R&D Focus
The Alchemy for Textile Industry Automated Textile Sorter
A total of 206 brands worldwide, including H&M and PUMA, have promised to increase by 25% their usage of recycled polyester (rPET) by 2025, and to use 100% recycled or other sustainable textile products before 2030. However, the major challenge lies in the source of rPET, a new gold that the textile industry is pursuing.
While rPET coming from recycled plastic bottles is rather limited and producing rPET from textile products is difficult and expensive, ITRI’s smart sorting system acts as an alchemist to produce this gold and help textile manufacturers fulfill their promises.
ITRI researchers found that existing clothes sorting systems have been struggling with textile blends and the PET therein cannot be easily sorted from other fibers. Moreover, any sorting accuracy below 90% is not conducive to textile factory systems.
Integrated with a conveyor belt, ITRI’s sorting solution has a throughput of one article of clothing per second.
To recover higher-purity PET from textile blends, ITRI created an automated textile sorting system that can identify and separate textile blends based on optical analysis of fabric compositions. Besides its ultra-high sorting speed of one article of clothing per second, it has a pioneering accuracy in identifying post-consumer textiles of over 96%. Integrated with an industrial conveyor belt, this system can achieve a throughput of 900 kg of textiles per hour, regardless of color, material, and texture.
So how does this technology achieve a high accuracy in sorting textiles? According to Ryan Chu, ITRI’s Technical Manager, ITRI integrated near-infrared (NIR) excitation and auto-focus Raman spectrum technology, and trained the AI model with spectral property data of different fabrics. The sorting purity for pure PET reaches more than 99%, and the identification accuracy of PET/cotton blends is above 95%. The sorted PET can then be recovered and processed effectively into rPET. The sorting technology is also capable of sorting PET-based fabrics that contain only a small proportion of other fibers. For example, it can sort PET blends with a minimum content of 5% cotton and PET blends with a minimum content of 1% Lycra. See below for the benchmark chart of ITRI’s sorting system.
This chart compares ITRI’s sorting solution to the technological benchmark in textile identification and separation.
Curious about what happens to the clothes put into the sorting system? When cast onto the moving conveyor (as in a factory), the wrinkly fabric is tracked by a laser profiler, so that the system can receive the best Raman signal regardless of the fabric’s topography. After the fabric is hit by 1,064 nm NIR excitation beams, it responds with distinctive peaks, which allows the AI model to analyze the fabric’s composition. The fabric is then categorized into, for example, 100% PET, 85% PET/15% cotton blend, or 99% PET/1% Lycra, and sent to different piles with air jets.
Once sorted, these clothes may be converted to reliable inputs for apparel manufacturing industry players to realize the textile-to-textile recycling that was once a castle in the air.