Forecasting global aluminium flows to demonstrate the need for improved sorting and recycling methods

1 Forecasting Global Aluminium Flows to 1 Demonstrate the Need for Improved 2 Sorting and Recycling Methods 3 4 Simon Van den Eynde1, Ellen Bracquené1, Dillam Diaz-Romero1,2, Isiah Zaplana1, Bart 5 Engelen1,3, Joost R. Duflou1,4, Jef R. Peeters1 6 Department of Mechanical Engineering - KU Leuven, Celestijnenlaan 300A, Box 2422, 3001 7 Leuven, Belgium1 8 PSI-EAVISE - KU Leuven, 2860 Sint-Katelijne-Waver, Belgium2 9 Technology Campus Diepenbeek - KU Leuven, Agoralaan Gebouw B, 3590 Diepenbeek, 10 Belgium3 11 Member of Flanders Make4 12 Abstract 13 The probable emergence of a global aluminium scrap surplus in the coming decade is one of 14 the main incentives for the aluminium recycling industry to invest in new methods and 15 technologies to collect, sort and recycle aluminium scrap. However, due to the considerable 16 uncertainty in the evolution of the global scrap surplus, it is difficult for policymakers and the 17 recycling industry to accurately estimate the economic and environmental advantages of 18 implementing enhanced sorting and recycling methods. The International Aluminium 19 Institute (IAI) has developed a model to track and forecast the global flows of aluminium, but 20 this model is not extensive enough to estimate the scrap surplus evolution. Therefore, this 21 paper introduces an alloy series resolution to the supply and demand of aluminium in the 22 IAI’s global flow model and estimates the composition of the recovered scrap flows to 23 improve the estimate of the technical potential of secondary alloy production. The estimated 24 scrap surplus evolution is subjected to a sensitivity analysis, considering the most critical 25 parameters, including the speed of electrification in the automotive sector, the recovered 26 scrap’s composition and the lifetime of aluminium products. In addition, the estimated 27 2 composition of the recovered aluminium scrap in the model is compared to composition 28 measurements of alumimium scrap collected at a Belgian recycling facility as a means of 29 validation. This study allows to estimate that the global aluminium scrap surplus will emerge 30 soon and reach a size of 5.4 million tonnes by 2030 and 8.7 million tonnes by 2040, if 31 currently adopted aluminium sorting and recycling methods are not improved. 32 Keywords: Aluminium, Forecasting, Material Flow Analysis, Alloys, Scrap surplus 33 1 Introduction 34 The demand for aluminium has been increasing drastically since 1950 due to the global 35 population's growth and the improved standard of living (European Aluminium Association, 36 2021). To date, aluminium is the second most-produced metal, preceded only by steel. 37 Aluminium is produced more than all other non-ferrous metals combined (Cullen and 38 Allwood, 2013). In the last two decades, the demand for aluminium has grown faster than 39 that for any other metal, increasing at a significantly faster rate than the global GDP (Fog, 40 2019). Its light weight, high strength, good corrosion resistance and high conductivity make 41 aluminium an attractive choice for many products, including food packaging, car parts, 42 airplane components and building features. The increased use of aluminium has led to 43 significant weight reductions of components in the automotive and aerospace sector which 44 have saved large amounts of fuel in the use phase of cars, trucks, and planes (European 45 Aluminium Association, 2013). However, aluminium production itself has substantial 46 environmental impact, in the form of toxicity, acidification, greenhouse gas emissions and 47 resource depletion (Schlesinger, 2017; The Economist, 2007). In 2020, the primary 48 production of aluminium was responsible for the emission of more than 1 billion metric 49 tonnes of CO2-equivalents, accounting for almost 2% of the global human-caused emissions 50 in that year (Saevarsdottir et al., 2020; Van Heusden et al., 2020). In order to reduce the 51 aluminium industry’s environmental impact, companies and policymakers increasingly focus 52 3 on aluminium recycling as a potential solution, with as main driver the substantial difference 53 in energy consumption: producing 1 kg of recycled aluminium requires on average 9.2 MJ, 54 compared to 144.6 MJ for producing 1 kg of primary aluminium (Peng et al., 2019). 55 The European Aluminium Association (EAA), the organisation representing the European 56 aluminium industry, forecasts a rise in the share of recycled aluminium in European end-use 57 products from 26% in 2000 to 49% in 2050 (European Aluminium Association, 2019). In its 58 “VISION 2050” report, the EAA explains that this is an ambitious but realistic evolution that 59 will significantly contribute to the European decarbonisation efforts. However, most collected 60 aluminium scrap today contains a mixture of different alloy types. As a result, different 61 alloying elements and impurities are present in the scrap. Removing these elements 62 metallurgically from the secondary aluminium is notoriously difficult (Nakajima et al., 2010). 63 Therefore, most collected aluminium scrap is “downcycled” and used for the production of 64 cast aluminium alloys, which have high tolerances for impurities (Paraskevas et al., 2015). A 65 smaller share of the collected scrap is used to produce wrought aluminium alloys, which have 66 much lower tolerances for alloying elements and impurities. To produce wrought alloys from 67 mixed scrap, it needs to be diluted with large amounts of primary aluminium. 