WRAP MDD018/23 WEEE Separation techniques. Visys Multi Frequency Laser trial report

Transcription

1 WRAP MDD018/23 WEEE Separation techniques Visys Multi Frequency Laser trial report Abstract This report details a trial conducted on the Spyder multi laser separator at Visys for WRAP project MDD018. The aim of the project was to trial innovative techniques to tackle some of the more difficult separations encountered by primary and secondary WEEE processors. A number of sensor based technologies have been trialled during the project to evaluate their ability to identify and segregate a number of material streams, including circuit boards, various polymer types and other impurities. The Spyder laser sorter uses multi frequency lasers to sort material into two fractions. The aim of the trial was to test the machine s ability to remove printed circuit boards (PCBs) from a number of samples containing printed circuit boards (PCBs), plastic, metal and other minor fractions of material. Four different samples of material were tested. The first was a sample of mixed WEEE plastic supplied by SWEEP, a primary WEEE processor. This material was tested in the form it arrived from SWEEP with no pre-processing. The second and third samples were SWEEP WEEE material that had been partially processed at Axion Polymer s plant at Salford, in order to separate a magnetic fraction using rare earth magnets and a non-ferrous fraction using an eddy current separator. The fourth sample was a specially made mix of the SWEEP WEEE plastic, with some handpicked circuit boards sourced from other WEEE material sourced from S Norton & Co. Each of the samples was processed by the Visys laser sorter with the aim being to remove the circuit boards and create a pure plastic fraction. The results from the trial varied for each test run as the type and quantity of PCBs in each sample differed. The machine demonstrated an ability to remove 99% of the PCBs from the fourth sample. With the other samples, although not all the PCBs were removed, the results were still promising. One issue arising from the trial is the PCB concentration of the eject fraction. This tended to be in the 50-60% range. The PCB product would be more valuable if this concentration could be increased. To achieve a higher PCB concentration in the eject stream there would need to be some additional work undertaken, either in the form of fine tuning the machine or a second pass of the PCB fractions.

2 As the technical results of the trial were promising, an economic assessment of the machine was completed. The evaluation showed that the payback time is highly dependent on the amount of PCBs removed from the feed and the value of the PCB fraction. In a worst case scenario where 5% is removed from the WEEE material feed as PCBs and the PCB fraction is worth 40 per tonne then the payback time is 274 months, nearly 23 years. However if 10% of the stream is PCBs and the fraction is worth 250 per tonne, then the payback time is 23 months which is much more promising and would potentially be of interest to a WEEE processor. The higher PCB value would normally only be achieved by processors handling a large proportion of computers where the circuit boards have a greater precious metal content. In conclusion the Visys Spyder machine demonstrated a strong technical performance but the detail of the economic case will affect how viable the technology would be in the WEEE recycling sector. 2

3 Table of Contents Abstract Information from Trial Description of Trial Equipment Photographs of the trial equipment Trial Objectives Sample Material Summary of Trial Sorts Trial 1 - Removal of PCBs from a mixture of plastic, metal and PCBs Feed Material Trial Photograph of result samples Analysis of results samples Discussion of results Conclusions from trial Trial 2 - Removal of PCBs from a magnetic fraction Feed Material Results Photograph of result samples Analysis of results samples Discussion of results Conclusions from trial Trial 3 - Removal of PCBs from a mixture of plastic, metal and PCBs Feed Material Trial Photograph of result samples Analysis of results samples Discussion of results Conclusions from trial Trial 4 - Removal of PCBs from non ferrous material Feed Material Results Photograph of result samples

4 Analysis of results samples Discussion of results Conclusions from trial Trial 5 - Removal of PCBs from plastic - PCB mixture, separation of plastics by colour Feed Material Results Photograph of result samples Analysis of results samples Discussion of results Conclusions from trial Economic assessment of the technique Overall conclusions of the trial

5 List of Figures Figure 1: Schematic of the Spyder Laser Sorter (courtesy of Visys)... 6 Figure 2: Photograph of the trial equipment... 9 Figure 3: Photograph of the trial equipment showing a side view of the system Figure 4: Photograph of the trial equipment Figure 5: Schematic indicating the feed samples for the trial Figure 6: Photograph of Trial 1 PCB eject fraction Figure 7: Photograph of Trial 1 Plastic reject fraction Figure 8: Schematic of trial 1 results Figure 9: Photograph of Trial 2 PCB Eject Fraction Figure 10: Photograph of Trial 2 Plastic/Metal Reject Fraction Figure 11: Schematic of the trial 2 results Figure 12: Photograph of Trial 3 PCB eject fraction Figure 13: Photograph of Trial 3 plastic reject fraction Figure 14: Schematic of the trial 3 results Figure 15: Photograph of Trial 4 PCB eject fraction Figure 16: Photograph of Trial 4 plastic/metal reject fraction Figure 17: Schematic of trial 4 results Figure 18: Photograph of Trial 5 plastic reject from sort Figure 19: Photograph of Trial 5 sort 1 PCB eject fraction Figure 20: Photograph of Trial 5 sort 3 eject - black plastic Figure 21: Photograph of Trial 5 sort 3 reject - white plastic Figure 22: Photograph of Trial 5 sort 3 reject white plastic reprocessed into different white fractions Figure 23: Schematic for trial 5 sort 1 results - first pass of plastic/pcb mixture, ejecting PCBs Figure 24: Schematic of trial 5 sort 2 results - sort 1 eject second pass, ejecting PCBs Figure 25: Schematic of trial 5 sort 3 results - sort 1 and 2 plastic fractions, ejecting black plastic List of Tables Table 1: Summary of the trial sorts Table 2: Hand sorting results for trial 1 fractions Table 3: Q and R separation efficiencies for trial Table 4: Hand sorting results for trial 2 fractions Table 5: Q and R separation efficiencies for trial Table 6: Hand sorting results for trial 3 fractions Table 7: Q and R separation efficiencies for trial Table 8: Hand sorting results for trial 4 fractions Table 9: Q and R separation efficiencies for trial Table 10: Hand sorting results for trial 5 fractions Table 11: Q and R separation efficiencies for trial 5 sort Table 12: Q and R separation efficiencies for trial 5 sort Table 12: Overall Q and R separation efficiencies for trial

