AI in the Workshop: How Geometric Similarity Search is Revolutionizing Mechanical Estimating

Introduction

In the precision mechanics sector, time is a resource as critical as micron-level accuracy. Yet, the estimating process often becomes a bottleneck: hours are spent searching through historical databases for a component similar to the one requested, relying on engineers' memory or exhausting manual searches.

At Envision Data, we recently developed an innovative solution for an european leader in high-precision mechanics, transforming their historical archive from a "static memory" into an "active business tool."

Mechanical Design and Technical Drawing Department

The Challenge: The Hidden Cost of Manual Search

For a company operating in the mechanical sector, every new Request for Quotation (RfQ) accompanied by a CAD drawing (STEP, DXF, or PDF) requires in-depth analysis. The traditional process involves:

• Mnemonic Search: Engineers try to remember if a similar part has been produced before.

• Manual Comparison: Analyzing thousands of past drawings to retrieve actual manufacturing costs.

• Risk of Inaccuracy: Quotes based on approximate estimates can lead to margin erosion or lost orders.

  
  

The Solution: The "Digital Fingerprint" of Components

We designed an AI system that doesn't just search for files—it "understands" the shape of the parts.

The technology is based on a geometric similarity search engine that operates in three steps:

1. Geometric Analysis: The system extracts a unique numerical descriptor from every file (STEP, DXF, PDF)—a true "digital fingerprint" of the solid geometry.

2. Vector Indexing: These fingerprints are stored in a vector database (a "super-directory") capable of handling tens of thousands of models.

3. Instant Search: When a new drawing arrives, the system compares it against the entire historical archive in seconds, returning the 5-10 most similar components produced in the past, complete with their actual historical costs and production times.

   
   

Tangible Business Benefits

Integrating this AI platform into the R&D process brings immediate advantages:

• Extreme Speed: Search time is reduced from hours to just seconds.

• Data-Driven Precision: Quotes are no longer based on "gut feeling" but on certain historical data and actual production costs.

• Leveraging Know-how: Decades of company experience become immediately accessible to the entire team.

  
  

A Technical Milestone: From 3D Files to 2D Drawings

A crucial aspect of the project is the ability to handle multiple formats. While the STEP format ensures near-perfect accuracy thanks to its rich 3D data, we implemented computer vision algorithms to extract information even from PDF and DXF (2D) files, ensuring useful results even when a three-dimensional model is unavailable.

Conclusion

AI is no longer a futuristic technology but a practical tool to make manufacturing companies more competitive and agile. With the “Similarity Search for Technical Drawings” project, Envision Data confirms its commitment to supporting Swiss industrial excellence through digital innovation.