68 Although this downcycling practice has been a successful strategy because of the high 69 demand for cast aluminium alloys for the production of combustion engines, this is expected 70 to change with the electrification of the automotive industry. Due to this transition, the global 71 demand for cast aluminium alloys will stagnate or is even expected to decline 72 (BloombergNEF, 2019; Modaresi and Müller, 2012). Simultaneously, the amount of 73 aluminium scrap collected from end-of-life products and the demand for wrought aluminium 74 alloys will keep growing. Previous research has suggested that, if the current practice of 75 systematic downcycling is maintained, the collected amount of aluminium scrap will soon 76 exceed the capacity of wrought and cast alloy production to absorb the secondary aluminium. 77 4 As such, an amount of aluminium scrap would be collected for which there is no suitable 78 application. This amount of aluminium scrap is commonly referred to as a scrap surplus. 79 Hatayama et al. (2012) estimate the scrap surplus size at 6.1 million tonnes in 2030. Modaresi 80 and Müller (2012) and Modaresi et al. (2014) expect a scrap surplus of 4.2 million tonnes in 81 2030 that will grow to a size of 14 million tonnes by 2050. However, they add that due to the 82 uncertainty in their parameters, the scrap surplus's actual size could lie anywhere between 3.3 83 and 18.3 million tonnes in 2050. 84 Therefore, this paper estimates the evolution of the global scrap surplus by expanding the 85 global flow model of the International Aluminium Institute (IAI). This model is a prominent 86 tool in the aluminium industry that tracks and predicts the volumes of aluminium throughout 87 the different life cycle stages. In this paper, an alloy series resolution is introduced in the 88 supply and demand data of the IAI and the composition of the recovered scrap flows is 89 estimated to improve the estimate of the technical potential of secondary alloy production. 90 The estimated evolution of the global scrap surplus is also subjected to a sensitivity analysis, 91 considering the most critical uncertain parameters that affect its growth. In addition, the 92 estimated composition of the recovered aluminium scrap in the model is compared to 93 composition measurements of aluminium scrap collected at a Belgian recycling facility as a 94 means of validation. These measurements are performed using a handheld X-Ray 95 Fluorescence (XRF) device. 96 2 Methodology 97 2.1 Demand for Aluminium Alloys 98 The IAI publishes annual data on the global flows of aluminium from different studies and 99 surveys. Bertram et al. (2009) combined these data into a single model, which resulted in the 100 first global flow model for aluminium, published in 2009. Ever since, the global aluminium 101 5 flow model of the IAI has been updated regularly (Bertram et al., 2017). Stakeholders in the 102 aluminium industry often refer to the model, that is freely accessible on the website of the IAI 103 (NTNU et al., 2020). Similar efforts to model and predict (global) flows of aluminium have 104 been made by other researchers as well (Dai et al., 2019; Zhu et al., 2021). 105 Figure 2.1 is adapted from Bertram et al. (2017) and shows the structure of the IAI’s global 106 flow model. It models the flows of aluminium throughout the life cycle stages per region and 107 then links all regions together. The global flow data of the IAI include the amounts of 108 aluminium that flow to the manufacturing phase in the different industrial sectors since 1950, 109 and projections are made for these flows until the year 2040. Because the IAI has access to 110 extensive databases from reliable sources worldwide, the accuracy of their data is 111 unparalleled. No other material industry has succeeded in quantitatively modelling global 112 material flows with a similar level of accuracy (Bertram et al., 2009). However, the major 113 drawback of the MFA model of the IAI is that the aluminium is treated as a single, uniform 114 material that seems unaltered when it goes from one stage in the lifecycle to the next. 115 However, in practice aluminium is mostly alloyed, and the aluminium material flows undergo 116 significant compositional changes, especially in the end-of-life phase. Another significant 117 drawback of the IAI’s model is that it assumes that all collected aluminium scrap can be 118 remelted into new alloys without verifying the allowable recycled content. Therefore, the 119 IAI’s model cannot predict a possible emergence of a scrap surplus. Furthermore, the authors 120 chose to rely on projections of the EAA (European Aluminium Association, 2019) for the 121 demand for aluminium between 2030 and 2040 instead of using the numbers of the IAI. The 122 main difference is that the EAA forecasts a demand that keeps growing significantly up to 123 2040 while the IAI expects the demand to stagnate more between 2030 and 2040. Even 124 though the methodology behind the projections of the EAA is not elaborately explained in the 125 published report itself, the authors chose to use these numbers because they are the result of 126 6 more recent research and because they are probably more accurate, since they were 127 specifically estimated by CRU, a market analysis firm specialised in metals, whereas the 128 projections of the IAI are based on a relatively simple time series. 