6 Table 13: Payback calculation for a Visys laser sorter separating PCBs Information from trial Trial host: Visys NV. The trial was conducted at a test facility in Hasselt, Belgium. Trial equipment: Spyder Digital Laser Sorter Trial date: 24 th February Description of trial equipment The Spyder Digital Laser Sorter is a fully digital advanced machine, which has a number of uses in a range of industries including food processing and recycling. The machine can identify and remove materials such as glass, stones, wood, metals and plastics. The system can sort material based on colour, shape and material type. A sort can be done on one of these factors or by combining a number of them together. Materials such as glass, stone, wood, metals and plastics can be identified and removed by the technology. Figure 1 is a schematic of the sorter which is followed by a detailed description. Figure 1: Schematic of the Spyder Laser Sorter (courtesy of Visys) Material is fed to the system by a vibratory feeder. The length of the feeder on this machine is an important factor, to ensure that the material is well spread out when it reaches the start of the chute. The material falls vertically of the end of the feeder and down between the lasers. The chute which the material passes down is a patented idea known as the chicane slide. For the trial the vibratory feeder was fed by hand but in a production plant it would be fed by a hopper or conveyor. 6

7 There are two laser units within the system, at the front and back, which are controlled by the user interface panel. Each laser creates a single beam of light which is directed by a static mirror onto a further set of spinning mirrors. The mirrors cause the laser beam to scan back and forth across a reference bar that is mounted on the far side of the chute from the laser source. The particles to be scanned fall down the chute in front of the reference bar. There is a front and back reference bar for the front and back lasers to shine on. The material falls between the two reference bars and it is at this point that the lasers scan the front and back of the material/particles. Different types of material give different reflective signals. The signal from the laser reflects back into the laser unit, bounces off the rotating mirrors and into the detection system. The detection system contains the optics and photomultiplier which is the laser light receiver. The photomultiplier converts the laser signal into an electrical signal which can then be displayed on the user interface screen. The Visys Spyder unit can be fitted with a front or back laser unit, or both, to scan one or both sides of each particle. Each laser unit can contain up to five lasers. These are positioned slightly apart so that the individual signals do not interfere with each other. The laser unit chamber is sealed and positively pressurised so that dust and dirt will not be able to enter the unit and damage the lasers. The lasers also have a heating/cooling system to ensure that they operate continuously at the right temperature. There are different types of signals which can be detected by the system. These are scattering, anti-scattering and normal reflection signals. Scattering is the diffusion around the laser point. Anti-scattering is the opposite of this and focuses on the signal from the laser point. Normal reflection is the standard light reflected from an object when a laser is shone on it. Different laser light frequencies can be detected by the machine. These include infra-red, red, green and blue. A decision must be made as to which signal should be used by the machine to determine if a material is to be identified and ejected. For example the red reflection signal for two objects may be very similar but the infra red antiscattering signal may be different, so this would be the one used to determine which material to eject. The user sets a threshold value via the control panel and the signal from the object is compared to the threshold value. The system can be set to eject material which has a signal above or below this threshold value. Once the material has passed through the laser beam and it has been identified, it is either ejected by the air jets or left to fall into the reject collection bins. The unit can be fitted with up to 134 air jets across the width of the chute. The ejected material falls into a vibrating channel which sends the material into a collection bin. The air jets can be set up to fire based on an object s shape. For an odd shaped object the air jets follow the shape of the object to ensure it is successfully removed. The system should be able to process up to 4 tonnes per hour of WEEE and this can increase to 6-7 tonnes per hour with certain types of material. The system can handle particles as small as 3mm but this will reduce the throughput of the unit. The unit typically works best on particles of 5mm or greater in size. 7

