Revolutionize Battery Recycling

Project description

The disposal of electronic devices can cause battery fires, which lead to health hazards and environmental damage.

We aim to prevent this by detecting whether devices contain (removable) batteries before they are processed.

Context

E-waste accounts for a large portion of potentially recyclable waste that is often diverted to landfills, causing a lot of pollution and with significant resources wasted. Batteries are often present in e-waste and are part of the reason for this low recycling rate. Due to their very reactive nature, they have a widespread impact on recycling operations where they can easily short-circuit or puncture and combust. Fires, exposure to fumes and other hazards are frequent, threatening the health and lives of the people in the vicinity of the facilities. The environment is harmed directly due to the pollution generated but also indirectly through the disruption of recycling operations. Finally, these accidents can lead to heavy losses in terms of machinery, productivity and revenue, but they can also incur heavier insurance costs.

Batteries are often attached to the e-waste and may be challenging to detect at scale. The process is heavily reliant on difficult manual work which is prone to mistakes and injury. Deep learning has been shown to have higher success rates in image recognition tasks than humans and can therefore improve this process and make it far more economically viable for companies to implement, helping solve our current e-waste problem.

Results

We are successfully able to predict battery removability using device dimensions and images. Using similar techniques, we have created a solution to detect whether monitors contain mercury. This allows factories to safely separate and recycle these devices.

In a real world scenario, devices may be stacked and may or may not contain batteries. In the next phase of the project, X-ray images will be used to improve the safety and efficiency of the recycling process.