MAGDEBURG, Germany—Engineers at the Fraunhofer Institute for Factory Operation and Automation (IFF) here are developing an automated process for electronics recycling. The goal of the Intelligent Disassembly of Electronics for Remanufacturing and Recycling (iDEAR) project is to create a nondestructive robotic method that will invigorate an advanced circular economy that reduces e-waste.

Worldwide annual e-waste generation could rise to as much as 74 million metric tons by 2030. However, only a small fraction of all electronic devices is currently recycled. More than 80 percent of the e-waste generated ends up in landfills or incinerators, including valuable raw materials such as precious metals and rare earths.

The small percentage of e-waste that undergoes treatment typically gets shredded, while only a limited portion is manually disassembled, cleaned of hazardous substances, broken down mechanically and sorted into different fractions. Manual disassembly entails high costs and is not very effective.

“We intend to revolutionize the disassembly of e‑waste,” says José Saenz, Ph.D., manager of the assistive, service and industrial robots group at Fraunhofer IFF. “Current solutions require substantial engineering and are limited to a particular product group.

“We are pursuing a data-driven methodology so that a wide variety of products, from computers to microwaves to home appliances, can be disassembled in real time with little engineering,” explains Saenz.

The engineers are initially focusing on the automated disassembly of computers. However, the system is intended to be upgradeable over time for other types of products, such as dishwashers and washing machines.

The process of identification and condition analysis uses optical sensor systems and 3D cameras with AI-powered algorithms to detect component types and locations, examine geometries and surfaces, assess the condition of fasteners and detect anomalies.

“Optical metrology helps scan labels and sort different parts, such as screws,” says Saenz. “Previously trained machine learning algorithms and AI interpret the image data, and enable the identification and classification of materials, plastics and components in real time based on sensor and spectral data.”

For instance, AI detects whether a screw is concealed or rusted. All the data is stored in a digital disassembly twin.

In the next step, Saenz and his colleagues define the disassembly sequence so that their software can determine whether to execute a complete disassembly or only focus on the recovery of specific, valuable components.

However, glued or otherwise mated components hinder nondestructive disassembly. Rusty or stripped screws, or deformed components, are not ideal either.

The disassembly process starts based on this high-level information. A robot receives a series of instructions and operations to complete, such as “remove two screws on the left of the housing.”

The skills specified in the disassembly sequences include robot actions, such as screwing, lifting, cutting, extracting, localizing, repositioning, releasing, moving levers, bending, breaking and cutting wires. According to Saenz, the disassembly robot can perform all of those tasks completely autonomously.