especially as file complexity increases.

Quality measures based on the geometry and construction features that characterise parametric CAD have long been studied. One of the most cited is the linguistic model, and two aspects of this, morphological and semantic relate directly to the sources of complexity within parametric CAD. Morphological quality is associated with the topological and geometrical correctness of the shape, while the semantic correlates to the level of design rationale contained in the construction information. Methodologies and strategies have been proposed in literature to improve the level of quality, but these are rarely evaluated by professional CAD engineers. Additionally, test parts published by researchers remain simpler than the most complex industrial equivalents.

This study examines parametric models using the PTC Creo CAD software application, concentrating on the ‘computational efficiency’ of these files, rather than more subjective ‘user efficiency’. Extensive complexity and quality measurements were taken from a diverse database of over 1000 industrial model files belong to Triumph Motorcycles Ltd. Each model was subject to various typical administrative actions, related to both the proprietary data (Open, Save, Regenerate) and the transfer of neutral data (Export, Import), and the duration to complete each was carefully recorded on calibrated equipment. The results were evaluated to examine general distribution of typical complexity and quality measures in the database and their effect on duration. They were then broken into more granular measures to construct the most efficient data analysis models to predict action durations. 

Using initial analysis, 12 high-complexity models were identified as worthy of optimisation. These were remodelled by the author, a Creo CAD expert with over 20 years of commercial experience, to determine how specific improvements affected computational efficiency. These parts enabled the auditing of modelling methodology as well as validating the predictive models constructed using the database. Analysis showed that the ability to predict neutral import duration was sufficiently high to use as a novel morphological complexity metric. Neutral data export prediction accuracy was also high, but the proprietary prediction accuracy less so. Nonetheless, they still allowed insight into the part characteristics having the greatest influence on durations.

Observations from the study in combination with those from other works allowed six modelling recommendations to be made of increasing importance to be presented to Triumph:

1. Simplify morphology of primary solid

2. Eliminate unnecessary reference geometry

3. Use semantically simple equivalent Model Tree Features (MTFs)

4. Evolve the model tree approach with respect to model requirements

5. Consider quality only at the most effective point in model optimisation

6. Continually review modelling approach and project methodology