Value Added Color Sortng of Recycled Plastc Flake from End-of- Lfe Electrcal and Electronc Equpment B. L. Rse*, L. E. Allen*, M. B. Bddle*, and Mchael M. Fsher * MBA Polymers, 500 W. Oho Ave., Rchmond, CA 94804, brse@earthlnk.net, 510-231-9031, FAX: 510-231-0302 Amercan Plastcs Councl, 1300 Wlson Blvd., Ste. 800, Arlngton, VA 22209, mke_fsher@plastcs.org, 703-253-0614, FAX: 703-253-0701 Introducton The propertes of products made from recycled plastcs are n part determned by the level and types of mpurtes found n the prmary plastc. The removal of other plastcs s essental f the propertes are to approach those of vrgn resn. Purfcaton of plastc resns s acheved by explotng dfferences n materal propertes of the dfferent plastc types. Dfferences n densty and surface propertes have been shown to allow for the separaton of a number of plastc materals. 1,2 In many cases, streams of recycled plastc are composed of flakes of a wde varety of colors. The flakes can also range from clear to opaque. Separaton of plastcs nto groups of smlar colors can greatly ncrease ther value because they can be colored to meet reasonable specfcatons much more easly. Sometmes, the dfferent colors mght correspond to dfferent types of plastcs, so ther separaton s desrable. For these reasons, the sortng of plastcs based on color may prove to be a valuable separaton technque for recycled plastc flakes. Fortunately, equpment s commercally avalable whch can sort granular materals based on color. Such equpment has been utlzed n the food 3,4,5 and plastcs 6,7 ndustres. In ths study undertaken jontly by the Amercan Plastcs Councl and MBA Polymers, color sortng was appled to the separaton of whte, gray and black plastcs from end-of-lfe electronc equpment. The composton dependence of sortng was nvestgated expermentally and compared wth a theoretcal model. Results ndcate that ncreased levels of mpurtes decrease the throughput rate and result n ncreased mpurty levels n both product and reject streams. The removal of black from a mxture of black and gray flakes was also demonstrated as a way to control product color. I. Background The color sorter used n ths study s a Scanmaster II manufactured by Satake. 8 Ths unt separates materals based on the gray scale ntensty of fallng partcles relatve to a background ntensty. The colors of the partcles are made dark or lght relatve to other colors and a background by selecton of approprate lght sources and flters. For example, red appears the same as whte under a red flter but the same as dark gray under a green flter. One can thus separate flakes based on how lght or dark ther colors appear wth a certan combnaton of lghts and flters. The user can decde whether to pneumatcally eject partcles that are darker ( dark trp ) and/or lghter ( lght trp ) than the background. In addton, the user can decde how much lghter or darker than the background the partcle must be n order to be ejected. Fg. 1 s a schematc representaton of fve partcles aganst such a background. Partcles A and B can be ejected wth a lght trp, although ejecton of B would requre a hgh senstvty. Partcle C s the same shade of gray as the background, so t cannot be ejected. Partcles D and E can be ejected wth a dark trp, although a greater senstvty would be requred to eject D. The lghtness of the background s controlled by physcally changng the background color (whte, black and varous other shades are avalable) or by rotatng the angle of the background relatve to the lght source and detectors.