129 130 Figure 2.1: Additions to the global flow model of the International 131 Aluminium Institute (red: alloy series resolution; green: composition 132 estimate; blue estimate scrap surplus size), based on Bertram et al. (2017) 133 To overcome these shortcomings, the structural contributions of this paper to the global flow 134 model of the IAI are threefold. Firstly, it introduces alloy series resolution into the modelled 135 supply and demand of aluminium. The demand for aluminium ingots, semis, and final 136 products, as well as the generation of aluminium EOL products, is modelled on an alloy level 137 in this research, whereas the global flow model of the IAI only estimates the volumes of 138 aluminium. The flows in red in Figure 2.1 (F11-21) are the ones for which the alloy series 139 resolution is added. Secondly, whereas the global flow model of the IAI only estimates the 140 total volumes of generated aluminium scrap, the presented research includes the elemental 141 composition of the recovered aluminium scrap, based on the estimated amounts of alloys in 142 recovered EOL products. The flows for which the elemental composition is calculated are 143 indicated in green (F23-26). The global flow model of the IAI only estimates the volumes of 144 generated aluminium scrap. Finally, the developed model estimates the size of the generated 145 7 scrap surplus, which is not considered at all in the global flow model of the IAI. This 146 contribution is indicated in blue (F27). The remainder of the model is unchanged with respect 147 to the original IAI model. 148 The alloy series resolution in the demand for aluminium is introduced by determining each 149 aluminium alloy series' share in the annual aluminium demand per industrial sector. This 150 demand (𝐷𝑆𝐸𝐶𝑇𝑂𝑅 𝑆𝐸𝑅𝐼𝐸𝑆) is calculated by multiplying the total demand for aluminium in a sector 151 (𝐷𝑆𝐸𝐶𝑇𝑂𝑅), according to the data of the IAI, by the share of that alloy series in the demand of 152 that sector (𝑆𝑆𝐸𝐶𝑇𝑂𝑅 𝑆𝐸𝑅𝐼𝐸𝑆), as expressed in Formula 2.1. The annual shares of the alloy series in the 153 total demand for aluminium per sector, considering evolutions in the demand over time, are 154 determined by performing an extensive literature study, as detailed in Appendix 1, combining 155 industry data, governmental data and data published by previous research on aluminium use 156 for all 12 sectors that are defined by the IAI: (1) “Building & Construction”, (2) 157 “Transportation – Auto & Light Truck”, (3) “Transportation – Aerospace”, (4) 158 “Transportation – Other”, (5) “Packaging – Cans”, (6) “Packaging – Other (Foil)”, (7) 159 “Machinery & Equipment”, (8) “Electrical – Cable”, (9) “Electrical – Other”, (10) 160 “Consumer Durables”, (11) “Other (except Destructive Uses)”, and (12) “Destructive Uses”. 161 Error! Reference source not found. in Appendix 2 illustrates the shares of the alloy series 162 in the total demand for aluminium in the defined sectors for the year 2020 and indicates on 163 which references the data are based. 164 𝐷𝑆𝐸𝐶𝑇𝑂𝑅 𝑆𝐸𝑅𝐼𝐸𝑆 = 𝐷𝑆𝐸𝐶𝑇𝑂𝑅⋅𝑆𝑆𝐸𝐶𝑇𝑂𝑅 𝑆𝐸𝑅𝐼𝐸𝑆 (2.1) 2.2 Scrap Generation 165 Secondary aluminium is sourced from “new scrap” and “old scrap”. New scrap, also referred 166 to as production scrap or pre-consumer scrap, is generated in the manufacturing phase due to 167 process inefficiencies. Old scrap originates from end-of-life products. The annual amounts of 168 8 new and old scrap collected for recycling from each industrial sector are included in the IAI 169 data. The developed model requires both the elemental composition of the collected scrap and 170 the volumes of these scrap flows. In the model's calculations, the composition of the scrap is 171 first determined on an alloy level and then converted to an elemental level. 172 For new scrap, it is assumed that the generated scrap in a certain year consists of the same 173 alloys that entered the manufacturing phase that year. As such, the amount of new scrap from 174 a specific alloy series that is generated in an industrial sector in a specific year 175 (𝐴𝑁𝐸𝑊 𝑆𝐸𝑅𝐼𝐸𝑆,𝑆𝐸𝐶𝑇𝑂𝑅) can be calculated by multiplying the total amount of generated new scrap in 176 the sector (𝐴𝑁𝐸𝑊 𝑆𝐸𝐶𝑇𝑂𝑅) with the share of the alloy series in the demand for aluminium in the 177 sector (𝑆𝑆𝐸𝐶𝑇𝑂𝑅 𝑆𝐸𝑅𝐼𝐸𝑆). The total amount of generated new scrap from a specific alloy series 178 (𝐴𝑁𝐸𝑊 𝑆𝐸𝑅𝐼𝐸𝑆) can be calculated by adding up the amounts of the different sectors. Dividing this 179 number by the total amount of collected new scrap (𝐴𝑁𝐸𝑊) gives the share of scrap from a 180 certain alloy series in the total amount of collected new scrap (𝐶𝑁𝐸𝑊 𝑆𝐸𝑅𝐼𝐸𝑆). These calculations 181 are summarised below in Formulas 2.2 to 2.4. Calculating every alloy series' share leads to a 182 complete alloy level composition of the collected new scrap. This alloy level composition 183 still has to be converted to an elemental level composition. 184 𝐴𝑁𝐸𝑊 𝑆𝐸𝑅𝐼𝐸𝑆,𝑆𝐸𝐶𝑇𝑂𝑅 = 𝐴𝑁𝐸𝑊 𝑆𝐸𝐶𝑇𝑂𝑅⋅𝑆𝑆𝐸𝐶𝑇𝑂𝑅 𝑆𝐸𝑅𝐼𝐸𝑆 (2.2) 𝐴𝑁𝐸𝑊 𝑆𝐸𝑅𝐼𝐸𝑆= ∑ (𝐴𝑁𝐸𝑊 𝑆𝐸𝐶𝑇𝑂𝑅⋅𝑆𝑆𝐸𝐶𝑇𝑂𝑅 𝑆𝐸𝑅𝐼𝐸𝑆) 𝑆𝐸𝐶𝑇𝑂𝑅 (2.3) 𝐶𝑁𝐸𝑊 𝑆𝐸𝑅𝐼𝐸𝑆= 𝐴𝑁𝐸𝑊 𝑆𝐸𝑅𝐼𝐸𝑆/ 𝐴𝑁𝐸𝑊 (2.4) 185 For old scrap, estimating which alloys can be expected in the collected scrap is more complex 186 since most aluminium products have a much longer lifetime than one year. Therefore, the 187 alloys collected from end-of-life products in a specific year are not identical to those that 188 entered the use phase during that year. The average time aluminium remains “in stock” in the 189 use phase is estimated by the IAI per sector and region. It varies from one year for packaging 190 9 to 60 years for aluminium in buildings (Bertram et al., 2017). For the developed MFA model, 191 it is assumed that the alloy series in the scrap of a sector are present in the same proportions 192 as the alloy series that entered the use phase one average lifetime ago for the products in that 193 sector. This assumption does not allow to consider possible variations in the lifetime of the 194 products within a sector. However, this approach still yields reasonable approximations since, 195 in most sectors, the use of alloys in the manufacturing process changes only gradually during 196 the products' average lifetime. 