AI-Based Optimization of Polymer Composite Recycling Processes

A three-day, hands-on workshop where you learn to apply AI across the composite-recycling chain—imaging-based sorting, surrogate-model reprocessing optimization, physics-informed AI and digital twins—culminating in pilot-ready tools, an LCA-lite CO₂e calculator, and DPP-compliant traceability.

• Build an end-to-end AI pipeline (sorting → reprocessing).

• Create datasets/imaging; train compact classifiers with active learning & uncertainty.

• Optimize via surrogates & multi-objective trade-offs; set KPIs/guardrails.

• Deploy & govern: edge constraints, digital twin, LCA-lite & DPP, MLOps.

📅 Day 1 – AI-Driven Sorting & Identification

• Mixed composite streams: dataset design, labeling standards, QA criteria
• Imaging stacks: RGB/NIR/HSI selection, illumination & capture setup
• ML pipelines: baselines → compact deep models; active learning & uncertainty
• Edge/line deployment: latency budgets, fail-safes, reject policies
• Hands-on: Build and benchmark a small classifier; set confidence thresholds

📅 Day 2 – Reprocessing Optimization

• Recycling routes & KPIs: yield, fiber-strength retention, resin recovery, energy/cost/CO₂e
• Data-to-decision: tidy historical runs, guardrails, acceptance criteria
• Surrogate modeling & multi-objective optimization (Pareto trade-offs)
• Interpretability & sensitivity (SHAP) to identify governing levers
• Hands-on: Train a surrogate, generate a Pareto set, issue a pilot run card

📅 Day 3 – Physics-Informed AI, Digital Twins & Traceability

• Physics-guided learning (PINNs/data fusion) for sparse, noisy regimes
• Line-level digital twin: soft sensing, guard-railed control, what-if analysis
• Sustainability accounting (LCA-lite) for cost and CO₂e per kg output
• Digital Product Passport readiness: data schema, QR payload, compliance
• MLOps & governance: versioning, drift, retraining cadence, audit trail
• Hands-on: Twin simulations under quality/CO₂e constraints; cost/CO₂e calculator; draft DPP payload

• Process & manufacturing engineers (thermoset/thermoplastic composites)

• Recycling R&D/operations teams (mechanical, chemical, pyro/solvolysis)

• QA/QC analysts & lab technicians (imaging, spectroscopy, materials testing)

• AI/ML engineers & data scientists (vision, edge/line deployment, optimization)

• Sustainability/LCA analysts; EHS/ESG & DPP/traceability leads

• Plant/operations managers; continuous-improvement professionals

• Senior UG/PG students in materials, chemical/mechanical, polymer, or AI/DS

• Build & calibrate an edge classifier with confidence thresholds; add active learning/uncertainty.

• Design datasets and imaging setups (RGB/NIR/HSI) with QA criteria.

• Define recycling KPIs/guardrails; tidy historical runs for data-to-decision.

• Train surrogates; run multi-objective (Pareto) optimization; issue a pilot run card.

• Use SHAP/sensitivity to identify governing levers.

• Apply physics-guided AI (PINNs/data fusion) for sparse/noisy regimes.

• Configure a line-level digital twin for soft sensing, guard-railed control, and what-ifs.

• Do LCA-lite (cost & CO₂e per kg) and prepare DPP-ready data schema/payload.

• Set up MLOps governance (versioning, drift, retraining, audit trail).