Fg. 1: Ejecton of partcles lghter or darker than the background ejected wth Ņlght trpó not ejected ejected wth Ņdark trpó A B C D E The color sorter s composed of several charge coupled devce (CCD) cameras whch observe partcles as they slde down a number of grooves n the color sorter. Each CCD camera has fve channels correspondng to nvddual grooves. Sgnals from the CCD cameras are processed by a computer whch decdes whether the partcles are too lght or dark and whether or not they should be ejected. The sortng can work n ether reflectve mode or translucent mode. The choce of mode depends on whether the flakes are clear or opaque. The reflectve sort mode employs both front and rear lamps, detectors and backgrounds, as shown n Fg. 2. Ths mode allows the removal of flakes that appear lghter and/or darker than the background. Materals are often opaque, but clear samples can also be ejected f they appear lghter or darker than the background. The unt can be run n translucent sort mode merely by changng a settng on the computer. The settng shuts off one par of lamps, allowng optcally clear partcles to appear much brghter than the background. Ths mode s useful for removng clear materals from opaque materals. When a lght (or dark) partcle s detected n a partcular channel, a sgnal s sent to a pneumatc ejector whch ejects the partcle wth a blast of ar. Ths ar blast occurs a few centmeters below the CCD camera detector. Because of dfferences n sze, shape and frctonal propertes of partcles, fallng veloctes may vary. It s therefore mportant to tune the delay tme (the tme between detecton and the ar blast) and dwell tme (the duraton of the ar blast) n order to remove as many of the undesrable partcles as possble. There are tradeoffs, of course, snce a longer dwell tme causes the ejecton of a larger number of good partcles n addton to the reject partcles. Fg. 3 shows a schematc of the probablty (P(t)) that a reject partcle detected at tme=0 passes the pneumatc ejector at tme t. The shape and sharpness of ths curve depend on the color sorter throughput rate as well as the partcle sze, shape and frctonal propertes. When bad flakes are ejected, good flakes are also carred wth the ejector ar. In order to reduce the total amount of good flakes gong to the reject stream, many color sorters contan a recycle loop. Only partcles ejected twce are sent to the reject stream R. Fg. 4 s a schematc flow dagram of a color sorter wth such a recycle loop. Fg. 2: Schematc dagram of the Scanmaster II n the reflectve sort mode. llumnated fallng partcles lamps CCD background ejector computer product CCD background ejected partcle reject lamps Fg. 3: Probablty that a reject partcle wll pass ejectors at a tme t. Typcal tmng parameters ( delay and dwell are shown. delay dwell tme t after detecton In the followng sectons, we descrbe the performance of the color sorter for two partcular separatons. Snce many of the plastcs n electroncs equpment are whte, gray or black, we concentrate
on separatons of whte from black and the removal of black from gray. The expermental detals and results are separated nto sectons descrbng the partcular type of separaton. Fg 4: Flow dagram of color sorter wth a resort loop F prmary sort P R resort II. Composton dependence of the removal of whte from black A. Introducton In many of the exstng applcatons of color sortng, the amount of mpurtes s very small. In addton, there may be very lttle change n mpurty composton over tme. In the separaton of engneerng thermoplastcs, however, the composton of a partcular color can vary from trace quanttes to 100%. One of the bg uncertantes n color sorter performance s how well t wll handle a wde varety of feed materals wth a varety of mpurty compostons. Varatons n composton wll change materal throughputs by changng the proporton of materal ejected. Fg. 5: Flow dagram of a smplfed color sorter wth no secondary sort F, f prmary sort P, p R, r In order to expermentally nvestgate the effects of composton, varous amounts of whte plastc were mxed nto black plastc and removed usng a sngle pass through the color sorter (no resort loop). Ths smple separaton wth feed rate F, product flow rate P and reject flow rate R s shown n Fg. 5. These expermental results are then compared wth a model derved n the Appendx. The model predcts how the mpurty compostons n the reject stream (r ) and product stream (p ) depend on the feed mpurty composton (f I ), the fracton of detected reject partcles that are actually removed () and the number of partcles that are ejected durng each ejecton (N). The model also relates the reject flow rate to the feed flow rate (R/F). The man results from the model are equatons (1), (2) and (3). r = 1 f N + N p = ( 1 )f N +1 R F = f N +1 + f N (1) (2) (3) B. Expermental We prepared 30 kg plastc flake samples wth varous amounts of whte flakes as an mpurty n black flakes. Only one half of the prmary sorter was used n order to accommodate ths relatvely small sample sze. The samples were color sorted wth a lght trp aganst a black background. The same senstvty, feed rate and ejecton tmng were used for all the samples. After purgng the prevous sample materal, samples were collected from both the product (P) and reject (R) streams for between 30 and 60 seconds. The tmed samples were weghed to determne P and R. The compostons f, p and r were determned by hand sortng 30 g portons by color nto whte and black. C. Results Fg. 6 s a plot of the mass fracton of whte flakes n the reject as a functon of feed composton. Accordng to equaton (1), ths plot should be lnear wth a slope of (1-/N) and an ntercept of /N. The best ft of ths equaton s when /N=0.15. Fg. 7 s a plot of the mass fracton of whte flakes n the product as a functon of feed composton.
