DRAFT SYNTHESIS REPORT FOR PUBLICATION. PROJECT COORDINATOR : F raunhofer-institut fir Lasertechnik, Germany (ILT)

DRAFT SYNTHESIS REPORT FOR PUBLICATION CONTRACT NO : 13RE2-CT93 0589 PROJECT N“ : BE 7717 TITLE : High Speed Laser Identification of Plastics and Pol...
1 downloads 1 Views 540KB Size
DRAFT SYNTHESIS REPORT FOR PUBLICATION CONTRACT NO : 13RE2-CT93 0589 PROJECT N“ : BE 7717

TITLE : High Speed Laser Identification of Plastics and Poljrmers f r o m Domestic Waste for Recycling Purposes

PROJECT COORDINATOR : F’raunhofer-Institut fir Lasertechnik, Germany (ILT)

!,

PARTNERS :1. Foundation for Research and Technology-Hellas, Greece (FORTH)

I

I

2. Gaiker, Spain (GAIKER) ,,I

I

REFERENCE PERIOD FROM 01.12.1993 to 31.05.1996

*n*

STARTING DATE: 01.12.1993

* * *

I

Sa 95/11 ~1.212.3109

***

* * *

DURATION: 30 MONTHS

PROJECT FUNDED BY THE EUROPEAN COMMUNITY UNDER THE BRITE/EURAM PROGRAMME

)

DATE: 31.05.1996

J

!

BE 7717- Draft synthesis report

I

.

.

High speed Iaser identification of plastics and polymers from domestic waste for recycling purposes R. NoII, R. Sattmann Fraunhofer-Institut for Laser Technology, Steinbachstra13e 15, D-52074 Aachen, Germany, phone (++49) 241 / 8906-0, fax -12 i S. Couris, A. Hatziapostolou Foundation for Research and Technology - Hells; (F. O. R.T.H.), P.O. box ,1527, Heraklion 71110, Greece, phone (++30) 81 /39 1470, fax391318 E. Larrauri Centro Tecnologico Gaiker, Parque Tecnologico Edificio 202, E-48170 Zamudio, Spain, phone (9)4 -4522323, fax 4522236

Abstract Laser-induced breakdown spectroscopy (LIBS) is investigated for the identification of the polymers PE, PP, PET and PVC. About 10 spectral features are measured, e.g. sp:ctral lines of carbon, hydrogen and chlorine and emission bands of C2. The intensities are evaluated with multivariate statistical analysis and neural networks. Identification accuracies of = 90-95 % for PE and PP and of> 99% for PVC and PET are achieved. The measuring time for an identification is less than 100 p. An autofocus system for the focussing of the laser beam onto bottles with varying geometry and for plasma imaging has been assembled and tested with coloured samples. A handling device for the singularization, transport and sorting of waste bottles has been assembled, allowing sorting rates of up to 3 bottles/s. An assessment for the feasibility of industrial-scale sorting machines is given.

I

1

I

Version: .W31.05.1996

d’ippex.Sattmarm>SN6219 .dod

.—

d

1 Introduction

Polymers represent a signific~nt ~mourtt of the municipal solid waste. mainly due to bottles. From the 138.556.000 tonnes of [otal domestic woste generated in 1993 in Western Europe [ 1]. 10.928.000 tonnes are plastics (zbout 8 ?O in weight and 40 % in volume), cf. Fig. 1. The increasingly restrictive lilws about residues calI for new methods for its reduction, recovery or elimination avoiding damping.

