A Clean Slate for White Goods

When major household appliances break down, consumers often face the question: Which is more expensive – buying a new one or repairing it? A few years ago, Stiftung Warentest investigated exactly that. Their conclusion: Repairing a washing machine makes ecological sense, but it’s only marginally worthwhile financially. The testers assumed that a washing machine requires an average of two repairs over its 15-year lifespan, with the first occurring after about eight years. The repairs would typically cost so much that, in the long run, buying a new one would be only slightly more expensive than the repairs. 

However, it is also clear that using washing machines for as long as possible and repairing them when necessary, helps protect the environment. This is because the appliances are energy- and resource-intensive to manufacture. They consist mainly of steel, plastics, glass, and copper – materials that can be recycled and used to manufacture new electrical appliances. The proper disposal of old appliances at recycling centers, through retailers, or by the manufacturers themselves is therefore also a key environmental factor. According to the Federal Ministry for the Environment, Climate Action, Nature Conservation and Nuclear Safety (BMUKN), 1.8 million tons of electronic waste are generated in Germany every year. Fifty percent of this is disposed of improperly, in household trash or through illegal channels, the ministry reports. In a joint campaign with the German Environment Agency, the BMUKN therefore aims to encourage consumers to ensure that old electrical appliances undergo a proper recycling process. 

Recycling old electrical appliances with AI

As part of the KIKERP project, researchers at Fraunhofer IPK are investigating how household appliances can be reused at the end of their service life. Their goal is to use artificial intelligence to collect information about old appliances and determine, based on parameters such as model and condition, whether end-of-life refrigerators, washing machines, or other white goods should be refurbished or recycled. In recycling, the appliances are disassembled into their components and the raw materials are returned to the production cycle. In refurbishment, functional or easily repairable appliances are reconditioned and brought back to market.

Together with YES Ecosystems Technology GmbH and HaKiGo GmbH, they are developing a multimodal assistance system for this purpose, based on modern AI methods for understanding image and video data. Through a dialogue-based application, the AI is fed data until a classification is available for the proper refurbishment, reuse, and recycling of the examined electrical appliance. The system is being integrated into a cloud-based management platform and is intended to help both end consumers and manufacturers determine the quality and price of household appliances in the future, enabling them to sell used and repaired products on e-commerce portals. In this way, the project partners aim to support the sustainable management of old electrical appliances in the spirit of a true circular economy and reduce waste.

Rating on a scale of one to three

Using an app developed as part of the project, users can enter information such as the brand, product type, color, and model number of a household appliance via a dialog on their smartphone or tablet and then photograph it from various angles and perspectives to document any defects, such as scratches. Based on this, the AI visually assesses the quality of the item on a scale from one (okay/adequate) to three (very good). This allows for the derivation of further actions, as well as the determination of parameters such as price, condition, etc. 

To develop the AI, the researchers are using both manufacturer data and synthetically generated training data. In doing so, they are also investigating whether synthetic data can be used to train the system for visual inspection. »Our goal is to test and qualify more than 5,000 household appliances using AI by the end of the project and to achieve a detection rate of over 97 percent,« explains Vivek Chavan, project manager at Fraunhofer IPK.
 

Cloud-based platform

In addition to the AI assistant, Chavan and his team are developing a data management system that carefully sorts the data sets obtained during the process using an AI-based evaluation methodology and selects them for future training processes. Based on the data management system, the team aims to determine when and with which data the AI needs to be retrained to achieve the most significant improvement possible. In addition, repetitive or time-consuming and error-prone manual tasks are to be automated and taken over, for example, by software bots. »With our cloud-based management platform, we are designing a process landscape to remanufacture and reuse large household appliances and implementing it as an application demonstrator. Looking ahead, we are planning to synchronize it with e-commerce portals for the sale of the analyzed products,« says Chavan. Ultimately, the project is intended to help reduce the carbon footprint of the manufacture and use of household appliances in Germany, Europe, and other markets. Furthermore, it aims to promote scientific openness and the easy adoption of AI technologies by industry.