Accordng to equaton (2), ths plot should be lnear and through the orgn. The best ft of ths equaton when /N=0.15 s for =0.996. Ths suggests that 99.6% of the detected reject partcles are ejected and that about 6 or 7 flakes are ejected per ejecton (N=6.7). Fg. 6: Reject composton as a functon of feed composton whte r 0.5 0.4 0.3 0.2 0.1 r whte /N=0.15 0 0 0.05 0.1 0.15 0.2 0.25 f whte Fg. 8 s a plot of the rato R/F as a functon of feed composton. The expermental data compare favorably wth the model when /N=0.15 and =0.996 are substtuted nto equaton (3). Fg 7: Product composton as a functon of feed composton whte p 0.008 0.006 0.004 0.002 p whte /N=0.15, =0.996 0 0 0.05 0.1 0.15 0.2 0.25 f whte These results demonstrate that the color sorter can effectvely reduce the amount of mpurty n the product stream up to relatvely hgh levels of mpurty n the feed. The rato R/F, however, Based on the measured throughput of 55 lb/hr per channel and the average mass per partcle, roughly 0.1 to 0.2 partcles pass the ejector per mlllsecond of dwell tme. Wth a dwell of 10 mllseconds, we expect N to be between 1 and 2. The hgher number (N=6.7) may be due to the ejecton of glossy black partcles that appear lghter than the background. becomes large enough to drastcally reduce the producton rate P once the mpurty composton n the feed s above 5 or 10%. Fg. 8: Reject rato as a functon of feed composton R/F 0.8 0.6 0.4 0.2 III. R/F /N=0.15, =0.996 0 0 0.05 0.1 0.15 0.2 0.25 f whte Preparaton of dark and lght plastc products from a mxed feed A. Introducton Streams of plastc from electronc equpment often occur as mxtures of gray and black plastc flakes. In some cases, the value of the product can be ncreased f the materal s sorted nto a product stream wth very lttle black materal and another product stream wth very lttle gray materal. In the gray stream, any black materal wll make the product pellets a darker gray. Ths color may be fne for many applcatons, but t prevents the materal from beng colored wth any color other than black (to make dark gray). In the prmarly black stream, the presence of TO 2 (as occurs n gray plastc) makes t dffcult (or mpossble) to make the plastc black. Snce many buyers prefer a black materal, t s mportant to remove as much gray as possble from the black product. B. Expermental A typcal lot of flake materal was passed through the color sorter twce. Dark colors (black and dark gray) were ejected nto the reject stream. The feed materal, the product from the frst pass, the product from the second pass and the reject from the frst pass were collected for color analyss, pelletzaton, njecton moldng and ash tests (to determne TO 2 loadng). Color analyss conssted of categorzng all materal n a 15-25 gram sample of flakes as ether lght or dark. The lght materal ncluded gray,
clear, whte and colors. The dark materal ncluded black flakes. Ash tests were performed on portons of test specmens that were molded from pelletzed flakes. C. Results Table 1 gves the weght fractons of lght and dark and the ash content n the four samples. Table 1: Analyss of color sorted flake materal sample % lght % dark % TO 2 Feed 78 22 3.01 Product 1 95 5 3.50 Product 2 98 2 3.70 Reject 1 44 56 1.98 These results ndcate a sharp mprovement n lght composton wth the frst pass, and a smaller mprovement wth the second pass through the color sorter. The amount of TO 2 n the frst pass reject s roughly 2/3 of the amount n the feed stream. Ths s close to the rato of % lght n the two streams. The reject wll requre sgnfcant cleanng (wth low throughput rates) to remove enough of the TO 2 to allow t to become black. As shown n Fg. 9, the products are lghter and the reject darker than the feed materal. There s lttle vsual dfference between the frst and second pass products. There s a slght dfference n the % lght and % solds, however. These results suggest that color sortng s a useful tool for the producton of lght and dark streams from a sngle materal stream. A sngle pass can elmnate about 80% of the dark flakes from a prmarly lght materal. Further purfcaton also helps wth the purty, but to a lesser extent. The reject stream s enrched n the dark materal. Purfcaton of ths stream wll be slow, especally f nearly all of the lght materal must be removed. Fg. 9: Molded specmens of feed, products and rejects from the color sorter. IV. Conclusons Separatons based on color can be mportant for a number of reasons, ncludng the mprovement of product qualty or a more consstent product color. In ths study, color sortng was appled to the separaton of whte, gray and black plastcs. Results for the removal of whte from black plastc flakes are compared wth a model predctng the feed composton dependence of product purtes and throughputs (see Appendx). Both the expermental results and the model ndcate that ncreased levels of feed mpurtes decrease the product throughput rate and result n ncreased mpurty levels n both product and reject streams. Another set of experments demonstrated the creaton of lght and dark products by removng black flakes from gray flakes. Extruson of the feed materal and products showed that one product s lghter, and the other darker, than the feed materal. V. Appendx A model for product and reject compostons and throughputs as functons of feed composton When a partcle s sgnfcantly darker (or lghter) than the background, a computer sgnals a pneumatc ejector to fre a blast of ar begnnng at a tme delay past detecton for a duraton of tme dwell. We assume that all partcles are the same weght so that number fractons can be equated to mass fractons. Though ths s not the actual case, the results should be qualtatvely correct. The mpurty partcle, once detected, s ejected wth a probablty. Ths probablty depends on the flow of partcles and s a functon of both delay and dwell. If the delay s properly selected, s expected to ncrease wth dwell and approach a value of unty. Other partcles are also ejected along wth mpurty partcles causng the ejecton. The number of partcles per ejecton N s gven by N = V dwell (A.1) where V s the number of partcles passng the ejector per mllsecond (f dwell s n unts of mllseconds). The total number of mpurty partcles N ejected durng an ejecton ncludes both the prmary ejected partcle and a random number of other mpurty partcles (proportonal to the mpurty composton n the feed).