8 A Visys Spyder unit costs approximately 180,000 for a single laser unit and 250,000 for a unit with two lasers, which can detect signals on both sides of each particle simultaneously. The Visys Spyder system can be set up to operate in two different ways. The first option is to remove as much of a specific material as possible. This approach will create an almost pure reject fraction and an eject fraction which may be contaminated with some of the reject material. For example, the unit could be set up to remove as many PCB particles as possible in order to produce a pure plastics fraction. The second approach is to create as pure an eject fraction as possible. This may mean some material is missed and ends up in the reject stream. For example the PCB s could be identified and removed to create a pure eject fraction. The decision on which approach to take depends on whether it is more useful to maximise removal of the target material from the reject fraction or to maximise the purity of the eject fraction. It should be noted that the Visys Spyder system requires a degree of fine tuning to achieve the desired results. Identifying the best settings for the equipment is entirely dependent on the type of material being processed. When a Visys machine is installed, an engineer will often need to work with the unit for at least a few days, until the system is running smoothly. If the feed material changes the system set up will need adjusting accordingly. The Visys Spyder machine can sometimes have difficulties in identifying materials, if a confusing signal is received. For example the machine can separate material based on colours of particles and so can separate black and white streams. White material gives a very clear signal with a peak, whilst black material gives a signal with a trough. However if a piece of black plastic has been bent or scratched, so that it has white marks, it will give a mixed signal of peaks and troughs as the laser scans across it. This can result in the machine misidentifying the material. However the machine does have an advantage over other sensor based sorting techniques, in that it can see black particles in the first place. The high intensity of the laser light source is able to generate a weak reflection from even very dark particles. 8

9 Control panel 1.2 Photographs of the trial equipment Chycane slide chute Manual Feed Point Front laser chamber unit Vibratory feeder Reject collection bins Rear laser chamber unit Figure 2: Photograph of the trial equipment 9

10 Slide chute Front laser unit Rear laser unit Laser beam Reference bars Air jets located beneath front reference bar Figure 3: Photograph of the trial equipment showing a side view of the system Control panel Reject vibrating channel into collection bin Figure 4: Photograph of the trial equipment 10

11 1.3 Trial objectives The objective for the trial was to test the Visys machine s ability to identify and remove printed circuit boards (PCBs) from a mixed WEEE plastic stream. The aim for the trial was to remove the circuit boards. However, the separation strategy was dependent on which product fraction was more valuable and important to recover. Typically in the WEEE reprocessing sector, it is advantageous to remove as many circuit boards as possible in order to aid further processing of the plastic stream. 1.4 Sample material Four samples of material were used in this trial: 1) Mixed WEEE plastics from SWEEP, a leading UK primary WEEE processor as produced by SWEEP with no additional processing at Axion Polymer s Salford plant; 2) The magnetic fraction from processing the SWEEP WEEE over a rare earth magnet; 3) The non-ferrous fraction from processing the SWEEP WEEE over an eddy current separator; and 4) A mixture of the bulk plastic from SWEEP WEEE after the eddy current separation system but with the addition of extra PCBs to enrich the PCB content of the material and challenge the separator (plastic with additional PCBs). Samples 2, 3 and 4 were produced using rare earth magnets and an eddy current separator at Axion s plant in Salford to remove ferrous and non-ferrous metals. Both the rare earth magnet and the eddy current separator tend to remove part of the PCBs in the feed. Figure 5 shows which part of the process each sample was sourced from. All the material was over 15mm in size, with some pieces above 50mm. SAMPLE 1 Sweep WEEE as it is Rare Earth Magnet SAMPLE 2 Magnetic Fraction Eddy Current System SAMPLE 3 Non ferrous Fraction SAMPLE 4 Bulk plastic with additional PCBs Figure 5: Schematic indicating the feed samples for the trial 11

12 1.5 Summary of trial sorts Table 1 is a summary of each of the individual sorts conducted during the trial. It shows the feed material, which material was ejected and what was left in the reject fraction. Trial Input fraction Output fractions Eject Reject Material Material Material 1 Plastic with additonal PCBs PCBs Plastic 2 Rare earth magnet fraction PCBs Plastic/Metals 3 Plastic with additonal PCBs PCBs Plastic 4 Non ferrous fraction PCBs Plastic/Metals 5 Sort 1 Sweep WEEE PCBs Plastic 5 Sort 2 Trial 5 sort 1 eject PCBs Plastic 5 Sort 3 Trial 5 sort 1 reject and trial 5 sort 2 reject (plastic) Black White Table 1: Summary of the trial sorts 12

13 2.0 Trial 1: Removal of PCBs from a mixture of plastic, metal and PCBs 2.1 Feed material The feed material for this trial was a specially made mix of SWEEP WEEE plastic and handpicked PCBs sourced from WEEE plastic material produced by S Norton & Co in Manchester. The PCB content of the sample after blending was 15% by weight. This sample was made because the SWEEP WEEE received by Axion at Salford contains very few PCBs, approximately 2%. SWEEP use magnets and eddy current separators in their own process which tend to remove a large proportion of the circuit boards. The aim was therefore to create a fraction with a high PCB content to challenge the machine. 2.2 Trial 1 The first trial was an initial test of the system to see how well the PCBs could be removed and to allow the Visys engineers to conduct some initial set up tests of the system. The unit was configured to eject the PCBs and create a pure plastic fraction in the reject stream. The red reflection signal and the infra-red anti-scattering signal were used to determine which material to eject. 2.3 Photograph of result samples Figure 6: Photograph of Trial 1 PCB eject fraction There were a handful of pieces of plastic in the eject fraction, which can be seen in Figure 6, along with a few items such as copper wire lump and rubber tubing. The eject fraction appeared to contain a high concentration of circuit boards. 13

14 Figure 7: Photograph of Trial 1 Plastic reject fraction There were a few PCBs in the plastic fraction. A visual assessment indicated that a good separation had been achieved. Increasing the sensitivity of the machine to remove more PCBs would mean the plastic stream could be purified further, but the loss of good plastic to the PCB fraction would increase. 14

