The Silicone Contamination Problem

Silicone rubber (polydimethylsiloxane, PDMS) is ubiquitous in modern manufacturing. It appears in consumer electronics, automotive components, medical devices, kitchenware, and — critically for recycling operations — cable insulation and connectors. When end-of-life electronics and cable waste enters the recycling stream, silicone fragments contaminate both metal and plastic fractions.

The economic impact is severe. When PET flake containing more than 0.1% silicone is processed into food-contact pellets, the resulting material fails mechanical testing and must be downgraded to non-food applications — a value loss of 20–35%. In cable copper granulation, silicone contamination on copper particles creates processing problems in electrolytic refining and reduces conductor conductivity. The problem is growing: global silicone demand is projected to reach 5.5 million tonnes by 2028, driven by the electronics and EV industries.

Why Traditional Sorting Methods Fail

Every conventional recycling sorting technology has a fundamental blind spot when it comes to silicone rubber:

In cable recycling specifically, silicone is found as the primary insulation on high-temperature cables (silicone wire), connector gaskets, and protective boots. Without effective silicone removal, the plastic granulate cannot meet market specifications for recycled polymers. Our cable separation line integrates AI sorting as a final purification step for this reason.

AI-Powered NIR Sorting: How It Works

Near-infrared (NIR) spectroscopy is the only reliable method for detecting silicone rubber in mixed material streams. Each polymer absorbs NIR light at unique wavelengths — silicone's characteristic absorption peaks appear at 4,500–5,000 nm and 8,000–10,000 nm, which are chemically distinct from PET, PE, PP, PVC, and rubber. AI-enhanced sorting systems combine NIR spectroscopy with deep neural networks to deliver unprecedented accuracy:

Sorting Accuracy and Performance Metrics

Performance data from field installations demonstrates the clear superiority of AI-enhanced NIR sorting over conventional spectroscopic methods. The deep learning component addresses one of the fundamental limitations of rule-based spectroscopy: particles that are partially occluded, abnormally shaped, or have surface contamination produce distorted spectra that rule-based systems misclassify.

For plastic recycling operations targeting food-contact compliance, the AI sorter typically achieves final silicone contamination below 0.05% — well below the 0.1% threshold for FDA and EFSA food-grade certification. For more details on achieving food-grade plastic quality, see our comprehensive plastic recycling guide.

Applications in Cable and Plastic Recycling

AI-powered silicone sorting integrates into recycling operations at two critical points:

• Cable recycling lines: After granulation and air separation, the copper and plastic fractions are each passed through the AI sorter. Silicone contamination on copper reduces electrolytic refining efficiency; silicone in the plastic granulate limits end-market applications. LVKESORT's cable separation line incorporates AI sorting for final purification of both output streams.
• PET/PE/PP recycling lines: Silicone rubber from appliance casings, cookware, and electronics enclosures enters the plastic stream during grinding. The AI sorter removes silicone before the washing stage, preventing contamination of the final recycled pellets.
• WEEE recycling: Waste electrical and electronic equipment contains numerous silicone components (keypads, seals, thermal pads). AI sorting enables recovery of clean plastic fractions from complex WEEE streams.

Frequently Asked Questions

Why is silicone rubber difficult to sort from plastics?

Silicone rubber presents unique sorting challenges that traditional methods cannot overcome. First, silicone has a density of 1.1–1.3 g/cm³, which overlaps significantly with common plastics like PET (1.38 g/cm³) and PP (0.90 g/cm³), making density-based float-sink separation ineffective. Second, silicone rubber is available in virtually any color — transparent, white, black, and every shade in between — so color-based optical sorting fails entirely. Third, silicone often appears in small forms: gaskets, seals, O-rings, and cable jacket fragments as small as 5–10mm. Traditional air classification, magnetic separation, and eddy current systems are all blind to silicone. This is why AI-powered near-infrared (NIR) spectroscopy is now the industry standard for effective silicone removal.

How accurate are AI-powered sorting machines for silicone removal?

Modern AI-powered optical sorters achieve sorting accuracy of 99.0–99.9% for silicone rubber identification, depending on particle size, feed rate, and material complexity. At particle sizes above 15mm, accuracies consistently exceed 99.5%. Below 10mm, accuracy typically drops to 98–99% due to the reduced spectral signature of smaller fragments. The AI component — deep neural networks trained on millions of labeled material spectra — enables the system to distinguish silicone from chemically similar materials (silicone vs. TPV, silicone vs. liquid silicone rubber LSR) that pure spectroscopic analysis alone cannot reliably separate. Throughput ranges from 1 to 8 tonnes/hour depending on particle size and system configuration.