A Brief Look at Methods & Technology
How Does it Work?
It's a common misconception that shape-search tools compare CAD models directly. That is simply too complex an approach to be scalable. Instead, every shape-search application uses some method to represent geometry. The vast majority use a parameter such as surface area or volume, using the assumption that if two models have very similar parameters, that they might be identical or near-duplicate models. The limitations of these approach are well known.
"Yet most engineering departments are still limited to out-of-date parametric search tools that produce either thousands of hits, or no hits at all.
A typical part search would turn up as many as 2,000 hits, forcing frustrated engineers to simply design new parts from scratch."
Accelerating Innovation with New Data Strategies, tdwi.org
A New Approach
CADseek was the first and is still the only shape-search application to work purely from geometry.
Since all models have shape... no model can be invisible to a shape search simply because it lacks a text descriptor, or falls outside a parametric range.
CADseek uses pattern recognition and a mathematical transformation to fingerprint each model so that the entire geometry is transformed into a numerical representation at the full resolution of CAD.
Concise -- Comprehensive -- Comparable
In simplistic terms, the more similar any two models are to each other, the more similar their numerical fingerprints will be, and from these signatures a highly accurate degree of similarity can be computed.
CADseek is able to produce search results that are simultaneously comprehensive and without clutter.
INTELLIGENCE for SIMILARITY
Similarity is much harder to detect than identical models because of the endless number of ways that parts can be different but still similar. CADseek's greatest strength is its ability to not only see similarity even as size and features change, but also accurately score the degree of the similarity. This capability is a tremendous benefit for usability and critical for geometry-based analytics.
A common use case for shape-search is using a new model, perhaps one received from a customer or supplier, as a search target to search the CAD library for existing similar designs. CADseek's numerical signatures provide this on-the-fly search capability, allowing new models to be used as search targets with search results produced in just a few seconds.
CADseek has demonstrated its ability to maintain sub 2 second response in multi-million item libraries.
CADseek is not a learning algorithm but instead works on any 3-dimensional geometry, from large mechanical assemblies, to tiny protein models, without any customization required.
The algorithm can encode all types of native CAD (both models and assemblies), intermediate and neutral formats such as STEP, IGES or STL, and even 3D laser scans.
The processes of encoding a CAD library to create the shape-search index is fully automated with no human requirement for sorting or tagging.
Updating processes are also fully automated and can occur on a daily, weekly or monthly process.
Additional technical information is available, which you can request using CONTACT US.