cl%

. . .,,.,-----. . . . . ..&._ . ...=;-= -_-*

~ ORGANIC PRODUCTS 43,5 %

PAP BOARD 29,3 %

PLASTIC

. . %,5 %

S

-.. tiw’t 15.6

PP 6,4 %

OTIERS 4,7%

p= 6,2 %

Ps+ps r!?%

% ~

LDPEfLIDPE 42,5 %

Figure ]: Piasfics and domestic wasre

A high-level recycling of the polymers requires sorting the waste bottles into pure polymer fractions. Several approaches are investigated for the sorting of polymers- Mass , methods like swim-sink-techniques cannot separate the polyoleflnes PE and PP. A major problem is the separation of PET and PVC, since these po~ymers are highly incompatible. Probably [he mostly investigated single-piece idec.tification technique is nearinfrared {NIR) reflection spectrometry [2,3], which is also implemented in pilot facilities. Nevertheless. NIR suffers from disturbances as soiled or labeled surfaces. This report describes the method of laser-induced breakdown spectroscopy (LH3S) for the identification of the polymers FE, PP, PET and PVC, evaluation algorithms, an autofocus system to adapt to bottles of varying geometry ar:d a handling device for the singuhrization, transport and sorting of waste bottles. The developed system for the separation of post consumed plastics will be a tool to reduce the increasing amount of packaging pIastics that go to landfills, reducing the environmental impact of the pkstic waste.

, ‘!

1

I

I

,

I

I



I

SA96219.dod

Version: M31.05.1996

1

,

BE 7717- Draft synthesis report

3

,

.

2 Approach

2.1 Laser-induced breakdown spectroscopy Laser-induced breakdown spectroscopy makes use of the element and/or molecule characteristic emission of excited species like atoms or molecules [5, 6]. hltel~se laser pulses are focussed onto the specimen. The laser radiation vaporizes a small amount of material and induces a plasma with excited atoms, ions and ~molecules. After calibration with reference samples of known composition, the irradiated sample can be anaiy sed by temporally and spectrally resolved measurements of the plasma emission.

[i,~~~~~ ——— ____ .._.

Nd:YAG-Laser

Iwl PMT

Pc

FO

u Monochromator

+

Multichannel gated + integrators

b PIN u ,F

,.,

‘“’$ Plasma Specimen

Figure 2: Scheme oflhe iabot-afory sefup for polymer iden@cafion wit% LJBS [1.PD4: intensified photodiode array, PMT: photornuhiplier tube, FO: jlber oplics, PIN: pin photodiode, IF: interjlererzce filfer, L- knse,). The laboratory setup is shown schematically in Fig. 2. Laser pulses of a pulsed pumped, Q - s w i t c h e d Nd:YAG-laser are guided by mirrors to a lens L (planoconvex, ‘ ~= 200 mm), which focusses the pulses onto the specimen. The focus position can be changed by shifting the lens. The front end of a fiber optics (bundle of quartz/quartz fibers) is mounted at a fixed distance to the specimen. The fiber optics guide the piasma radiation to the spectrometers. For most of the experiments, a Czerny-Turner spectrometer is used (Jobin Yvon, type HR 320, g-rating 600 I/mm). The rear end of the fiI ber is mounted in front of the entrance slit of the spectrometer. An intensified photodiode array (IPDA} is mounted in the exit focal plane of the spectrometer for the spectrally and temporally resolved detection of the emissiun spectra. The data of the IPDA are sent to a personal computer (PC) via a controller.

I

I

cT1ppex.Sattmann> SA96219.dod

Version: A/31.05.1996

I

I ,; ( I

BE 7717- Drafl synthesis report.

4

.

For the simultaneous measurement of line intensities beyond the spectral region detected by the lPDA, bifurcated fiber optics are used. SeveraI arms of the bifurcated fiber optics transmitt the plasma radiation to separate spectrometers. Besides the Czerny Tumer spectrometer, Czemy-Tumer monochromators (~bin Yvon, type HR 250, grating 2400 I/mm) and interference filters are used to register spectrally resolved data. The intensities are measured with photomultipIier tubes (PMTs) and PIN-photodiodes (PDs). Time resolution with the lPDA is achieved by gating the multichannel plate with high-voltage pulses. The signals of the PMTs and PDs are integrated for time windows ~ing by means of a multichannel time-gating integrator with a delay tdel to the laser puises. After integration, the signals are digitized and sampled by a PC. Table 1: Samples used for the experiments. The samples marked with * are mizde from virgin polymer, the other samples are recycled imzterials. ~ Polymer I

