Recycling of printed circuit boards by robot manipulator: A Deep Learning Approach

摘要

Electronic circuit boards in mobile phones, PCs, home appliances, etc. that are no longer in use contain many useful parts such as resin and metals. In addition, there are toxic elements that need a specific treatment. Japan now depends almost entirely on foreign countries for the required metal resources. Since most electronic waste is composed of complex materials such as plastics and metals, it is difficult to isolate them and return them to a single resource. In addition, dedicated equipment is required to recover resources, and sensors are often used, which is a factor in increasing costs. In this paper we develop a system for recycling of PCB boards through classification, recovery and management of electronic parts. In our method, the PCBs are divided in small pieces which moves in a belt conveyer. For component recognition, we trained and compared the performance of two Convolution Neural Networks (CNN). The camera detects each PCB component and classify it. In our implementation we classified the components in five classes. Based on the camera information and the conveyer speed the robot manipulator motion is generated and the classified parts are sorted in separate containers. In addition, we investigated the shape and material of the gripper for easy grasping of PCBs components.