How AI can support the Circular Economy

Facebook IconFacebook IconFacebook Icon
Arnaud De Decker
Mar 19, 2024
On Wednesday 16th February 2022, a webinar was held on the technological challenges of Artificial Intelligence in relation to the circular economy. This event was organized by Digital Wallonia4AI and brought together several companies including Sagacify to share their experiences.

We live in a consumer society that has long been based only on a linear economic model. However, the planet's resources are running out and people have realized that this way of consuming is not sustainable in the long term. This is why society is now moving towards a more sustainable economy based on the circular economic model. This model aims to extend the life cycle of a product through refusing, reducing, reusing, repurposing, and recycling. In this context, one of the approaches to facilitate the development of the circular economy is artificial intelligence (AI). Indeed, concrete applications have proven that AI can help reduce waste and improve its sorting among others.

The case of Besix: Waste management optimisation

Besix called on Sagacify to solve a problem concerning the recycling of waste disposal from its containers. This Belgian construction company regularly finds itself with incorrectly sorted containers, which costs them money and prevents them from recycling in a 100% efficient way. Sagacify's solution to this problem was to constantly monitor the waste in the containers using on-site cameras and an AI algorithm to detect sorting errors and alert Besix directly so that they can correct it.

Detecting sorting errors in containers can be solved by building an object detection or image classification algorithm. However, because of the large diversity of shapes and textures of contaminants to be detected in containers, the quantity of images to annotate is potentially very important in order to build a powerful model. To alleviate this issue and maximize the model performance with a small amount of data, Sagacify adopted a particular approach called self-supervised learning, which is a family of techniques for converting an unsupervised learning problem into a supervised one by creating surrogate labels from the unlabeled dataset. In this project, we used the SIM-CLR algorithm, to improve the feature representation quality of our model. It consists of two main steps; the unsupervised training step and the supervised classification step. The first step aims to increase the model's knowledge of the dataset. To do this, it performs small tasks on the data such as reconstructing images. After that, the model is asked to do the targeted task of classifying the data. More information on self-supervised learning can be found here if you are interested in this topic.

Sagacify's results from this approach should increase the amount of waste recycled and reduce Besix's costs.

If you want to learn more about this solution, listen to the podcast here.

The case of Resideo: Production chain waste reduction

Resideo is a company that creates products and systems to build smart-homes. However, it faces a waste and loss problem. Indeed, when a final product has a defect, the company is forced to disassemble it before reworking the raw material, which is very expensive. In the worst case, products can be thrown away. Early identification of defects on production lines, through visual inspection, would reduce the amount of necessary rework on defectuous parts or customer returns.

Sagacify has therefore designed an automatic defect detection system based on AI to spot these flaws. Technically speaking, they are presented to cameras by a robot installed on site and a set of computer vision models decide whether the part should be placed in a rejected bin or left on the production line. That way, Sagacify has made it possible to optimize the production chain, reduce costs and decrease the amount of waste generated by Resideo.

If you want to learn more about this solution, listen to the podcast here.

Bottom Line

These are just two examples of how AI can be used as a tool to accelerate the transition to the circular economy. The potential of AI for a greener world is therefore huge. Indeed, sustainable solutions are almost infinite and can be adapted to many situations since AI-based models are able to learn continuously. AI is encouraging companies to take the plunge and be more responsible and sustainable. After all, who says that technology and the environment do not go together?

At Sagacify we already know that technology is helping to make the world a more responsible place. We are developing Sagavision, a solution that will help facilitate and optimize the deployment of quality control models based on the vision in the industrial sector. For more information on Sagacify’s capabilities and use cases in visual inspection systems, don’t hesitate to schedule a call with Arnaud De Decker.

For more similar content, check out the Digital Wallonia and follow them on LinkedIn here so you don't miss any news.