Delta Electronics debuts dual-arm robot for AI server production lines

Delta Electronics is expanding its AI-driven manufacturing capabilities with the unveiling of its "embodied intelligence dual-arm collaborative robot" solution at the Taipei International Automation Exhibition. This advanced system integrates an AI robot control software platform with AI models and equipment control, enabling robots to autonomously learn, analyze, and execute tasks based on voice or visual input.

The company aims to apply this technology to complex, high-mix, low-volume assembly processes, particularly for AI servers. Delta is also enhancing its entire production line with AI-powered digital twins, a technology already implemented in its Thailand factory for AI server power supply production. This move signifies a shift from automation towards greater autonomy in manufacturing.

“

As global localization advances cross-domain intelligent manufacturing models, and the manufacturing industry faces labor shortages, factories will move from automation to autonomy, with AI, intelligent robots, and digital twins becoming key.

Central to the dual-arm robot solution is its AI robot control software platform, which connects diverse AI models and equipment controls. Utilizing various data types like language and images, the robot can independently learn tasks, analyze workflows, execute actions, and make corrections, thereby simplifying the integration of AI robots for manufacturers. The system incorporates NVIDIA's Jetson AGX Orin module and Isaac ROS software suite, providing real-time 3D environment modeling and dynamic obstacle avoidance capabilities.

“

The dual-arm collaborative robot can autonomously learn, analyze, and execute tasks through voice or visual commands.

Beyond robotics, Delta is leveraging AI for its digital twin technology across entire production lines. This includes simulating and optimizing product assembly, production line design, equipment validation, and defect detection. For instance, in PCB spray coating processes, Delta uses its virtual machine development platform, DIATwin, and NVIDIA's Omniverse libraries to automatically generate optimal paths and convert them into machine recipes. This virtual validation significantly shortens the preparation cycle for mass production.

Additionally, Delta is employing AI-driven defect image generation technology, powered by NVIDIA's Cosmos open-world foundation models, to create diverse defect patterns and surface textures. This accelerates the training of AI defect detection models, improving the detection rate of automated optical inspection (AOI) by 17%. The preparation time for defect models of new, unknown products has been reduced from three months to just two weeks, further speeding up new product introductions and production line deployment.

“

This technology can be further applied to multi-process, complex assembly processes such as AI server assembly, increasing production line changeover and multi-tasking flexibility.