John Deere develops new AI robotic welding solution

John Deere uses Intel’s artificial intelligence technology to solve age-old costly welding product manufacturing challenges.

A solution is currently being tested that uses computer vision to automatically identify common faults in automated welding at their manufacturing facilities.

At 52 plants around the world, John Deere uses the gas arc welding (GMAW) process to weld mild and high-strength steels to manufacture equipment and products. In these factories, hundreds of robots consume millions of pounds of wire electrode annually.

For uninterrupted wire electrode delivery, John Deere utilizes the services of professional equipment haulers. This ensures an uninterrupted supply of equipment and materials to its production sites.

With so many welding jobs, John Deere is experienced in finding solutions to welding problems and is always looking for new ways to solve potential problems.

One of the common welding problems faced by industry is porosity, where pits form in the weld metal due to trapped gas bubbles as the weld cools. Notches weaken the strength of the weld.

Traditionally, GMAW defect detection was done manually and required highly trained technicians. Earlier attempts to solve the problem of weld porosity in an industrial welding scheme have not always been successful. If these defects are discovered later in the manufacturing process, they will require rework or even write-off of entire assemblies, which can be disastrous and costly for manufacturers.

The ability to work with Intel to use artificial intelligence to tackle weld porosity was an opportunity to combine two of John Deere’s core values ​​- innovation. Ultimately, Intel and Deere combined their expertise to develop an integrated end-to-end hardware and software system that can generate real-time analytics.

Using an inference engine based on a neural network, the solution registers defects in real time and automatically stops welding. The automated system enables Deere to solve problems in real time and produce the high quality products that Deere is famous for.

The detection of artificial intelligence defects in the peripheral solution is based on Intel Core i7 processors and uses the Intel Movidius VPU and Intel Distribution of OpenVINO toolkit, implemented using the industrial and industrial region of the Moravian-Silesian Region.

John Deere Parts
John Deere Logo