Toshiba AI Technology Catalog

  • Operation and Control

Laser welding control technology using reinforcement learning

Optimizes complex system control without modeling the system.


  • Using a simple laser welding simulator, learns methods for dynamically controlling welding conditions (e.g., laser power, welding speed, spot diameter) adjusting for gaps between materials and melting pond conditions.

Applications



  • Welding process automation/optimization
  • Laser processing automation/optimization

Benchmarks, strengths, and track record



  • Can optimize controls without system modeling by a specialist or teaching by an expert because it learns control methods for minimizing error of bead width through trial and error on a simulator.
  • Uses an original learning method (a new reward definition and a learning parameter update method) because welding speed cannot be optimized by applying conventional reinforcement learning methods.

Inquiries



Contact the Toshiba Corporate Research & Development Center

Please include the title “Toshiba AI Technology Catalog: Laser welding control technology using reinforcement learning” or the URL in the inquiry text.
Please note that because this technology is currently the subject of R&D activities, immediate responses to inquiries may not be possible.