Residual waste sorting: Milan model shows nearly 28% of mixed waste could still be recovered

Residual waste sorting has long been a side issue in circular economy debates: whatever goes into the black bin, the usual reasoning goes, is already lost to the material loop. A project team in Milan has now tested that assumption in numbers, in a city whose separate collection ranks among Europe’s best. The answer: not quite. A mass-balance model built on data from AMSA, Milan’s waste management operator, and A2A, which runs the city’s incinerator, suggests that an integrated sorting line could recover around 64,000 tonnes of secondary material a year from the roughly 231,000 tonnes of mixed municipal waste Milan generates, before the remainder is sent to incineration.

The researchers also presented their findings on the illuminem platform on 8 November 2026. Below we look at what they examined, how they calculated, which numbers deserve a careful reading, and what the results mean for Hungarian waste management.

In short: According to the Milan model, residual waste sorting with bag opening, trommel screening, magnetic and eddy-current metal separation and NIR optical sorting could recover 27.9% of the 231,000-tonne mixed waste stream, or 64,342 tonnes a year, as secondary material. The model estimates this would avoid about 120,000 tonnes of CO₂ a year and still deliver a net energy saving after deducting the electricity the incinerator would no longer generate.

Milan: a high separate collection rate, yet 231,000 tonnes go to incineration

Milan is a particularly useful case study because there is no missing collection system to build. According to AMSA, the city collects about 638,000 tonnes of municipal waste a year, 63.8% of it separately. The remaining 36.2%, roughly 231,000 tonnes of mixed waste, is currently sent in its entirety to energy recovery, in other words to incineration.

In the researchers’ words, this is not a “build a recycling system from scratch” problem but a marginal-recovery problem within a mature collection system: how much material can still be saved at the very end of the chain?

The question is driven by EU regulation. Directive (EU) 2018/851, amending the Waste Framework Directive, requires preparation for re-use and recycling of municipal waste to reach at least 55% by 2025, 60% by 2030 and 65% by 2035, while Directive (EU) 2018/850, amending the Landfill Directive, caps landfilling at 10% by 2035. Referring to the 10R circularity framework, the researchers frame the goal for mixed waste as using technology to move it one step up the ladder, from R9 (recover) to R8 (recycle).

What is in the black bag?

The first step was to define the composition of the mixed waste stream. Mixed waste from door-to-door collection, street bins and other sources was treated separately; for the latter, about a quarter of the total, no dedicated composition dataset was available, so the door-to-door profile was applied. The weighted composition:

• paper and cardboard: 70,676 t/y (30.6%)
• other: 44,402 t/y (19.2%)
• plastic bottles and packaging: 43,913 t/y (19.0%)
• textiles: 19,376 t/y (8.4%)
• plastic bags: 16,653 t/y (7.2%)
• paper towels and tissues: 12,615 t/y (5.5%)
• food and organics: 11,852 t/y (5.1%)
• metals: 5,656 t/y (2.4%)
• rigid plastic items: 4,557 t/y (2.0%)
• glass: 1,283 t/y (0.6%)

Two things stand out. First, paper and plastic packaging together make up nearly half of the mixed waste, meaning a substantial amount of material slips through the separate collection system. Second, organics account for only 5.1%, a result of Milan’s extensive bio-waste collection. As we will see, this shapes the design of the sorting line decisively.

No Milan-specific particle-size data were available, so the researchers mapped the residual household waste dataset of ADEME, the French environment agency, onto the Milan composition. It is an approximation, but it reflects a well-known pattern: the fine fraction contains more organic and sanitary material, the coarser fractions more paper, plastics and textiles.

No single technology is enough

The project began with a technology review. The researchers started from the European Commission’s BAT reference document for waste treatment, then refined their assumptions with industry data, manufacturer specifications and input from A2A. The technologies examined and the main parameters used in the model:

• bag opener: tears open closed bags, with a 95% opening rate; a preparation step that recovers no material on its own;
• trommel screen: separation by size at 60 mm (at 60 and 200 mm in the three-stream variant);
• overband magnet: ferrous metals at 90% efficiency;
• eddy-current separator: aluminium and other non-ferrous metals at 92.5% efficiency;
• NIR/SWIR spectroscopy with air jets: recognises and ejects paper, plastics, textiles, glass and organics;
• AI-guided robotic arm: detects and picks individual objects;
• manual picking: used as a benchmark.