15 2.4 Analysis of results samples The eject and reject fractions were both hand sorted to determine the compositions of the fraction. Table 2 shows the results of the hand sorts. Weight Trial Fraction Material of fraction Weight of samples Sample composition PCBs Plastic Metal Wood PVC Wires kg g g % g % g % g % g % g % g % g % g % g % T1 Eject PCBs % % 0.0 0% % % % % % % % T1 Reject Plastic % % 0.0 0% % 188 5% % 139 4% % % 131 4% Copper Wires Table 2: Hand sorting results for Trial 1 fractions Rubber Stone/ Glass Lights/ Foils/ Foams Others

16 Trial 1 Eject PCBs Plastic with additional PCBs kg % Total % PCBs % Plastic % Metal % Wood % PVC Wires % Feed kg % Copper Wires % Total % Rubber % PCBs % Stone/Glass % Plastic % Lights/Foils % Metal % Others % Wood % Visys Laser Sorter PVC Wires % Trial 1 Copper Wires % Reject Rubber % Loss 0.04 kg % Stone/Glass % Total % Lights/Foils % PCBs % Others % Plastic % Metal % Wood % PVC Wires % Copper Wires % Rubber % Stone/Glass % Lights/Foils % Others % Figure 8: Schematic of Trial 1 results 2.5 Discussion of results None of the feed material was kept for hand sorting after the trial by mistake and therefore the feed composition in Figure 8 is back calculated from the eject and reject composition. The results indicate that the feed contained approximately 15% circuit boards. The eject fraction contained 56% circuit boards and the reject stream contained only 0.2% circuit boards. Therefore the machine was able to remove nearly all the circuit boards from the feed material. This was the intention of this first trial as the machine had been set up to create a pure plastic fraction. The reject fraction contained 83% plastic which was an increase from the starting feed material which contained 70% of plastic. The polymer loss to the PCB eject fraction is low but a question could be raised about whether the eject fraction has any value with a composition of 56% PCBs. The PCB eject fraction probably needs to contain a higher composition of PCBs in order to increase the value of this stream and to make the processing worthwhile. It may be that in its current

17 form the material could be hammer milled to recover the metals and the PCB level is acceptable. The loss of material during the separation was very low. Axion s Q and R convention has been used to measure the success of the separation. In this case: The product separation efficiency Q = the probability that PCBs are correctly sorted into the eject fraction.; and The reject separation efficiency R = the probability that non PCB material is correctly sorted into the reject fraction. Q R Trial 1 99% 87% Table 3: Q and R separation efficiencies for Trial 1 For this trial product separation efficiency, Q, was 99% which is very high for a sorting machine and the reject separation efficiency, R, is lower at 87%. This is because the machine was set up to allow some loss of non PCB material into the eject fraction in order to ensure maximum removal of PCBs. It was observed during the trial that there were some small PCBs and pieces of metal in the feed material. Ideally the material should be screened to remove these small pieces as they could not be detected by the sorting machine. During the trial the PCBs which ended up in the reject stream were picked out by hand and held in front of the lasers to determine if the machine could identify them. The air jets were activated, indicating the PCBs were detected and hence should not have been in the eject fraction. It is thought that there are two possible explanations for these mis-sorts. The first is that the air jets fired at the wrong time and missed the PCB material completely. This could be because the air jets were activated too early or too late. By adjusting the timing of the air jets this may result in improved removal of the PCBs. The second possible explanation is a mechanical issue with the machine. Some of the particles bounced off the splitter plate after they had been ejected and fell back into the reject stream. It appeared that the gap through which the material had to fall after ejection was not large enough. This caused some of the material to hit the walls and bounce back. The splitter plate was designed in this way by Visys so that as the material lands on the sides of the splitter plate the kinetic energy of the particles is reduced. Reducing the speed of the particles cuts wear on other parts of the machine. If the splitter plate on the eject side could be realigned this could potentially reduce the chance of particles bouncing back. 17

18 Visys is aware of this issue with the design of their splitter plate. It has only been observed when processing waste materials for recycling. They believe that a simple change should alleviate this problem. 2.6 Conclusions from trial The trial was a success as 99% of the circuit boards were removed from the feed fraction leaving a circuit board-free reject fraction. The amount of non PCB material which was ejected in the eject fraction was quite high, which meant that the PCB concentration in the eject fraction was only 56%. It may be cost-effective to re-run the eject fraction through the machine in order to upgrade it and obtain a higher value for the recovered material. 18

19 3.0 Trial 2: Removal of PCBs from a magnetic fraction 3.1 Feed material The feed material for this trial was the magnetic fraction extracted by the rare earth magnet when the SWEEP WEEE was processed at Axion s Salford plant. It contained around 30% PCBs by weight, roughly double the content of the feed material for trial Results As with trial 1, the machine was set up to remove the PCBs and create a clean reject fraction. The observation during the trial was that the material processed well and eject fraction contained a large amount of PCBs. 3.3 Photograph of result samples Figure 9: Photograph of Trial 2 PCB Eject Fraction The photograph in Figure 9 clearly shows a high concentration of PCBs. 19

20 Figure 10: Photograph of Trial 2 Plastic/Metal Reject Fraction The photograph in Figure 10 shows mainly plastic and metal pieces with a few pieces of green PCB visible. 20