197 With this assumption, the amount of old scrap from a particular alloy series from a certain 198 sector that is collected for recycling (𝐴𝑂𝐿𝐷 𝑆𝐸𝑅𝐼𝐸𝑆,𝑆𝐸𝐶𝑇𝑂𝑅) can be calculated similarly as for the 199 new scrap (see Formula 2.5). 𝐴𝑂𝐿𝐷 𝑆𝐸𝐶𝑇𝑂𝑅 is the amount of old scrap from an industrial sector 200 collected for recycling. Annual numbers for these scrap flows are included in the IAI data, as 201 well as the amount of aluminium end-of-life scrap that is not collected for recycling. This 202 scrap mostly ends up in landfills. The apostrophe in the symbol 𝑆′𝑆𝐸𝐶𝑇𝑂𝑅 𝑆𝐸𝑅𝐼𝐸𝑆 stresses the time 203 delay between the manufacturing phase and the end-of-life phase of the aluminium products 204 in the sector, which must be considered. 205 The total amount of scrap from each alloy series in the collected old scrap from all sectors 206 (𝐴𝑂𝐿𝐷 𝑆𝐸𝑅𝐼𝐸𝑆) can be calculated by summing up the amounts from the different industrial sectors, 207 as expressed in Formula 2.6. An exceptional flow in the developed MFA model is the 208 collected aluminium scrap from used beverage cans (UBC). The researchers that contributed 209 to the global flow model of the IAI indicate that scrap from UBC reaches cast houses mostly 210 separately from casting scrap, extruded scrap, rolled scrap and other scrap from different 211 sources (Bertram et al., 2017). Therefore, UBC recycling is modelled as a closed-loop 212 system, separate from the remainder of the collected scrap. As such, the “Packaging – Cans” 213 sector is not included in the summation of Formula 2.6. The amount of aluminium that has to 214 16 between the demand for the wrought alloy series (𝐷𝑊𝑅.𝑆𝐸𝑅) and the amount of recycled 312 aluminium that flows to the alloy series (𝑆𝑆𝐸𝐶 𝑊𝑅.𝑆𝐸𝑅). As long as the amount of generated 313 aluminium scrap that is not used for wrought alloy production is smaller than the demand for 314 cast alloys, there is no scrap surplus (𝑆𝑃). In this case, the scrap that is not used for wrought 315 alloy production can be absorbed entirely by the cast alloys. The amount of primary 316 aluminium used in the production of the cast alloys (𝑆𝑃𝑅𝐼 𝐶𝐴𝑆𝑇) is then equal to the difference 317 between the demand for cast alloys (𝐷𝐶𝐴𝑆𝑇) and the amount of secondary aluminium used in 318 the production of cast alloys (𝑆𝑆𝐸𝐶 𝐶𝐴𝑆𝑇). However, if the amount of recycled aluminium exceeds 319 the capacity of both wrought and cast alloys to absorb this material, there is no destination 320 left for this flow. Then, the size of the scrap surplus is equal to the generated amount of scrap 321 after melting and premelting losses (𝐴′𝑆𝐸𝐶) minus the demand for cast alloys and the amount 322 of secondary aluminium used in wrought alloy production. These calculations are 323 summarised in Formulas 2.13 to 2.17. All the model calculations have been made in a 324 Microsoft Excel Workbook and are visualised in the form of a Sankey diagram using the 325 Python programming language and the FloWeaver library (Lupton and Allwood, 2017; 326 Lupton, 2017). 327 𝑆𝑆𝐸𝐶 𝑊𝑅.𝑆𝐸𝑅= 𝑅𝐶𝑊𝑅.𝑆𝐸𝑅⋅ 𝐷𝑊𝑅.𝑆𝐸𝑅 (2.13) 𝑆𝑃𝑅𝐼 𝑊𝑅.𝑆𝐸𝑅= 𝐷𝑊𝑅.𝑆𝐸𝑅− 𝑆𝑆𝐸𝐶 𝑊𝑅.𝑆𝐸𝑅 (2.14) 𝑆𝑆𝐸𝐶 𝐶𝐴𝑆𝑇= { 𝐴′𝑆𝐸𝐶−∑𝑆𝑆𝐸𝐶 𝑊𝑅.𝑆𝐸𝑅 𝑆𝐸𝑅𝐼𝐸𝑆 𝑖𝑓 𝐴′𝑆𝐸𝐶−∑𝑆𝑆𝐸𝐶 𝑊𝑅.𝑆𝐸𝑅 𝑆𝐸𝑅𝐼𝐸𝑆 < 𝐷𝐶𝐴𝑆𝑇 𝐷𝐶𝐴𝑆𝑇 𝑖𝑓 𝐴′𝑆𝐸𝐶−∑𝑆𝑆𝐸𝐶 𝑊𝑅.𝑆𝐸𝑅 𝑆𝐸𝑅𝐼𝐸𝑆 > 𝐷𝐶𝐴𝑆𝑇 (2.15) 𝑆𝑃𝑅𝐼 𝐶𝐴𝑆𝑇= { 𝐷𝐶𝐴𝑆𝑇− (𝐴′𝑆𝐸𝐶−∑𝑆𝑆𝐸𝐶 𝑊𝑅.𝑆𝐸𝑅 𝑆𝐸𝑅𝐼𝐸𝑆 ) 𝑖𝑓 𝐴′𝑆𝐸𝐶−∑𝑆𝑆𝐸𝐶 𝑊𝑅.𝑆𝐸𝑅 𝑆𝐸𝑅𝐼𝐸𝑆 < 𝐷𝐶𝐴𝑆𝑇 0 𝑖𝑓 𝐴′𝑆𝐸𝐶−∑𝑆𝑆𝐸𝐶 𝑊𝑅.𝑆𝐸𝑅 𝑆𝐸𝑅𝐼𝐸𝑆 > 𝐷𝐶𝐴𝑆𝑇 (2.16) 𝑆𝑃= { 0 𝑖𝑓 𝐴′𝑆𝐸𝐶−∑𝑆𝑆𝐸𝐶 𝑊𝑅.𝑆𝐸𝑅 𝑆𝐸𝑅𝐼𝐸𝑆 < 𝐷𝐶𝐴𝑆𝑇 𝐴′𝑆𝐸𝐶−𝐷𝐶𝐴𝑆𝑇−∑𝑆𝑆𝐸𝐶 𝑊𝑅.𝑆𝐸𝑅 𝑆𝐸𝑅𝐼𝐸𝑆 𝑖𝑓 𝐴′𝑆𝐸𝐶−∑𝑆𝑆𝐸𝐶 𝑊𝑅.𝑆𝐸𝑅 𝑆𝐸𝑅𝐼𝐸𝑆 > 𝐷𝐶𝐴𝑆𝑇 (2.17) 17 328 2.5 Aluminium Scrap Sampling 329 The demand for aluminium alloys and the composition of the collected aluminium scrap are 330 calculated in the model based on data from literature. To be able to compare the literature 331 data with the actual composition of aluminium scrap collected at recycling facilities, 332 composition measurements have been conducted on aluminium scrap samples collected at a 333 Belgian recycling facility. A batch of 275 aluminium scrap samples with a total mass of over 334 10 kg was collected from the so called “Twitch (40-120 mm)” fraction at the recycling 335 facility of Galloo in Menen, Belgium. This fraction consists almost exclusively of aluminium 336 scrap. The aluminium in the Twitch fraction of Galloo is separated from other materials in the 337 treated end-of-life waste streams by magnetic separation, density separation, eddy current 338 separation, and optical separation (Eggers et al., 2019). According to the representatives of 339 Galloo, the aluminium in the Twitch fraction originates on average for 40% from 340 construction waste, for 40% from automotive waste and for 20% from waste from consumer 341 durables. 342 The composition of the collected samples was measured with a handheld XRF device 343 (Thermo Scientific Niton XL2). Based on the XRF measurements, the 275 samples were 344 divided in five categories related to their alloy type. The 1000 series, 3000 series, and cast 345 alloys are three separate categories. The 2000, 4000, 7000, and 8000 series alloys are 346 combined in the “Other Wrought” category since they are significantly less popular than the 347 other series and barely encountered during the measurements. The alloys of the 5000 and 348 6000 series are bundled in one category due to the difficulty of distinguishing between these 349 series with the used measuring method. 