For metals, the model treated ferrous and non-ferrous fractions separately: based on regional data from ARPA Lombardia, the mixed metal fraction was assumed to be 60.32% iron, 11.11% aluminium and 28.57% other metals. This matters, because treating all “metals” as one recoverable category would overstate the role of any single technology.

Optical sorting was tested in four configurations (basic NIR, basic SWIR, hyperspectral NIR and hyperspectral SWIR). The more advanced set-ups performed better in isolation, but the reference plant uses basic NIR for economic reasons. Modelled recognition efficiency is 87% for paper, 86% for plastics, 77% for textiles, 68% for organics and 70% for glass, with 99% jet accuracy.

The most instructive result concerns the robotic arm. The model assumed 76% efficiency and about 52.5 picks per minute, which at the average object weight works out at only around 0.09 tonnes per hour per unit. Object-by-object picking can therefore be valuable for quality control or targeted recovery, but it is capacity-intensive and probably uneconomic for processing a bulk municipal stream. Manual picking ran into an even stronger economic limit and was kept only as a benchmark.

The key lesson of the technology review was structural: no single machine could address more than a subset of Milan’s mixed waste composition. That is what moved the analysis from individual technologies to an integrated sorting line.

How the mass-balance model works

The reference plant follows a simple sequence: bag opening → size separation → ferrous removal → eddy-current separation → NIR sorting → recovered or residual material. Each stage is a separate module: incoming tonnes are multiplied by the relevant opening, size-allocation or recovery parameter, and unrecovered material passes on to the next stage. This prevents the same tonne being counted twice and lets the residual composition evolve step by step.

This is what makes the model auditable: change a technology efficiency or a composition assumption and the effect propagates through the whole line rather than appearing only in the top-level total. Two configurations were developed. The two-stream version (below and above 60 mm) is the reference, because it is simpler and offers higher throughput. The three-stream version (below 60 mm, 60–200 mm, above 200 mm) tests whether the largest items can be recovered with a robotic arm.

A key design choice is that the line is essentially dry mechanical sorting. Many mixed-waste plants use wet/dry separation or biological pre-treatment when organics and moisture are the main contaminants. With organics at 5.1% in Milan, the researchers argue this is less necessary.

The model’s most important, and most debatable, element is the final recoverability factor. Mechanical or optical separation is not the same as producing a market-ready secondary material. The reference values provided by AMSA:

• plastic bottles and packaging, rigid plastics, metals, glass: 100%
• paper and cardboard: 50%
• textiles: 25%
• plastic bags, paper towels, organics, other: 0%

These factors convert gross sorting output into tonnes counted as recovered, which is why the final result is well below what the NIR module can physically identify and eject.

Residual waste sorting: 64,000 tonnes a year

In the reference case, residual waste sorting would recover 64,341.6 tonnes a year from a mixed stream of 230,983 tonnes. That is 27.9% of the mixed waste and about 10.1% of all municipal waste collected in Milan. The recovered output breaks down as follows:

• plastic bottles and packaging: 29,587 t/y (46.0%)
• paper and cardboard: 24,490 t/y (38.1%)
• metals: 3,469 t/y (5.4%)
• textiles: 3,360 t/y (5.2%)
• rigid plastic items: 3,166 t/y (4.9%)
• glass: 271 t/y (0.4%)

Plastic packaging and paper together account for about 84% of recovered material, which is no surprise given their share of the input. Plastic bags, sanitary paper, organics and the “other” fraction do not appear on the recovered side because their recoverability factor is zero. The model thus clearly separates whether an item can be technically separated from whether it can be counted as an accepted recyclable output.

Energy, carbon and water: the environmental balance

The researchers estimated environmental effects using proxy factors from the Ansys Granta materials database.