21 3.4 Analysis of results samples As with trial 1 each of the product fractions were hand sorted to determine the fraction composition. The results are shown in Table 4. Trial Fraction Material Weight of fraction Weight of samples Sample composition Lights/ Copper Stone/ PCBs Plastic Metal Wood PVC Wires Rubber Foils/ Others Wires Glass Foams kg g g % g % g % g % g % g % g % g % g % g % T2 Eject PCBs % % % % % % % 0.0 0% 111 2% 0.0 0% T2 Reject Plastic/Metal % % % 146 2% 141 2% % 170 2% % % % Table 4: Hand sorting results for Trial 2 fractions

22 Trial 2 Eject PCBs Rare earth magnetic fraction kg % Total % PCBs % Plastic % Metal % Wood % PVC Wires % Feed kg % Copper Wires % Total % Rubber % PCBs % Stone/Glass % Plastic % Lights/Foils % Metal % Others % Wood % Visys Laser Sorter PVC Wires % Trial 2 Copper Wires % Reject Rubber % Loss 0.04 kg % Stone/Glass % Total % Lights/Foils % PCBs % Others % Plastic % Metal % Wood % PVC Wires % Copper Wires % Rubber % Stone/Glass % Lights/Foils % Others % Figure 11: Schematic of the Trial 2 results Q R Trial 2 69% 84% Table 5: Q and R separation efficiencies for Trial Discussion of results The feed composition was back calculated from the eject and reject fractions, which gave a feed composition of 29% PCBs. The eject fraction contained 64% PCBs, whilst the reject stream had 13% PCBs. Having a higher PCB concentration in the eject fraction means that it is less likely to require re-working. But the amount of PCB s remaining in the reject is enough to mean that it may require another pass in order to remove a greater proportion of the PCBs.

23 The product separation efficiency, Q, for this trial was lower than for trial 1 at 69% and the reject separation efficiency, R, was similar at 84%. This feed material contained a lot more metal than the previous sample. It may be that this had some effect on the ability of the machine to identify the PCBs. The feed material was rather dirty and could be considered to be a worst case scenario for WEEE material. Ideally the material should have been screened at 15mm and de-dusted in order to obtain the best results from the machine. 3.6 Conclusions from trial Overall the second trial did demonstrate that the PCBs could be removed from a PCBcontaining fraction that had been separated by rare earth magnet. 69% of the PCBs were removed to the eject fraction at a concentration of 64%. This type of material was harder for the machine to sort into valuable streams. It could be that the higher PCB content in the feed (30%) meant that air jets were unable to keep up with the removal rate required. Alternatively the machine set-up may not have been quite right for the material and further fine tuning of the equipment was required. This was not possible in the time available during the trial. 23

24 4.0 Trial 3: Removal of PCBs from a mixture of plastic, metal and PCBs 4.1 Feed material The feed material for trial 3 was a specially made mix of bulk WEEE plastic with extra PCBs added to the material to give an average PCB content in the feed of 15% by weight. The original material prior to the addition of the PCBs on had a 2% PCB content. This was the same material as used in trial 1. Trial 1 was just a test to configure the machine, whilst the aim of trial 3 was to process the sample to obtain a throughput result. 4.2 Trial procedure The system was set up as with the first two trials in order to remove all the PCBs from the plastic fraction. 4.3 Photograph of result samples Figure 12: Photograph of Trial 3 PCB eject fraction Figure 12 shows a photograph of the eject fraction from trial 3. The PCBs are clearly visible but so are quite a few pieces of plastic. 24

25 Figure 13: Photograph of Trial 3 plastic reject fraction The photograph in Figure 13 shows the plastic reject fraction. It appears to contain very few pieces of circuit board and the majority of the sample is plastic. 25

26 4.4 Analysis of results samples The results from hand sorting of the eject and reject fractions from trial 3 are shown in Table 6. Weight Trial Fraction Material of fraction Weight of samples Sample composition Lights/ Copper Stone/ PCBs Plastic Metal Wood PVC Wires Rubber Foils/ Others Wires Glass Foams kg g g % g % g % g % g % g % g % g % g % g % T3 Eject PCBs % % % 147 1% 236 2% % % 131 1% % % T3 Reject Plastic % % % % 146 4% % % % 0.0 0% 0.0 0% Table 6: Hand sorting results for trial 3 fractions

27 Trial 3 Eject PCB's Plastic with additional PCBs kg % Total % PCBs % Plastic % Metal % Wood % PVC Wires % Feed kg % Copper Wires % Total % Rubber % PCBs % Stone/Glass % Plastic % Lights/Foils % Metal % Others % Wood % Visys Laser Sorter PVC Wires % Trial 3 Copper Wires % Reject Rubber % Loss 0.0 kg % Stone/Glass % Throughput 930 kg/hr Total % Lights/Foils % PCBs % Others % Plastic % Metal % Wood % PVC Wires % Copper Wires % Rubber % Stone/Glass % Lights/Foils % Others % Figure 14: Schematic of the Trial 3 results Q R Trial 3 92% 86% Table 7: Q and R separation efficiencies for Trial Discussion of results This trial used the same material as was tested in the trial 1 sort. Therefore similar results should be expected and indeed the results do show similarities. The back calculated PCB feed composition is 13% for trial 3, which compares well with the trial 1 value of 15%. The majority of the PCBs were removed from the feed with a separation efficiency of 92%, leaving only 1% PCBs in the reject fraction. However the content of PCBs in the eject fraction was just under 50%. As for trial 1, the PCB content of the eject fraction may be too low for it to have a worthwhile end value. A second pass through the machine may upgrade it to a more saleable PCB content. The plastic content of the reject fraction was 88%, compared to 77% in the feed.

