Specific methods and technical skills are being developed in the iDEAR research project to enable the disassembly of end-of-life electronics by robots. The biggest challenge is the wide variety of end-of-life electronics model types and conditions based on their life cycles. Standard automated operations are especially efficient when they can be executed repetitively over a longer period, when the components and assemblies are known and when these are in a defined condition. Automating the disassembly of end-of-life electronics consequently requires new approaches: Every device must be analyzed separately. Disassembly processes must be planned based on the condition of individual products. To achieve this, robots must be able to execute complex actions, sometimes simultaneously. Until now, human experience and creativity was essential to this job. Manual processes are often inefficient and expensive, though.

Rising resource prices and diminishing availability of critical materials will make it increasingly interesting for manufacturers to integrate recycling actions, such as product recycling and remanufacturing, in their processes as well. Legal regulations, such as the right to repair, are reinforcing this trend with appropriate standards. Automated solutions, such as robot-guided disassembly, are one option to recover valuable materials efficiently and cost-effectively and simultaneously to counteract the shortage of skilled labor.

Industry has initial solutions for the automated disassembly of specific products. These are highly specialized, though, and not applicable to a wider range of products. The substantial engineering required is a particular challenge for automated solutions designed for flexibility. Reducing the engineering required by using data models end-to-end and developing new AI methods and robotic skills are other research priorities in the iDEAR project.

“In the iDEAR research project, we are pursuing a data-driven methodology so that as many different products as possible can be disassembled with little engineering effort,” says Dr. José Saenz, project manager at Fraunhofer IFF.

The research scientists in the project initially concentrated on the disassembly of computers. They have a high material value for their size and constitute a significant share of electronic waste. Various technological developments were tested and refined on this model product.

The first step is the individual analysis of the device. Important fasteners, such as screws and rivets, are scanned and evaluated by a camera. Ultimately, this analysis digitally scans and saves the location and condition of individual screws and components for each product.

The scanned data is used together with potentially available product data (manuals and experiences with prior disassembly) to generate the disassembly sequence, which specifies the sequence of steps and the tools necessary for disassembly. This sequence is used to generate the sequence of actions for the robots to execute. Here separately developed robotic skills (e.g., unscrewing and cover removal) are interconnected for complex processes. The execution of each complete subprocess from picking the tool needed for disassembly to unscrewing screws to placing the component removed down is automated.

New, adaptive solution strategies were developed for more complex robot handling, such as removal of the motherboard from a computer. Among other things, reinforcement learning and imitation learning approaches are being tested to execute complex robot actions.

Such a solution strategy evaluates the current state in every step and infers the optimal action from this, for instance, by adjusting the robot’s movement. This way, whole components can be extracted for reuse or disassembled down to its separate materials.

An iterative process of analysis, sequence modification and robot disassembly is generated for the disassembly depth desired. The Fraunhofer IFF research scientists developed new assessment methods to identify the economically and environmentally effective disassembly depth, which, for instance, incorporate current prices of secondary materials or even product valuation based on the analysis of it.

A data platform that collects and supplies product and equipment data in the form of a standardized digital twin (using so-called asset administration shells) is being built to be able to access product data continuously and consistently in all subprocesses. This “disassembly hub” is the knowledge base that enables subsequent use of data collected during the disassembly process and inclusion of experiential knowledge about the disassembly of types of products. This makes it possible to optimize other processes with identical product types based on completed disassembly actions. The methods developed are also intended to be transferrable to other devices and contexts. The “disassembly hub” can serve as a source of data for individual automated disassembly processes.

Pilot test systems and demonstrators developed to test and demonstrate the methods and technologies developed are currently being used to test and refine the analysis and disassembly of products and the interconnection of these processes by data technology.

“We integrated the solution strategies learned in an industrial robot controller, for instance. Methods of machine learning were used for subtasks that are hard to perform with analytical methods,” says Saenz, explaining the approach.

The tests with the separate demonstrators that identify and analyze computers, open the housing and remove the motherboard from the housing were so successful that the development and testing of other disassembly processes is already being worked on.

The methods and technical skills developed in the project are easily transferrable to other product types and sizes. Apart from sectors, such as white goods, in which recycling is currently the main focus, there are also product types, such as jet engines, for which disassembly is a key process in product recycling and remanufacturing.

Once the technology is mature enough, the iDEAR solutions developed will be tested in the real-world environment at industrial companies and refined. The experiences will enter into other research studies and the transfer to industrial solutions.

Dr. José Saenz is confident: “We can make an important contribution with this project so that fewer materials are wasted in the future and electronics are recycled faster, more easily and more cost-effectively.”