Energy. Replacing primary production with recycled material would save 1.725 million GJ (about 479 GWh) of gross primary energy a year. Because the sorted combustible material no longer goes to the incinerator, the model also deducts the electricity that would no longer be generated: at 35% gross electrical efficiency, the upper end of the BAT range, that is 0.490 million GJ (about 136 GWh) a year. The net benefit still comes to 1.235 million GJ (about 343 GWh). Plastics are the largest contributor, as the energy gap between virgin and recycled production is widest there.

Carbon dioxide. The model adds two mechanisms. Replacing virgin production avoids about 42,000 tonnes of CO₂ a year, and keeping fossil-based material (mainly plastics) out of the incinerator avoids another 77,700 tonnes or so. Combined, that is around 120,000 tonnes of CO₂ a year, almost two-thirds of it from avoided incineration.

Water. The gross avoided virgin-production water footprint is about 59.7 million m³ a year, of which 41.7 million m³ relates to paper and 13.3 million m³ to textiles. The researchers stress that this is a directional resource indicator, not water physically saved in Milan. The textile figure uses an average of cotton and polyester, and cotton’s water factor is orders of magnitude larger than that of most industrial materials.

A professional view: strengths and question marks

The study’s greatest merit is its method. A modular, sequential mass balance links composition, physical sorting, downstream recoverability and environmental effects in a transparent framework that can be transferred to any collection system. The authors themselves present it as a screening tool, not a final plant design. At several points, however, the numbers deserve a careful reading.

Recoverability factors drive everything. The 50% factor for paper and 25% for textiles are among the main drivers of the result. Paper sorted from mixed waste is wet, greasy and contaminated with sanitary material, so paper mills tend to treat it with reservation. For textiles, regulation narrows the options too: under the recent proposal from the Commission’s Joint Research Centre, textiles recovered from mixed municipal waste cannot leave waste status via re-use or remanufacturing. The authors also note that these factors need validating with recyclers and quality specifications.

The energy bridge subtracts apples from oranges. The gross saving is primary energy, while the deducted item is electricity. Producing a unit of grid electricity takes more primary energy, so the real net benefit may be smaller. The model also counts only electricity: if the incinerator sells heat as well, as around 60% of European plants do in combined heat and power mode, the foregone heat should also be accounted for.

The avoided-incineration assumption is optimistic. The 77,700-tonne item assumes that the sorted plastic is not burned anywhere. In practice, a significant share of the rejects from recycling processes also ends up in energy recovery, which the model does not net out. The authors are clear that the environmental calculation is not a full ISO life-cycle assessment: sorting-plant electricity, civil works, changes in transport and downstream yield losses are not included.

The economics are missing. Although economic viability was a central objective, the study publishes no figures on it, because they rely on sensitive information from AMSA and A2A. Yet this is the decisive question. Zero Waste Europe estimates that secondary sorting of mixed waste before incineration costs €39–50 per tonne; in our analysis of the EU ETS rules for waste incineration we explained why the carbon price due from 2031 may not be enough on its own to drive that investment.

Proxy data. The particle-size distribution relies on French data, the composition of a quarter of the mixed stream is assumed, and machine efficiencies may vary with belt loading, moisture, contamination and maintenance.

Even so, the direction is convincing: even alongside excellent separate collection, a meaningful amount of material remains in mixed waste, and recovering part of it may well pay off.

What does this mean for Hungary?

The Milan result cannot be transferred one-to-one to Hungarian conditions, but the question may be even more pressing here.

A different starting point. Hungarian mixed waste has a much higher organic content: the University of Pannonia network model assumed a 20% organic fraction in mixed waste based on laboratory analyses, compared with Milan’s 5.1%. A purely dry sorting line like Milan’s would therefore not work here without biological pre-treatment or wet/dry separation, and wet, organically contaminated paper and textiles would find it even harder to meet recyclers’ requirements.

Infrastructure already exists. A large share of Hungary’s mixed waste goes to mechanical-biological treatment (MBT) plants, where bag opening, screening and magnetic metal separation already take place and the high-calorific fraction is turned into refuse-derived fuel (RDF). In effect, the Milan model asks whether part of the paper and plastic packaging now ending up in RDF could instead be recycled as material. That would require an NIR-based sorting stage on the output side of MBT plants, and a market willing to take material recovered from mixed waste.