28 For this trial the throughput was measured at 930 kg per hr, however it is believed that a higher throughput of up to tonnes per hour should be possible on the machine, based on the experience of Visys with this type of material. The product separation efficiency, Q, for this sort was 92% which is very good and would be suitable for a commercial separation. The reject separation efficiency, R, of 86% is also good. These values are line with the separation efficiencies measured in trial 1. During the trial it was noted that there was an issue of the bounce back of material because of the design of the splitter plate, as discussed in an earlier section of this report. An important observation during the trial was that the particle size of the material in the reject fraction was smaller than the particles in the eject fraction. Screening the feed material to remove the fines would assist the sorter as the material being processed would all be within the size range the air jets can eject. Processing tight size classifications may improve the separation efficiency. 4.6 Conclusions from trial The trial demonstrated that the machine is able to sort PCBs from the other material in the feed and produce a reject fraction with a low PCB content. The machine was set up with this objective as it is important to have a PCB free plastic fraction for further processing. This resulted in an eject fraction which included around 50% non PCB. However by reworking the eject fraction or fine tuning the system, it should be possible to achieve a final PCB product fraction containing a higher level of PCBs. 28

29 5.0 Trial 4: Removal of PCBs from non-ferrous material 5.1 Feed material The feed material for trial 4 was the non-ferrous fraction from processing SWEEP WEEE material through the eddy current separator at Axion s plant in Salford. This produced a non-ferrous fraction containing around 16% PCBs by weight. 5.2 Results No changes were made to the system set up for trial 4. As for the other trials it was run using the red reflection signal and the IR scattering signal. Although the throughput was not measured for this sort, it was observed that it took longer to process this material than the previous sorts, as it consisted of smaller particles. It was observed during the trial was that there was some enrichment of the PCB fraction but the separation was not as good as achieved in the previous trials. 5.3 Photograph of result samples Figure 15: Photograph of Trial 4 PCB eject fraction The eject fraction photograph shows a high proportion of circuit boards but there are a number of other items present. 29

30 Figure 16: Photograph of Trial 4 plastic/metal reject fraction A few circuit boards can be seen in the reject photograph along with a wide range of other items including metal and plastic. 30

31 5.4 Analysis of results samples The results from the hand sorts of the trial 4 fractions are shown in Table 8. Trial Fraction Material Weight of fraction Weight of samples Sample composition PCBs Plastic Metal Wood PVC Wires Copper Rubber Stone/ Lights/ Others kg g g % g % g % g % g % g % g % g % g % g % T4 Eject PCBs % % % 0.0 0% 0.0 0% % % % % % T4 Reject Plastic/Metal % % % % % % % % 0.0 0% 0.0 0% Table 8: Hand sorting results for Trial 4 fractions

32 Trial 4 Eject PCBs Non Ferrous fraction kg % Total % PCBs % Plastic % Metal % Wood % PVC Wires % Feed kg % Copper Wires % Total % Rubber % PCBs % Stone/Glass % Plastic % Lights/Foils % Metal % Others % Wood % Visys Laser Sorter PVC Wires % Trial 4 Copper Wires % Reject Rubber % Loss 0.01 kg % Stone/Glass % Total % Lights/Foils % PCBs % Others % Plastic % Metal % Wood % PVC Wires % Copper Wires % Rubber % Stone/Glass % Lights/Foils % Others % Figure 17: Schematic of Trial 4 results Q R Trial 4 66% 89% Table 9: Q and R separation efficiencies for Trial Discussion of results The back calculated feed composition for this sort was 16% PCBs. The separation produced an eject fraction with a composition of 52% PCB and a reject fraction with 7% PCBs. The results of this trial are similar to the results from trial 2, where some but not all of the PCBs were removed by the process. Therefore, both of the fractions may require further work to produce a plastic fraction with a lower PCB content and a PCB fraction with less plastic. It may simply be a case that the machine could be further tuned and these fractions could be achieved in the first pass. The product separation efficiency, Q, for this trial is 66%, which is lower than the previous trials and means that a third of the PCB material was not correctly separated. The reject separation efficiency, R, at 89% was similar to the other trials. As for the other trials this

33 resulted in significant loss of non PCB materials into the PCB product fraction and dilution of the PCB content to 52%. A second pass of the product fraction through the machine may be required to upgrade the PCB product to a commercially viable value. 5.6 Conclusions from trial The trial showed that some 66% of the PCBs were removed from the eddy current separated non-ferrous fraction but a significant amount was left in the reject fraction. 33