350 18 2.6 Method for sensitivity analysis 351 Predicting the composition of the global aluminium flows and the size of the scrap surplus 352 involves considerable uncertainties. Since the evolution of the scrap surplus is estimated by 353 expanding the IAI’s global flow model, not only the uncertainties that arise from the 354 assumptions in this paper have to be considered, but also the uncertain parameters within the 355 IAI model. Two sources of uncertainty that are introduced in this paper and that have a 356 particularly high influence on the scrap surplus growth are the changing demand for cast 357 alloys in the automotive sector and the composition of the collected scrap. The estimated 358 demand for cast alloys in the automotive sector in the presented MFA model relies on the 359 projections of Modaresi et al. (2014). However, predictions for the future demand of cast 360 alloys for cars and light trucks vary significantly, depending on the consulted source, due to 361 the considerable uncertainty in the rate at which electric vehicles will gain market share in the 362 automotive sector in the coming decades (Hatayama et al., 2012; Buchner et al., 2017). 363 Furthermore, the most critical parameter within the global flow model of the IAI is the 364 lifetime of the products in the different sectors. To show how the estimated evolution of the 365 scrap surplus is affected by these different uncertainties in the data, a scenario-based 366 sensitivity analysis is performed. 367 Seven alternative scenarios are investigated in addition to the baseline scenario. In the first 368 alternative scenario (“slow electrification”), the share of cast alloys in the aluminium demand 369 in the automotive sector diverges gradually from the share assumed in the baseline scenario 370 between 2020 and 2040. In 2020, this share is still assumed the same as in the baseline 371 scenario, but it linearly increases until 2040 when it is ten percentage points higher than in 372 the baseline scenario. This increased demand for cast alloys corresponds to a situation in 373 which the transition towards electric vehicles takes place slower than anticipated in the 374 baseline scenario. In the second alternative scenario (“fast electrification”), the share of cast 375 19 alloys in the automotive sector’s aluminium demand is gradually decreased from 2020 376 onwards until it is ten percentage points lower than in the baseline scenario in 2040. This 377 scenario represents a faster than anticipated electrification in the automotive sector. 378 The third and fourth alternative scenarios illustrate the influence of the composition of the 379 collected aluminium scrap. In the baseline scenario, the secondary scrap composition is 380 calculated based on the average concentration of alloying elements in the wrought alloys, as 381 presented in the “Average” rows of Table 2.1. In the third (“Min Alloys”) and fourth (“Max 382 Alloys”) alternative scenarios, the secondary scrap composition is calculated based on the 383 values in the “Min” and “Max” rows in Table 2.2, respectively. In the third alternative 384 scenario, an alloying element's concentration in an alloy is assumed to be 0.01wt% when no 385 lower limit is specified for the production of that alloy. This is a very low value, even for 386 wrought alloys, and considered the minimum concentration of an alloying element in 387 collected scrap since, in a realistic instance, the average concentration of alloying elements 388 and impurities will never be exactly zero. This specification is important to avoid too 389 unrealistic values in this scenario. These two scenarios also offer an idea about the magnitude 390 of the change in the size of the scrap surplus that results from the presence of tramp elements 391 in the collected aluminium scrap. According to Soo et al. (2018), the iron and copper 392 impurities due to tramp elements account for 0.03wt% to 0.36wt% and 0.13wt% to 0.26wt% 393 of the aluminium scrap, respectively. In the third and fourth alternative scenario, the 394 considered deviations in the iron and copper content are slightly larger than these numbers. 395 Therefore, the considered deviations in these scenarios cover the range of impurity 396 concentrations that can be realistically expected in aluminium scrap. 397 The fifth alternative scenario (“Galloo”) considers the results of the scrap measurements. 398 Whereas in the baseline scenario, the alloy-level composition of the scrap from the sectors 399 “Building & Construction”, “Transportation – Auto & Light Truck”, and “Consumer 400 20 Durables” is determined based on literature data, in this alternative scenario the compositions 401 are adjusted so that they match the results of the scrap measurements. The estimated share of 402 the 1000 series, 3000 series, and cast alloys in the recovered scrap is increased, while the 403 share of 5000, 6000, and 7000 series alloys is decreased with respect to the literature data. 404 The purpose of including this scenario is to demonstrate the sensitivity of the scrap surplus 405 growth to the alloy-level composition of the collected old scrap. 406 To demonstrate the sensitivity of the scrap surplus evolution to the product lifetimes assumed 407 in the IAI global flow model, a sixth and seventh alternative scenario have been developed. 408 In the global flow model, global average lifetimes of aluminium products are estimated per 409 sector (see Appendix 3). In the sixth scenario, these average lifetimes are reduced in a linear 410 way with 10% between 2020 and 2040. In the seventh alternative scenario, the average 411 lifetimes are reduced in a linear way by 20% between 2020 and 2040. 