Speeding Up Non-Destructive Testing of Welded Parts by Assisting Human Inspection of X-ray Images

The Problem

The existing manual inspection suffered from high escape rates, high overkill rates, and high costs from wasted work.

DarwinAI developed and deployed a vision-based AI that assists human inspectors by detecting localized defects on digital X-ray images.

The Results

fewer false positives

50%

faster inspections

90%

annual ROI

10x

Difficulties with the existing inspection process

• High Escape Rates
• High Overkill Rates
• Unnecessary Rework

Build versus buy?

The manufacturer recognized that a vision-inspection aid could help to address these issues, but they also knew that:

• developing such systems is time consuming and often requires specialized expertise
• most solutions on the market depend upon large volumes of training data to become reliable—often exceeding what is available or economical

DarwinAI offered a compelling alternative: a fast, economical—and reliable—solution that requires significantly less training data.

DarwinAI’s Solution

In only two months, DarwinAI developed and deployed a purpose-built AI system.

By accelerating and improving the quality control process, DarwinAI’s AI-assisted X-ray inspection process paid for itself within its first month of operation and is projected to deliver 10x ROI, annually.