N = + f N ( ) (A.2) The reject mpurty composton r s therefore gven by ( ) r = + f N N = + f ( V dwell ) V dwell (A.3) For a gven set of ejecton parameters, r should be lnear wth the feed composton. Plottng r as a functon of f should gve a slope of 1-/N and an ntercept of /N. The parameter /N tells us what fracton of the ejected partcles are actually the recognzed ejected partcle. Ths parameter would equal one f no extra partcles were ejected. We would also lke to relate the mpurty concentraton n the feed to that expected n the product. For relatvely small levels of mpurty n the feed, we assume that the detectors can detect all mpurty partcles and that the ejectors can fre for each partcle detected. In ths case the only mpurtes reachng the product are 1- for each ejecton. The mpurty composton n the product s equal to ths amount dvded by the total amount of partcles that are not ejected (N p ). p = 1 N P (A.4) Ths number N p s the average number of product partcles between each ejecton event. N p can be related to N and the flow rates F and R as follows. N p = F R 1 N (A.5) A speces balance for yelds an equaton for F/R as a functon of compostons n the varous streams. The product mpurty concentraton can be plotted as a functon of the feed mpurty composton. The slope (combned wth /N determned from the plot of r versus f ) should gve the fracton of partcles that are detected and ejected. Ths number should be close to one. One can also express the rato of the reject flow rate to the feed rate (R/F) as a functon of feed composton. Combnng the prevous equatons gves the result. R F = f N +1 + f N (A.8) The rato R/F s thus a functon of the feed mpurty composton and parameters /N and N (whch are assocated wth the color sorter settngs). References 1. Morley, N., Polymer Recyclng, 3 (3), 217-226, 1998. 2. Bddle, M. B. and Fsher,, M. M. Proceeedngs of the SPI 22 nd Annual Conference, Structural Plastcs Dvson, Washngton, DC, 1994. 3. Larson, B. E. and Robe, K., Food Processng, USA, 45 (5) 174, 176, 1984. 4. Satake, S., Ito, T. and Ikeda, N., US Patent 5638961, 1997. 5. Zhang, M., Ludas, L.I., Morgan, M. T., Krutz, G. W. and Precett, C. J., Proceedngs of 1998 Precson Agrculture and Bologcal Qualty (SPIE), 208-219, 1999. 6. Schut, J. H., Plastcs Technology, 38 (10), 15, 1992. 7. Ban, D. R., Botje, G. and Scholt, J. C., Proceeedngs of the Recycle 94 Conference, 1994. 8. Satake, USA, Houston, TX, www.satake-usa.com. F R = ( r f ) p f 1 ( p f ) ( ) (A.6) The prevous three equatons can be combned to yeld an equaton for p that s lnear n f. p = ( 1 )f N +1 (A.7)
1 Morley, N., Polymer Recyclng, 3 (3), 217-226, 1998. 2 Bddle, M. B. and Fsher,, M. M. Proceeedngs of the SPI 22 nd Annual Conference, Structural Plastcs Dvson, Washngton, DC, 1994. 3 Larson, B. E. and Robe, K., Food Processng, USA, 45 (5) 174, 176, 1984. 4 Satake, S., Ito, T. and Ikeda, N., US Patent 5638961, 1997. 5 Zhang, M., Ludas, L.I., Morgan, M. T., Krutz, G. W. and Precett, C. J., Proceedngs of 1998 Precson Agrculture and Bologcal Qualty (SPIE), 208-219, 1999. 6 Schut, J. H., Plastcs Technology, 38 (10), 15, 1992. 7 Ban, D. R., Botje, G. and Scholt, J. C., Proceeedngs of the Recycle 94 Conference, 1994. 8 Satake, USA, Houston, TX, www.satake-usa.com.