412 3 Results & Discussion 413 3.1 Limits to recycled content for wrought alloys 414 Table 3.1 shows the limit that each element in the collected aluminium scrap imposes on the 415 recycled content of the different wrought alloy series (𝐿𝑆𝐸𝑅𝐼𝐸𝑆 𝐸𝐿𝐸𝑀𝐸𝑁𝑇) in 2020. The recycled 416 content (𝑅𝐶𝑊𝑅.𝑆𝐸𝑅), the most critical limit for each alloy series, is marked in yellow. 417 Table 3.1 Calculated Upper Limits for the Recycled Content of the Wrought 418 Alloy Series in 2020 419 Element 1000 2000 3000 4000 5000 6000 7000 8000 Cu 5% 100% 26% 31% 10% 10% 100% 5% Fe 100% 100% 100% 100% 93% 93% 32% 100% Mg 4% 4% 100% 4% 100% 89% 100% 4% Mn 17% 100% 100% 17% 34% 34% 20% 17% Si 9% 19% 11% 100% 10% 23% 4% 5% Zn 8% 42% 42% 16% 16% 16% 10% 16% Other 36% 90% 90% 90% 90% 90% 90% 90% 420 21 3.2 Sankey diagram 421 Figure 3.1 shows the Sankey diagram for the calculated aluminium flows in 2030. Diagrams 422 for 2035 and 2040 are included in Appendix 4. In the Sankey diagram, the nodes are ordered 423 according to the different lifecycle stages, and the colors indicate the alloy series or the type 424 of aluminium that flows between the nodes. The scrap from UBC is recycled in a closed-loop 425 recycling scheme, in contrast to the aluminium flows for the other end-of-life products. The 426 scrap surplus is represented as a flow that starts in the supply phase, which cannot be 427 connected to the manufacturing phase. 428 According to the model's calculations, the scrap surplus would constitute 11.4% of all 429 collected aluminium scrap in 2030. This would mean that in 2030, only 88.6% of all collected 430 and processed aluminium scrap could be used for the production of new wrought and cast 431 alloys. By 2040, this share is estimated to decline to 84.0%. In such situation, the use of 432 primary aluminium would continue to grow despite the abundant availability of aluminium 433 scrap. 434 22 Figure 3.1: Sankey diagram representing the global aluminium flows in 2030, shades of green represent the alloy series (1000-8000 + cast), data in Appendix 5 23 3.3 Composition Measurements 435 Figure 3.2 shows the measured shares of the different alloy series in the set of aluminium 436 (Twitch) scrap samples collected at Galloo, as well as the model’s prediction for the 437 composition of old scrap collected in 2020 from a mix of 40% scrap from construction, 40% 438 automotive scrap and 20% scrap from consumer durables. The measured share of cast alloys 439 in the Twitch fraction is relatively close to the estimated share. Among the wrought alloys, 440 there are significant differences between the measured and estimated shares of the different 441 alloy series. There are multiple explanations for these differences. First of all, it could be that 442 the gathered sample does not exactly reflect the average composition of the Twitch fraction 443 over the course of one year. The assumed origins of the measured scrap (40% construction, 444 40% automotive, and 20% consumer durables) are a company estimate of yearly averages. 445 However, the composition of the scrap variates significantly throughout the year. Another 446 important consideration is that the global average composition of a mix of old scrap is 447 compared with the composition of the scrap at a local Belgian recycling facility. Since there 448 are regional differences in the use of alloys for the production of aluminium products, it is not 449 unexpected that the scrap collected in Belgium has a slightly different composition than the 450 global average. The importance of an accurate estimate of the collected scrap’s alloy-level 451 composition for estimating the growth of the scrap surplus is shown in the sensitivity analysis 452 in the next section. 453 24 Figure 3.2: Alloy-level composition of collected scrap: Predicted by Model and Measured 3.4 Scenario-based sensitivity analysis for the evolution of the scrap surplus 454 Error! Reference source not found. shows the estimated evolution of the global scrap 455 surplus between 2020 and 2040. In this figure, the baseline scenario of the predicted 456 evolution is compared to seven other scenarios. These scenarios demonstrate the sensitivity 457 of the scrap surplus’s growth to the most important sources of uncertainty in the model. 458 According to the calculations in the baseline scenario, a small global scrap surplus of 0.5 459 million tonnes first surfaced for one year in 2009, the year that the global economy contracted 460 by almost 2% (The World Bank, 2021). In this year of recession, the London Metal Exchange 461 suffered significant financial losses when its stock of aluminium increased by 152% in one 462 year to 2.34 million tonnes (Salazar and McNutt, 2010). However, it is difficult to prove the 463 scrap surplus’s contribution to this incident since the most significant part of these stocks was 464 probably the result of the overproduction of primary aluminium, as the sudden collapse in 465 demand took many aluminium producers by surprise. Starting from 2023, an annual global 466 25 scrap surplus is predicted to arise and to continue to grow quickly to 5.4 million tonnes of 467 aluminium in 2030 and then to around 8.7 million tonnes in 2040. This prediction is 468 relatively close to what other researchers have predicted (Hatayama et al., 2008, 2012; 469 Modaresi and Müller, 2012; Modaresi et al., 2014). There is a kink in the scrap surplus 470 evolution in the year 2030, due to the decrease in the annual growth rate of the demand for 471 aluminium that is expected by the EAA. As a consequence, the generated amount of new 472 scrap and the amount of old scrap from products with very short lifetimes increase less 473 rapidly. 