A carbon price changes the maths. From 2031 the EU Emissions Trading System is set to extend gradually to waste incineration; we have also summarised the industry’s reaction to the proposal. As burning fossil-carbon materials, above all plastics, becomes more expensive, the business case for sorting improves. An earlier report by Zero Waste Europe and Reloop showed exactly how a carbon price could make mixed waste sorting competitive.

Landfill is the bigger problem. In Milan, mixed waste goes to incineration; in Hungary a large share still goes to landfill. The first task for Hungarian waste management is therefore to curb landfilling and deal with biodegradable waste, as we discussed in connection with the European Landfill Association’s position. According to the Pannonia model, the planned measures will not secure the 2035 EU targets even combined, so material recovery from mixed waste could be one source of the missing percentage points.

The Milan lesson holds in Hungary too: sorting mixed waste does not replace prevention and separate collection, but as a complementary tool it deserves to be taken seriously.

Conclusion

The Milan project shows that circular economy thinking cannot stop where waste is classified as mixed. Even with an advanced separate collection system, nearly 28% of mixed waste may be recoverable if technologies work as a sequence: bag opening and screening prepare the stream, magnets and eddy-current separators remove metals, and optical sorting handles the mixed mass of paper, plastics and textiles. No single technology is sufficient on its own.

The numbers, especially the recoverability factors, the energy bridge and the water footprint, call for careful interpretation, and economic viability remains an open question. The study’s main contribution is therefore its method: a transparent, transferable framework that any city or region can use to work out under which conditions sorting mixed waste pays off.

Residual waste sorting – frequently asked questions

How much material could be recovered from Milan’s residual waste?

According to the model, 64,342 tonnes a year, which is 27.9% of the 230,983-tonne mixed waste stream and about 10.1% of all municipal waste collected in Milan. Of the recovered output, 46% is plastic bottles and packaging and 38% paper and cardboard, with the rest made up of metals, textiles, rigid plastics and glass.

Which technologies make up the Milan sorting line?

The reference line has five consecutive stages: a bag opener, a 60 mm trommel screen, an overband magnet for ferrous metals, an eddy-current separator for aluminium and, finally, an NIR optical sorter for paper, plastics, textiles and glass. Robotic arms and manual picking were also assessed, but their low throughput kept them out of the reference plant.

Why is the recovered amount lower than what the machines sort out?

Because the model filters sorted material through recoverability factors. Plastic packaging, metals and glass count as 100% recovered, paper as 50% and textiles as 25%, while plastic bags, sanitary paper and organics do not count at all. These factors reflect contamination and the acceptance requirements of downstream recyclers.

What environmental benefits does residual waste sorting bring in Milan?

The model estimates about 120,000 tonnes of avoided CO₂ a year: 42,000 tonnes from replacing virgin production and 77,700 tonnes from plastics no longer burned. Even after deducting the electricity the incinerator would no longer generate, a net energy saving of 1.235 million GJ, or about 343 GWh, remains each year.

Can the Milan model be applied in Hungary?

The method can, but the numbers cannot be transferred directly. Hungarian mixed waste has an organic content of roughly 20%, compared with Milan’s 5.1%, so a purely dry sorting line would not work without biological pre-treatment. Existing mechanical-biological treatment plants, extended with NIR sorting, could however divert some of the paper and plastic now ending up in RDF to material recycling.

🔗 Official sources and references: 👉

• Coppa, G., Echevin, M., Ghilotti, G., van Heulen, C., Sibilla, E., Jessen, J. (2026): Rethinking Residual Waste: The Future of Circularity in Milan and beyond. Project report, with input from AMSA and A2A.
• Coppa, G. et al. (2026): Beyond the recycling bin: can Milan recover more of its residual waste? illuminem Voices, 8 November 2026.
• European Parliament and Council: Directive (EU) 2018/851 amending Directive 2008/98/EC on waste; Directive (EU) 2018/850 amending Directive 1999/31/EC on the landfill of waste.
• European Commission, Joint Research Centre (2018): Best Available Techniques (BAT) Reference Document for Waste Treatment.

Consistent structure, quick answers: every article is built so that both readers and AI search tools can find the key points easily. Our content is produced with the help of AI tools and reviewed by our editor.