34 6.0 Trial 5: Removal of PCBs from plastic - PCB mixture, separation of plastics by colour 6.1 Feed material The feed material for the fifth trial was SWEEP WEEE plastic fraction as received in Salford, with no pre-processing undertaken on the material. The feed material contained about 2% PCBs. The plastic in the feed was a mix of black and white particles. The original trial objective of PCB removal was tested along with a second objective which was to separate the plastics in the fraction by a colour sort into black and white fractions. 6.2 Results Several runs were conducted on the SWEEP WEEE material to find the best setting for the machine. The initial plan was test the black/white plastic separation. The aim was to remove all the white plastic into the eject fraction, leaving the black and dark grey plastic in the reject. After testing various pieces of plastic the system was set up using the red reflection signal and the green anti-scattering signal. Although this separation was possible an issue were encountered. The material was rather dirty; it is thought that the dirt came from computer printer ink cartridges which were not removed prior to shredding the WEEE material. The shredding process released the ink from the cartridges and contaminated the material. This created a dark layer on the plastic material and the white particles became grey. This may have had an effect on the ability of the machine to determine whether the particles were black or white. It was then decided to try to eject the black material rather than the white particles so that the reject fraction was clean white plastic. This put a large demand on the machine as a large proportion of the material was black plastic. Neither of these colour sorts were successful so all the materials were remixed and processed again. The machine was reset to eject PCBs as in the previous trials and the remixed material was run through on these settings (Sort 1). The reject fraction from this sort was very clean but the PCB fraction contained a lot of plastic. Figure 19 shows this fraction. From this separation the PCB eject fraction was then processed again with the same system settings to remove the PCB s from the plastic (Sort 2). The plastic content of the second processed PCB fraction was still significant at 77%. It also contained some metal pieces. The metal was picked out by hand; some of it was clearly magnetic and could have been removed by a magnet. The reject plastic from Sort 1, Figure 18, was then mixed with the reject plastic from Sort 2 and processed to separate the plastic by colour. The system was set up to eject the black plastic using the red reflection and red anti-scattering signal. Figure 20 shows the black plastic eject fraction. 34

35 A few PCBs were found in the eject fraction. It is thought that during the first sort of the material some PCBs fell back into the reject fraction due to the design of the splitter plate, as the PCB s could be identified by the machine. The white material from the sort was then processed again to show the degrees of whiteness which could be identified by the equipment. Figure 22 shows a photograph of the white material. The fractions on the left and in the middle were ejected, whilst the sample on the right was the reject. The white material was then mixed back together. 6.3 Photograph of result samples Figure 18: Photograph of Trial 5 plastic reject from sort 1 35

36 Figure 19: Photograph of Trial 5 sort 1 PCB eject fraction Figure 20: Photograph of Trial 5 sort 3 eject - black plastic 36

37 Figure 21: Photograph of Trial 5 sort 3 reject - white plastic Figure 22: Photograph of Trial 5 sort 3 reject white plastic reprocessed into different white fractions (Note the fractions were remixed after the photograph). 37

38 6.4 Analysis of results samples Table 10 shows the results of the hand sorting for trial 5. Trial Fraction Material Weight of fraction Weight of samples Sample composition PCBs Plastic Metal Wood PVC Wires kg g g % g % g % g % g % g % g % g % g % g % T5 sort 2 Eejct PCBs % % % % % % % 0.0 0% % 0.0 0% T5 sort 3 Eject Black Plastic % % % 0.0 0% % 0.0 0% 108 3% % % % T5 Sort 3 Reject White Plastic % % 112 3% % % % % 0.0 0% 0.0 0% 0.0 0% Copper Wires Rubber Stone/ Glass Lights/ Foils/ Foams Others Table 10: Hand sorting results for Trial 5 fractions

39 Feed kg % Eject PCB's kg % Total % Total % PCB's % Plastic % Metal % Wood % Visys Laser Sorter PVC Wires % Trial 5 Sort 1 Copper Wires % Reject Plastic Rubber % kg % Stone/Glass % Total % Lights/Foils % PCB's % Others % Plastic % Metal % Wood % PVC Wires % Copper Wires % Rubber % Stone/Glass % Lights/Foils % Others % Figure 23: Schematic for Trial 5 Sort 1 results - first pass of plastic/pcb mixture, ejecting PCBs The product separation efficiency, Q = the probability that PCBs are correctly sorted into the eject fraction. The reject separation efficiency, R = the probability that non PCB material is correctly sorted into the reject fraction.

40 Q R Trial 5 Sort 1 69% 89% Table 11: Q and R separation efficiencies for Trial 5 Sort 1 Eject PCB's kg % Total % PCB's % Plastic % Metal % Wood % PVC Wires % Copper Wires % Rubber % Stone/Glass % Lights/Foils % T5 S1 Eject PCB's Visys Laser Sorter Others % kg % Trial 5 Sort 2 Total % Reject Plastic kg % Total % Figure 24: Schematic of Trial 5 Sort 2 results - sort 1 eject second pass, ejecting PCBs The product separation efficiency, Q = the probability that PCBs are correctly sorted into the eject fraction. 40

41 The reject separation efficiency, R = the probability that non PCB material is correctly sorted into the reject fraction. Q R Trial 5 Sort 2 98% - Table 12: Q and R separation efficiencies for Trial 5 Sort 1 41

42 Sort 2 Reject Plastic kg % Total % Eject Black Plastic kg % Sort 1 Total % Reject Plastic Feed Plastic PCB's % kg % kg % Plastic % Total % Total % Metal % PCB's % PCB's % Wood % Plastic % Plastic % PVC Wires % Metal % Metal % Copper Wires % Wood % Wood % Rubber % PVC Wires % PVC Wires % Stone/Glass % Copper Wires % Copper Wires % Visys Laser Sorter Lights/Foils % Rubber % Rubber % Trial 5 Sort 3 Others % Stone/Glass % Stone/Glass % Lights/Foils % Lights/Foils % Others % Others % Reject White Plastic kg % Total % PCB's % Plastic % Metal % Wood % PVC Wires % Copper Wires % Rubber % Stone/Glass % Lights/Foils % Others % Figure 25: Schematic of Trial 5 Sort 3 results - sort 1 and 2 plastic fractions, ejecting black plastic 42

