How much material was kept out of landfill or incineration?

The facility reported that its unrecoverable share fell from as much as 50% to 30%. This represents a 20 percentage-point reduction, although the source does not provide the material's total weight over a defined reporting period.

Will this equipment fit an existing sorting line?

The documented scanner measured 5 by 2 meters and operated with a series of conveyor belts. Fit still depends on infeed spacing, downstream lanes, reject handling, controls, and available floor area, all of which should be checked during site assessment mapping.

What should a textile recycler measure during a pilot?

Start with the current unrecoverable share, classification accuracy by fiber blend, throughput, false-sort patterns, downtime, and downstream acceptance. The case benchmark is informative, but each plant needs results from its own feedstock and recovery specifications.

Does AI sorting eliminate the need for operators?

The documented workflow still used workers to load textiles onto conveyor belts. The evidence supports faster and more precise classification, but it does not report a headcount reduction, so staffing claims would go beyond the source.

a commercial textile-recycling facility in eastern China

AI Textile Sorting Cuts Unrecoverable Material to 30%

Q: Can an AI textile sorting system distinguish close fiber blends?

That was the central advantage reported in this case. Facility staff said people could not accurately distinguish 80% polyester from 90%, while the scanner read material composition in less than one second and applied customer-defined thresholds.

See how an AI textile sorting system cut unrecoverable material from up to 50% to 30% at a commercial recycling facility in eastern China.

30%
unrecoverable after installation: 20 pts
reduction in unrecoverable share: <1 sec
fiber reading per item: 2 tons/hr
reported machine capacity

Based on a documented real-world deployment. Figures are from public reporting; the organization is not named.

The Fiber Blend Bottleneck

Incoming garments could look alike while carrying materially different fiber blends. Human sorters could identify broad categories, but they could not reliably distinguish close compositions, such as an item containing 80% polyester from one containing 90%.

That uncertainty narrowed the recoverable stream. Before the AI textile sorting system was installed, as much as 50% of processed material was classified as unrecyclable and sent to landfill or incineration.

The constraint was not simply sorting speed. The facility needed composition data precise enough to apply customer-defined recovery thresholds and send suitable material into the correct downstream process.

Close fiber blends were difficult to separate accurately by hand.
Variable composition complicated recovery decisions.
Misclassification pushed potentially useful feedstock into the reject stream.

Composition Scanning at Conveyor Speed

The facility installed the system in 2025. Workers placed stacks of used textiles onto conveyor belts, which carried each item through a 5-by-2-meter scanner and displayed the composition reading live.

The scanner analyzed each item in less than one second against customer-defined benchmarks. Conveyor lanes then directed qualifying textiles toward nylon or polyester recycling areas, while material below the required threshold moved to the reject stream.

The Associated Press account does not describe the vendor-selection process, a phased rollout, formal staff training, or the service agreement. The documented evidence supports the operating sequence below, not invented deployment details.

Load used garments onto the conveyor infeed.
Read fiber composition with the AI scanner.
Compare the reading with the customer's recovery benchmark.
Route qualifying nylon and polyester material for recycling.
Divert material below the benchmark to the unrecoverable stream.

A Smaller Reject Stream and Faster Classification

After installation, the reported share of processed textiles deemed unrecoverable fell to 30%, down from as much as 50%. That is a 20 percentage-point reduction in material sent mainly to landfill or incineration.

The system reportedly sorted 100 kilograms of clothing in two to three minutes, a task that took one worker around four hours. Reported machine capacity reached two tons per hour, while two people would need two days to process the same quantity and would do so with lower accuracy.

These figures were reported by facility management to the Associated Press. They are useful operating evidence, but the article does not present an independent audit, financial return calculation, or long-term reliability record.

From Demonstrated Capability to a Deployable US Project

This was not a Service Robot Co. deployment. It is a documented real-world example showing what composition-aware automation can change when manual identification is the recovery bottleneck.

For a comparable US project, Service Robot Co. would serve as the vendor neutral robot integrator. We assess the process, select suitable equipment across manufacturers, arrange financing, handle robot deployment and integration, train the operating team, and support the installed units through a nationwide US engineer network. One vendor carries the whole lifecycle.

A disciplined site assessment would establish the incoming material profile, recovery thresholds, conveyor interfaces, reject handling, and downstream capacity before equipment selection. A robot pilot program could then test classification performance on the facility's own feedstock and inform a phased deployment with no shutdown.

Commercial structures can include purchase, robot as a service, or monthly payment programs when available for the selected equipment. The buying model follows the operating case, with maintenance and service responsibilities defined before go-live.

Frequently asked questions