Automated Arc Welding Defect Detection

PathPartner, in collaboration with Intel, has launched a very low latency automated defect detection solution for robotic arc welding (GMAW) in heavy equipment manufacturing. The camera-based solution runs on the edge, detects weld porosity defects using a proven AI algorithm and can be used to stop the robotic arm immediately in case an anomaly is identified. This allows manufacturers to find weld defects and reduce delays in the production process, optimize costs and increase overall productivity.

The solution uses deep neural networks based AI model at the edge and is tested to deliver an accuracy of 97%. It is powered by scalable Intel® Core™ i7 processors and uses Intel® Movidius™ VPUs and the Intel® Distribution of OpenVINO™ toolkit to detect arc welding defects.

AI-Based Automated Weld Defect Detection
Move away from Manual QA

The solution consists of a pre-trained machine learning algorithm running on Intel Edge processors. It takes input from a ruggedized industrial camera and runs on an industrial PC based on Intel Edge AI processors. Multiple cameras can be connected to the industrial PC to enable cost optimizations.

  • AI defect detection model : Pytorch Action Recognition
  • Input Resolution: 640 X 512; Video input : 30-60fps
  • Output : Weld defect detection and probability
  • Works with Intel® Distribution of OpenVINO™ toolkit
  • Ruggedized USB/GiGE Camera
  • Industrial PC based on Intel Edge Processors with support for USB/GigE/Modbus/OPC communication. Example: Intel® Core™ i7 processors with Intel® Movidius™ VPUs

Download Solution Brief

Know more about all the features of our solution, the benefits it can drive and how to get started for your custom weld defect detection scenario.

Features and Specifications:
Drive weld defect detection with accuracy, reliability and speed

High Accuracy

Delivers 97% accuracy on tested dataset, highly reliable

Low Latency

Optimized to run at very low latency on scalable Intel platforms on the edge


Can process inputs from multiple camera sources, concurrently

Optimize costs

Less wastage, Less scraping, Savings on materials cost, Saving on operational costs .

Reduce delays

Don’t wait for QA to tell if there are defects and retain. Take corrective action.

Increase productivity

Overall process becomes faster and efficient as well as automated

We make AI work for you :
Weld Defect Detection Algorithm For Your Use-Case

Every welding process and set up is unique. An AI model that works for one use-case, may need to be fine tuned for other use-cases. PathPartner leverages its wealth of industry experience, custom AI algorithm development expertise and AI engineering talent with extensive hands-on experience on Intel platforms, to build the perfect AI solution that works for your use-case

Get Started in 3 Easy Steps
Evaluate us on AWS

The defect detection AI algorithm is available for a free evaluation on the AWS marketplace. All you need to get started is an input stream from your welding setup that you want to test. Get started without worrying about the hardware details or porting challenges.

Back to Top