43 Overall Q Overall R Trial 5 68% 95.0% Table 13: Overall Q and R separation efficiencies for Trial Discussion of results This trial sort was more complicated than the previous trials and tested two features of the machine: its ability to remove PCBs and its ability to separate plastic by colour. The PCB fraction was processed twice; the first time to remove the PCB material from the bulk plastic and the second sort was to refine PCB fraction. The results show that this did not work too well. Of the PCB fraction, the sort 2 eject fraction contains only 17% PCBs and 77% plastic which, even after a second pass through the machine, is too low to be commercially interesting. The separation efficiency for the removal of the PCBs, after two passes through the machine, was only 68%. This means the machine removed two thirds of the PCBs from the feed. The amount of mis-sorted material in the eject fraction is expected, as there is such a low quantity of PCBs in the feed. The low separation efficiency could be due to the PCBs being fairly small in comparison to some of the plastic pieces. The air jets may fire to eject the PCBs but because they are small the air jets miss or the larger pieces of plastic are caught by the jets instead. This meant that 1% PCBs remained in the reject. Because the sort 2 reject was remixed and processed and R value cannot be determined as the sort 2 composition is not known. Sort 2 removed 44% of the feed to the reject. The photographs in Figure 20 and Figure 21 clearly show the split into black and white plastic. This was a quick demonstration of the machine capabilities, so with more time the separation could be fine tuned and improved. The analysis of the results shows that 44% of the feed plastic was ejected whilst 56% of the plastic went to the reject. The photographs show that the plastic present covers the whole scale from white to black. Deciding at what point the material stops being classified as white and should no longer be ejected is difficult. Because of this the eject and reject fractions were not individually sorted into distinct black and white fractions. The level of PCBs in the feed for this trial was low. This is probably because it had been processed for circuit board removal prior to arriving at Axion s plant at Salford. It may be that because the level of PCBs was so low in the feed material, the machine had difficulty finding the few circuit boards amongst the bulk plastic and hence the results are poor. 6.6 Conclusions from trial The trial was not as successful as some of the other trials and this may be due to the low level of PCBs present in the feed material. The eject fraction after two passes through the machine only contained 17% PCBs. This is not acceptable as a saleable PCB fraction. The demonstration of the machine s ability to identify and separate plastic by colour, just black and white in this case, was promising and further tests with multicoloured plastic are recommended.

44 7.0 Economic assessment of the technique The Visys Spyder laser separator showed good technical performance and could be used to remove PCBs and potentially also for colour sorting. A simple economic assessment of the technology has been completed through a payback calculation. Table 14 shows the results of the payback calculation and is followed by an explanation of the assumptions. Pay back calculation Trial Equipment Visys Laser Sorter Capacity te/hr 4 Cost of unit Basis of operation hr/yr 3000 Overall Equipment Effectiveness OEE % 70% Plant Input te/yr 8400 Operating Costs Power Quantity kw 30 Cost (assuming 10p/kW hr) /hr 3.00 Power costs /te of feed 0.75 Power costs /yr 6300 Labour costs /yr Total Operating Costs Revenue 5% feed is PCB's PCB Product extracted te/yr /te from sale of PCBs. 60/te land fill cost saving Value of product /te 100 /yr Margin /yr Payback time (months) 329 Table 14: Payback calculation for a Visys laser sorter separating PCBs 44

45 It is assumed that the machine could process 4 tonnes per hour. A unit with two lasers costs 250,000 Euros. The installed cost assumed to be double this and the full cost is converted to GBP. It has been assumed that the unit would operate for 12 hours a day, 5 days a week, 50 weeks a year giving 3,000 hours of operation per year. Power is required to run feeders, the detection system and to provide the compressed air required to eject the material. If the system requires 30 kw hr/hr to operate this costs 3.00 per hour. The power cost per tonne of feed is Assuming an overall equipment effectiveness of 70%, this gives a plant input of 8,400 tonnes per year with a power cost of 3,150 per year. Assuming one operator is required at a full job cost of 20,000 per year this gives total operating costs of 26,300 per year. Assuming that 5% of the feed is PCBs this means that 410 tonnes per year of PCBs are recovered. The value of the PCB is estimated at 40 per tonne. In addition a landfill cost of 60 per tonne is avoided for this material. This means the system will separate PCB product valued at 42,000 per year. For the installed equipment cost of 430,000, it would take 330 months (28 years) to pay back. The payback calculation is sensitive to the amount of PCBs in the feed and the value of the PCB fraction. PCBs from computers tend to have a higher value than PCBs from televisions and other small WEEE, as they contain more precious metals. Computer PCBs can sell for up to 500 per tonne. If this value can be achieved and feed material containing at least 400te/yr of these high grade PCBs can be sourced then the payback time reduces to 25 months, which should be acceptable. The economic evaluation demonstrates that use of the Visys Spyder sorter to remove low grade PCBs from WEEE plastic fractions is unlikely to be commercially attractive at current PCB values. However for processors handling higher grade PCBs from computers in significant volumes it may be an attractive alternative to hand sorting. 45