AI plus solvent-based recycling could close the plastics loop
A unified path to circular plastics: pair solvent-based recovery with AI-guided sorting and system optimization. Build hybrid networks that cut emissions, cost, and impurities.
Solvent-based recycling and AI: a practical path to a circular plastics economy
Only a small fraction of plastic gets a second life. Estimates put global recycling at about 9%, with the rest landfilled, burned, or leaked into the environment. That's a design and systems problem, not a material problem.
A recent review led by Dr. Aurora del Carmen Munguía-López at the University at Buffalo argues for a unified approach: combine solvent-based recycling, data-driven planning, and process systems engineering to close the loop. The takeaway is simple-optimize the full system, not a single technology.
Why plastics still matter
Plastics reduce food waste, lower transport emissions via lightweighting, and enable affordable medical devices. Eliminating them across the board isn't feasible right now, as Dr. Munguía-López notes.
The issue is post-use management. Poor collection, mixed streams, and weak sorting send plastics to landfills and ecosystems, with health risks tied to long-term exposure. Smarter recovery is the lever.
Solvent-based recycling: where it wins
Solvent-based routes selectively dissolve target polymers from mixed or contaminated streams, making materials like multilayer films recoverable. A study co-authored by Dr. Munguía-López found it to be the most cost-effective option for recycling multilayer coffee packaging films.
Trade-offs exist. Emissions can be lower than many advanced thermal routes, but process conditions matter. Cooling-based reforming steps tend to outperform heating-intensive ones on emissions. Solvent recovery, energy integration, and solvent choice drive both cost and footprint.
- Best fit: multilayer packaging, mixed-polymer waste, and contaminated streams where mechanical recycling fails.
- Key levers: closed-loop solvent recovery, heat integration, and contamination management.
- System move: combine solvent-based with conventional mechanical recycling to maximize recovery while controlling emissions.
AI for smarter sorting, planning, and policy testing
AI is turning sorting and planning into a data problem. Models like PlasticNet (University of Wisconsin-Madison) have reached over 87% accuracy for identification, achieving 100% for some resin classes. That precision cuts contamination and raises yield.
Natural language processing can mine literature for process parameters, solvent/polymer compatibility, and kinetics, accelerating design cycles. Beyond the plant, AI can optimize transportation, coordinate across collectors and processors, and simulate the effects of new policies before implementation. As Dr. Munguía-López notes, supply-chain-level models will be essential to meet demand and logistics constraints.
Are biobased plastics better? Only if the full life cycle checks out
Biobased and compostable options look attractive, but scale introduces land, water, and food-system trade-offs. Compostable polymers also need the right facilities, which many regions lack.
Life-cycle assessment is the filter. As Dr. Munguía-López emphasizes, value claims need cradle-to-grave accounting: feedstocks, processing energy, additives, infrastructure, and end-of-life in real collection systems. Until then, biobased adoption will remain selective.
Process systems engineering ties it together
The review calls for systems-level optimization: product design, collection networks, pre-processing, conversion pathways, and markets for outputs. Process systems engineering provides the tools-flowsheet synthesis, multi-objective optimization, stochastic planning, and scenario testing.
Most studies still isolate a single step. Integrating solvent-based recovery with mechanical recycling, and pairing plant-level decisions with supply chain and policy models, is where the gains accumulate.
What researchers and industry can do now
- Design for recovery: reduce multilayer complexity where possible; standardize labels and additives to improve sortability and dissolution selectivity.
- Expand data pipelines: build open datasets for NIR/hyperspectral sorting, solvent-polymer phase behavior, and degradation kinetics to improve AI models.
- Co-optimize plants and networks: couple TEA/LCA with logistics models (collection density, routing, MRF capacity, energy mix).
- Pilot hybrid flowsheets: mechanical pre-sorting followed by solvent recovery for complex fractions; validate solvent recovery rates at scale.
- Instrument for feedback: inline analytics for contamination, solvent carryover, and energy use; close the loop with model-predictive control.
Metrics that matter
- Material yield and purity by polymer (post-wash and post-reprocessing)
- Solvent losses and recovery energy per kg of product
- GHG intensity per kg of recycled resin vs. virgin baseline
- Cost per kg and revenue from product slate (including additives and fillers)
- Logistics intensity: km-ton hauled, route efficiency, contamination at MRF intake
What comes next
Solvent-based recycling is a practical bridge for hard-to-recycle streams. AI boosts accuracy and speed across sorting, process control, and network planning. Biobased options may contribute in specific contexts, but only where full life-cycle performance is verified.
The path forward is integration. As Dr. Munguía-López puts it, pursue holistic approaches and evaluate the pros and cons across the entire life cycle. The focus now: build hybrid systems that are clean, economical, and scalable.
Further reading
Upskill for AI-enabled sustainability
If you're building AI tools for materials, sorting, or supply chains, explore practical training for scientists and engineers: Latest AI courses.
