Printed circuit boards (PCBs) from waste electrical and electronic equipment (WEEE) constitute a valuable material flow, as they contain important raw materials such as gold, copper, palladium, tantalum and silver.

However, consumer electronic devices, information and telecommunication technology, and the growing number of smart devices pose a problem for the recycling industry: In order to determine the value of a larger amount of PCBs, they are presorted manually and the value is estimated on basis of a randomly selected sample set.
This process is relatively slow and prone to error. There are currently no suitable methods for determining the exact value of large quantities of PCBs that offer a high degree of accuracy as well as high throughput. 

The aim of the “PCBcycle” project is to develop a solution that provides a complete online evaluation of the PCBs, enabling them to be sorted subsequently. 
The result is a sorting system and process for the automatic sorting of waste printed circuit boards (WPCBs). The estimated value of the WPCBs can then be used to decide whether manual sorting and removal of valuable components is economically viable.

How it works

© Fraunhofer IIS
The system is capable of identifying individual components in a PCB. This enables raw materials to be evaluated automatically.

The system takes dual energy X-rays (DE-XRT) of the PCBs on a conveyor belt. These images are preprocessed and fed into a deep neural network that identifies the different components, such as ICs, BGAs/PGAs, tantalum capacitors, connectors, etc. The system calculates the value of each component in a PCB using a mathematical model based on characteristics such as component type and size. This valuation model can be adapted to the user’s individual needs. 

In contrast to other approaches such as visual or NIR imaging, these X-rays can show PCB components regardless of the positioning of the PCB, i.e., they can detect components on the front and back of the PCB at the same time. In addition, DE-XRT is a more robust imaging technology in terms of application in dirty or dusty environments.

© Fraunhofer IIS

Fully automated evaluation of recyclable materials

To create a system for the automated valuation of PCBs, several self-contained process steps need to be connected.

  1. PCBs are measured on a conveyor system using a dual energy X-ray system (DE-XRT)
  2. The relevant components are detected using machine learning and these data are used for the subsequent valuation. 
  3. The weight of the chemical elements contained in the detected components is checked against a customizable price model/market model. This enables the financial countervalue of the PCBs to be assessed.
© Fraunhofer IIS/EZRT
Specific information can be extracted from the comprehensive data using intelligent software-based methods.

A representative database for training and evaluating models is essential for the purposes of machine learning. The first step is therefore to create such a machine learning database from real PCB samples. The next steps involve taking the X-rays, followed by preparing and retrieving representative samples of different PCB component classes. The chemical composition of these samples is then analyzed (e.g., inductively coupled plasma optical emission spectrometry — ICP-OES) to determine the different concentrations of the valuable metals. This information provides the basis for calibrating the model that predicts the actual recycling value of the PCB.