Sorting out: Making Recycling Easier with Technology That Identifies Fibres and Fabrics

What drew you to this field? Which opportunities did you see that made you choose to operate in this market?
Lieve Vanrusselt: In 2020, it was estimated that worldwide 92 000 000 tonnes of textile waste is created each year which means that the equivalent of a rubbish truck full of clothes ends up on landfill sites every second. This figure is expected to grow by a further 60% by 2030. From this, today, only less than 1 percent is fibre-to-fibre recycled. Fibre-to-fibre recycling of post-consumer textiles is now one of the key strategic components to support the transition towards a circular fashion industry and increasing attention from governments and eco-awareness in general is pushing for increased recycling rates. This development is expected to lead to an increased demand for post-consumer textiles collection, sorting and recycling infrastructure across the EU and worldwide. This in turn will require an increased investment into infrastructure that can sort and prepare textiles for reuse and recycling.

The textile sorting system today relies heavily on manual processes which is likely to remain the first step for sorting any PCT with rewearable content but is not the optimal solution for recycling which requires identification of the specific fibre types. To feed the increasing demand of feedstock to these recycling markets, this purely manual sorting requires to be followed by semi-automated or automated sorting of the non-rewearable fraction by fibre type and also often colour. Due to rising consumption, increase in collection and the marked decrease in overall quality of the garments, the volume of this low value fraction is expected to grow substantially for many years to come and will need to be sorted. The sheer scale of these issues presents a large number of opportunities for solutions – such as ours.

For plastics, manual sorting is much less economical, but we strongly believe the opportunities here lie with the sorting of bulk waste (which is much heavier) that cannot be processed by the automatic sorting lines due to its size. Many of our PlasTell customers are busy in the field of pre-sorting while disassembling items such as (larger) household items for example.

Furthermore, both our technologies find applications in quality control in manufacturing and various other processes and services and are also being used for educational purposes. Our more compact material sensing module can also be integrated with small-scale automatic sorting machines; we have been working with a number of customers on integrating these modules for sorting fabrics and plastics.

Please take us through the process — explain the science of it in layman's language.
Lieve Vanrusselt: Our existing identification products are near-infrared (NIR) scanners that measure how different materials interact with infrared light, acquiring their infrared signatures (spectra). The spectra are then instantly processed by our industry-leading material identification ML algorithms that determine the composition of the material and, in the case of textiles, display this as the weight percentages of the detected textile components.

The infrared sensing and automatic material recognition core technologies are very scalable and currently deployed on our 4 device types (desktop, bench, handheld and sensing / OEM module), each with different footprints and suited for different use cases in multiple parts of the textiles and plastics supply chain, providing user-optimised solutions for our customers’ identification needs.

Our Cloud system and mobile app enable the saving of data, building of databases as well as providing data analytics to our customers. Particularly for textiles, this will play a key role establishing traceability and evidence of sorting but also in understanding the different waste streams and material flows.

Your website has two sections — one to identify plastics and the other textiles. Do your instruments detect synthetic blends?
Lieve Vanrusselt: Our textile identification technology supports 9 pure materials (cotton, polyester, viscose, polyamide, acrylic, wool, elastane, silk, acetate) and 13 different two-component blends with typical accuracy levels of +/- 5% for pures and +/-10% for blends. It works for all weaves and colours including black (except carbon black). Of the blends, our instruments can detect various synthetic blends such as polycottons, cotton blends with nylon and acrylic, wool blends with polyester, nylon and acrylic, viscose blends with polyester.

In blends with elastane, the presence of elastane can be detected down to 5% for some of the blends (ex. polyester-elastane). However, in general, at low percentages, some materials can look very similar (for example elastane and nylon) and our machine will say “95% cotton 5% contaminant” rather than “95% cotton 5% nylon”. Given that it is already highly challenging to recognise two-component blends, it means that in the case of 3- and more-component blends, our machine will most likely the two major components.

The performance and accuracy depend on a number of factors, such as the correct measurement technique, and these numbers represent the typical accuracy rather than guaranteed accuracy for every single sample. It is important to note that the pure and blends that are supported by our technology today have been identified as the most present in a post-consumer textile audit carried out by Refashion (France) (Refashion.fr - Refashion for a 100% circular textile industry) using our NIR scanners in 2021. However, we are continuously working on improving the identification accuracy and on expanding the detection capability of our technology to more blends.

For plastics identification, we are planning to implement the identification of blends in the future but for the moment we support 25 different plastics and have made it possible for customers to add new materials to the plastics library so they can identify specific plastics and blends.

What is the feedback like from your customers? What are the changes/modifications that they have demanded since you started?
Lieve Vanrusselt: Our primary customers are companies involved in the sorting and recycling of textiles and plastics. Our machines enable them to sort waste by material for all kinds of recycling processes, as well as perform quality control on feedstock or finished materials. There are a few secondary customer types, such as NGOs or educational organisations working in the field of circular economy that buy our machines for research, demonstration, and educational use. 

For example, our instruments were bought by Fashion For Good (Netherlands) and Refashion (France) to carry out their waste analysis surveys in Europe (2021) and India (still ongoing). We have recently started to collaborate with a few eco-minded brands since we believe that building awareness for textile recycling with the end consumer can only happen through a strong engagement and conviction by the brands; take-back schemes need to be encouraged and expanded across the fashion retail sector and need to be more efficiently organised, with a possible first ‘pre-sort’ in-store.

Feedback on both our technologies has been good and useful. From the start, we have worked closely with our first customers to develop our technology to their needs. This brought us to develop our ‘desktop’ model into our ‘bench’ model, which is better suited and more robust for continuous sorting and is now the preferred format being used in textile sorting halls. Based on this early feedback, our app (for iOS and Android) was written to configure the instrument as well as to save and analyse data our customers need for waste or feedstock analysis to go into the next process. 

Via the app, the colour of the instrument’s LED lights can be programmed to sort for the desired categories, and this allows for easy and simple sorting operation. For example, if you were sorting cotton for recycling and wanted to accept all samples with at least 90% cotton, you could configure it to show green light (=’accept’) for samples with >=90% cotton and red (=’reject’) otherwise.

Our communications library was set up to let you connect to our machines and sensing modules over a variety of interfaces (such as Bluetooth) and then extract the measurement results in real-time. For instance, this allows you to report the result to a robot or your own cloud system which will then take action based on the result. The library is open-source and free to use with our machines. While with our app, you can save and export measurements, sometimes you need more advanced access than this; for instance, if you would like to perform the export in an automated way. For this reason, we have created a data access API (application programming interface), currently still free of charge, which allows you to access data using scripts and obtain detailed insights into your processes.

Useful feedback has also come from our close collaboration on several textile waste analysis projects with some of these still ongoing. Our technology has played a key part in the post-consumer textile waste surveys done by Fashion For Good and Refashion. FFG published the ‘Sorting for Circularity in Europe’ project report (Sep 2022) along with their Sorters Handbook: how to conduct a sorting analysis using hand-held near infrared scanning technology. Both reports are exceptional, painting an accurate picture of the composition of our post-consumer textile waste today and providing a realistic insight into the opportunity and capability of textile recycling looking towards 2030. Through these collaborations, we have enjoyed a lot of exposure and got the opportunity to develop our tools to today’s industry gold standard for rapid textile analysis.

Moreover, we have gone for CE and UKCA Compliance.