Deep-Learning Based Robotic Manipulation of Flexible PCBs

In the past 10 years, due to the fast development of 3C industries such as mobile phones and computers, people have higher requirements for the automatic soldering technology of flexible PCBs. However, the deformation and the small size of flexible PCBs open up significant challenges to robotic soldering. This paper proposes a deep-learning based manipulation scheme for automatic soldering of flexible PCBs. The proposed controller can enable the robot to automatically contact the flexible PCB first, then actively control the flexible PCB to the desired position with the visual feedback, and finally, the soldering machine will solder the flexible PCBs smoothly. First, the approach of deep learning is used to detect the position of the solder pad (feature). Then, the vision-based controller drives the robot to manipulate the solder pad to the desired position, such that the soldering machine can work to solder two pieces of flexible PCBs together. The use of a deep learning approach can explore the human’s experience to improve the accuracy of detection and hence deals with the issues of clustered environment, change of illumination, and different initial position, etc. The proposed detection approach and control scheme is implemented in a soldering robot for flexible PCBs and the results validate the performance of the proposed methods.