474 Figure 3.3: Sensitivity analysis for the evolution of the annual global aluminium scrap surplus, data in Appendix 5 Figure 3.3 shows that both the demand for cast alloys in the automotive sector and the 475 amount of alloying elements in the collected scrap have a significant influence on the scrap 476 surplus growth. The “Galloo” scenario lies relatively close to the baseline scenario, even 477 though there are significant differences between the measured alloy-level composition of the 478 collected scrap at Galloo and the alloy-level composition predicted from literature data. 479 Figure 3.2 showed that, especially among the wrought alloys, the predicted and measured 480 0 2000 4000 6000 8000 10000 12000 14000 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 Mass [x1000 tonnes] 10% Shorter Life Fast Electrification Max Alloys Baseline Galloo Min Alloys Slow Electrification 10% Longer Life 32 https://aluminum.org/sites/default/files/Aluminum%20Building%20Wire%20Installation 603 %20&%20Terminations.pdf (accessed: Jun. 06, 2021) 604 CM Group, 2020. An Initial Assessment of the Impact of the Covid-19 Pandemic on Global 605 Aluminium Demand. IAI. https://www.world606 aluminium.org/media/filer_public/2020/05/28/initial_assessment_of_the_impact_of_the_ 607 covid-19_on_global_al_demand_.pdf (accessed: Apr. 28, 2021) 608 Cullen, J.M., Allwood, J.M., 2013. Mapping the Global Flow of Aluminum: From Liquid 609 Aluminum to End-Use Goods. Environmental science & technology, vol. 47, no. 7, pp. 610 3057–3064. https://doi.org/10.1021/es304256s. 611 Dai, M., Wang, P., Chen, W., Liu, G., 2019. Scenario analysis of China's aluminum cycle 612 reveals the coming scrap age and the end of primary aluminum boom. J ournal of Cleaner 613 Production, vol 226, pp. 793-804. https://doi.org/10.1016/j.jclepro.2019.04.029. 614 Das, S.K., Green, J.A.S., Kaufman, J.G., Emadi, D., Mahfoud, M., 2010. Aluminum 615 recycling—An integrated, industrywide approach. JOM, vol. 62, no. 2, pp. 23–26. 616 https://doi.org/10.1007/s11837-010-0026-6. 617 Djukanovic, G., 2016. Aluminium alloys in shipbuilding – a fast growing trend. Aluminium 618 Insider. https://aluminiuminsider.com/aluminium-alloys-in-shipbuilding-a-fast-growing619 trend/ (accessed: Apr. 09, 2021) 620 Djukanovic, G., 2017. Aluminium use in the production of trains steams ahead. Aluminium 621 Insider. https://aluminiuminsider.com/aluminium-use-production-trains-steams622 ahead/#:~:text=Made%20in%20China%20%E2%80%93%20the%20high,speed%20of% 623 20380%20km%2Fhr (accessed: Apr. 09, 2021). 624 Ducker Worldwide, 2011. Aluminum in 2012 North American Light Vehicles. Ducker 625 Worldwide, Troy, Michigan, USA. http://www.drivealuminum.org/wp626 33 content/uploads/2017/05/2012_Aluminum-in-2012-North-American-Light-Vehicles.pdf 627 (accessed: Apr. 09, 2021) 628 Ducker Worldwide, 2012. EAA Aluminium penetration in cars - Final Report. Ducker 629 Worldwide, Troy, Michigan, USA. https://www.european-aluminium.eu/media/1308/eaa630 aluminium-penetration-in-cars_final-report-public-version.pdf (accessed: Apr. 09, 2021) 631 Ducker Worldwide, 2017. Aluminum content in North American light vehicles 2016 to 2028. 632 Ducker Worldwide, Troy, Michigan, USA. http://www.drivealuminum.org/wp633 content/uploads/2017/10/Ducker-Public_FINAL.pdf (accessed: Apr. 09, 2021) 634 Eggers, A., Peeters, J.R., Waignein, L., Noppe, B., Dewulf, W., Vanierschot M., 2019. 635 Development of a computational fluid dynamics model of an industrial scale dense 636 medium drum separator. Engineering Applications of Computational Fluid Mechanics, 637 vol. 13, no. 1, pp. 1001–1012. https://doi.org/10.1080/19942060.2019.1663559. 638 European Aluminium Association, 2013. Aluminium in Cars - Unlocking the light-weighting 639 potential. https://european-aluminium.eu/media/1326/aluminium-in-cars-unlocking-the640 lightweighting-potential.pdf (accessed: Apr. 07, 2021) 641 European Aluminium Association, 2015. The Aluminium Automotive Manual. 642 https://www.european-aluminium.eu/resource-hub/aluminium-automotive-manual/ 643 (accessed: Apr. 09, 2021) 644 European Aluminium Association, 2019. VISION 2050: European Aluminium’s Contribution 645 to the EU’s Mid-Century Low-Carbon Roadmap. https://www.european646 aluminium.eu/media/2552/sample_executive-summary-vision647 2050_web_pages_20190408.pdf (accessed: Apr. 07, 2021) 648 Fog, K., 2019. Hydro Market Outlook. Norsk Hydro. 649 https://www.hydro.com/Document/Index?name=Market%20outlook%20by%20Senior% 650 34 20Vice%20President%20and%20Head%20of%20Corporate%20Strategy%20%26%20A 651 nalysis%20Kathrine%20Fog&id=7873 (accessed: Apr. 07, 2021) 652 Hannula, J., Godinho, J.R.A., Llamas, A.A., Luukkanen, S., Reuter, M.A., 2020. Simulation653 Based Exergy and LCA Analysis of Aluminum Recycling: Linking Predictive Physical 654 Separation and Re-melting Process Models with Specific Alloy Production. Journal of 655 Sustainable Metallurgy, vol. 6, no. 1, pp. 174–189. https://doi.org/10.1007/s40831-020656 00267-6. 657 Hatayama, H., Daigo, I., Matsuno, Y., Adachi, Y., 2008. Assessment of Recycling Potential 658 of Aluminium in Japan, the United States, Europe and China. Nihon Kinzoku Gakkai shi, 659 vol. 72, no. 10, pp. 812–818. https://doi.org/10.2320/jinstmet.72.812. 660 Hatayama, H., Daigo, I., Matsuno, Y., Adachi, Y., 2012. Evolution of aluminum recycling 661 initiated by the introduction of next-generation vehicles and scrap sorting technology. 662 Resources, conservation and recycling, vol. 66, pp. 8–14. 663 https://doi.org/10.1016/j.resconrec.2012.06.006. 664 Hatayama, H., Yamada, H., Daigo, I., Matsuno, Y., Adachi, Y., 2006. Dynamic Substance 665 Flow Analysis of Aluminum and Its Alloying Elements. Nihon Kinzoku Gakkai shi, vol. 666 70, no. 12, pp. 975–980. https://doi.org/10.2320/jinstmet.70.975. 667 Kinsey, B., Wu, X., 2011. Tailor Welded Blanks for Advanced Manufacturing, 1st ed. 668 Woodhead Publishing, 2011. https://www.elsevier.com/books/tailor-welded-blanks-for669 advanced-manufacturing/kinsey/978-1-84569-704-4 (accessed: Apr. 09, 2021) 670 Lamifil, 2019. Specialty wires. https://lamifil.be/wp-content/uploads/2016/04/04-10-2019671 specialities-brochure-lr.pdf (accessed: Apr. 09, 2021) 672 Lamifil, 2020. Aluminium alloys - Lamifil. https://lamifil.be/products/aluminium-alloys/ 673 (accessed: Apr. 09, 2020) 674 35 Lupton, R.C., 2017. ricklupton/sankeyview: v1.1.7. Cambridge, United Kingdom. 675 https://github.com/ricklupton/floweaver (accessed: Apr. 09, 2021) 676 Lupton, R.C., Allwood, J.M., 2017. Hybrid Sankey diagrams: Visual analysis of 677 multidimensional data for understanding resource use. Resources, conservation and 678 recycling, vol. 124, pp. 141–151. https://doi.org/10.1016/j.resconrec.2017.05.002. 679 MakeItFrom, 2020. Aluminum Alloys :: MakeItFrom.com. 680 https://www.makeitfrom.com/material-group/Aluminum-Alloy (accessed: Apr. 09, 681 2021). 682 Matweb, 2020. Aluminum 4043-H18. 683 http://www.matweb.com/search/datasheet_print.aspx?matguid=2541bb0127b34c7294612 684 c77393ead8c (accessed: Apr. 09, 2021) 685 Modaresi, R., Lovik, A.N., Müller, D.B., 2014. Componentand Alloy-Specific Modeling for 686 Evaluating Aluminum Recycling Strategies for Vehicles. JOM, vol. 66, no. 11, pp. 2262– 687 2271. https://doi.org/10.1007/s11837-014-0900-8. 688 Modaresi, R., Müller, D.B., 2012. The Role of Automobiles for the Future of Aluminum 689 Recycling. Environmental science & technology, vol. 46, no. 16, pp. 8587–8594. 690 https://doi.org/10.1021/es300648w. 691 Nakajima, K., Takeda, O., Miki, T., Matsubae, K., Nakamura, S., Nagasaka, T., 2010. 692 Thermodynamic Analysis of Contamination by Alloying Elements in Aluminum 693 Recycling. Environmental science & technology, vol. 44, no. 14, pp. 5594–5600. 694 https://doi.org/10.1021/es9038769. 695 NTNU, International Aluminium Institute, TruthStudio, Davies, J., 2020. Global Aluminium 696 Cycle. IAI. https://alucycle.world-aluminium.org/public-access/#global (accessed: Apr. 697 08, 2021) 698 36 Paraskevas, D., Kellens, K., Dewulf, W., Duflou, J.R., 2015. Environmental modelling of 699 aluminium recycling: a Life Cycle Assessment tool for sustainable metal management. 700 Journal of Cleaner Production, vol. 105, pp. 357–370. 701 https://doi.org/10.1016/j.jclepro.2014.09.102. 702 Peng, T., Ou, X., Yan, X., Wang, G., 2019. Life-cycle analysis of energy consumption and 703 GHG emissions of aluminium production in China. Energy Procedia, vol. 158, pp. 3937704 3943. https://doi.org/10.1016/j.egypro.2019.01.849. 705 Prasad, N.E., Wanhill, R.J.H., 2016. Aerospace Materials and Material Technologies: 706 Volume 1: Aerospace Materials. Singapore: Singapore: Springer Singapore Pte. Limited. 707 https://link-springer-com.kuleuven.ezproxy.kuleuven.be/book/10.1007%2F978-981-10708 2134-3 (accessed: Apr. 09, 2021) 709 Saevarsdottir, G., Kvande, H., Welch, B.J., 2020. Aluminum Production in the Times of 710 Climate Change: The Global Challenge to Reduce the Carbon Footprint and Prevent 711 Carbon Leakage. JOM, vol. 72, no. 1, pp. 296–308. https://doi.org/10.1007/s11837-019712 03918-6. 713 Salazar, K., McNutt, M.K., 2010. Minerals Yearbook - Metals and Minerals 2010, vol. 1. 714 Eatsern Region, Reston, Va., USA: Department of the interior. Bureau of mines. 715 Schlesinger, M.E., 2017. Aluminum Recycling, 2nd ed. CRC Press. 716 https://www.routledge.com/Aluminum-Recycling/Schlesinger/p/book/9781138073043 717 (accessed: Apr. 07, 2021) 718 Soo, V.K., Peeters, J., Paraskevas, D., Compston, P., Doolan, M., Duflou, J.R., 2018. 719 Sustainable aluminium recycling of end-of-life products: A joining techniques 720 perspective. Journal of Cleaner Production, vol. 178, pp. 119-132. 721 https://doi.org/10.1016/j.jclepro.2017.12.235. 722 37 Southwire, 2020. Products | Southwire Overhead Transmission. 723 https://overheadtransmission.southwire.com/products/ (Accessed: Apr. 09, 2020) 724 The Aluminum Association, 1989. Aluminum Electrical Conductor Handbook, 3rd ed. 725 Washington, DC, USA: Aluminum Association. 726 https://www.aluminum.org/sites/default/files/Aluminum%20Electrical%20Conductor%2 727 0Handbook.pdf (accessed: Jun. 06, 2021) 728 The Economist, 2007. The price of virtue. 729 https://www.economist.com/leaders/2007/06/07/the-price-of-virtue (accessed: Apr. 07, 730 2021) 731 The World Bank, 2021. GDP growth (annual %) | Data. 732 https://data.worldbank.org/indicator/ny.gdp.mktp.kd.zg (accessed: Apr. 09, 2021). 733 Van Heusden, R., Morrison, H., Puleo M., 2020. Why addressing the aluminium industry’s 734 carbon footprint is key to climate action, World Economic Forum. 735 https://www.weforum.org/agenda/2020/11/the-aluminium-industry-s-carbon-footprint-is736 higher-than-most-consumers-expect-heres-what-we-must-do-next/ (accessed: Apr. 07, 737 2021) 738 Zhu, Y., Cooper, D.R., 2019. An Optimal Reverse Material Supply Chain for U.S. Aluminum 739 Scrap. Procedia CIRP, vol. 80, pp. 677–682. https://doi.org/10.1016/j.procir.2019.01.065. 740 Zhu, Y., Chappuis, L.B., De Kleine, R., Kim, H.C., Wallington, T.,J., Luckey, G., Cooper, 741 D.R., 2021. The coming wave of aluminum sheet scrap from vehicle recycling in the United 742 States. Resources, Conservation and Recycling, vol. 164. 743 https://doi.org/10.1016/j.resconrec.2020.105208. 744