I Cdour

Polymer

I

blue blue brown not dyed yellow yellow 1

I LDPE bl* LDPE no LDPE rd* LDPE ye*

I blue not dyed red yellow

HDPE bl HDPE bl 2* HDPE br HDPE no HDPE ye 13DPE ye 2*

13DPE

LDPE

J

.--—

Sample

Sample II name

1[

PET

II PET no

I

PP

I PP bl Pl? gr PF no PF rd I

) blue green not dyed red I

Pvc

{ PVC bl

/ blue not dyed red red white

PVC no PVC rd PVC rd 2 Pvc Wh

‘cklh.wr 1 nut dyed

For most of the experiments, a Nd:YAG laser with multiple-pulse option is used [7, 8]. Single and double pulse bursts are generated by a single flash-lamp pulse. The duration of the pulses is = 15-25 ns, the interpulse separation of the pulses in a double puIse burst (or “double pulse”) amounts to 6 ps. Using doubIe pulses, the energy is equally distributed among the two pulses. The laser is operated at the fundamental wavelength of 1064 nm. The focus position is located 5 mm inside the samples to “avoid air breakdown. Most experiments are performed on flat, typically 5 mm thick samples of the materials HDPE, LDPE, PP, PET and PVC, produced from virgin and recycled material, cf. Tab. 1. Waste bottles of the materials HDPE, PP, PET and PVC have also been measured.

I

L

. .— F al z – 1

4000 3000 2000

I

I

l“’’I’’

6

’’1’

’’’1

’’”1 E c

n .

-J

:/

:

(2- hands

x:,~\

C

2 bands

h

. . -. . .

..,

““1’ ’I1IIIIIII[IIJ

E c

m 7

Lines of additives i

I

,,

:

., ,. ., . .

.,

‘.”

–z

––..--

: 0 ; A Ill

v

ill A, J,

.. . .

I

W&./

‘...

..-,

‘.

. . . . . . . . . . . . . . . ~. . . . . 000 ~ . . . .

0

~ lx100mJ }111111

450

.

.

‘ lx200mJ

llll![l,

460

.

470

,,,l,,

480

~?xlrm n-1.l

,,

— . ..- - - - - I,, ,IJ,,

490

500

-1

1!

510

520

530

A [rim] Figure 3: Emission spectra with C2 bands and additive lines tfter different iaser pulses onto sample HDPE bl {tdei = 1 ps, tint = 10 p,s, no pre-bumt, 1 burst measw-ed, 5 experiments averaged). Figure 3 displays the spectra of the sample HDPE bl in a spectral range comprising the lines H 486,13 nm, C 247.86 nm (in 2nd order) and C2-bands. The spectra are sampled for different laser pulse parameters. The spectra of the 1x100 mJ ‘and 2x100 mJ pulses show a similar spectral distribution, with a factor of > 2 higher emission with the 2x100 mJ doubIe pulses compared to the 1x 100 rnJ single pulses. The spectral distribution of the plasma induced by the 2x100 mJ double pulses shows as well significant differences to the spectra of the I x200 mJ singie puke: lower background intensity and higher Cz-band emission. The intensity of the C 247.86 nm line is slightly increased. Measurements of several samples yield approximately a factor of 2 higher intensities with 2x100 mJ double pulses than with 1x200 mJ pulses. Intensities of chlorine lines and additives like titanium are also increased. The behaviour of the hydrogen line is not uniform; in most cases the intensity of H 486.13 nm is reduced significantly, as in Fig. 3. Due to these results, we have mainly used double pulses in the following for polymer identification. In contrast to the other polymers, PVC contains the additional element chiorine. So it is straightforward to detect the Cl emission for the identification of PVC. Figure 4 shows the spectra of a blue HDPE and a blue PVC sample in the region of= 725 nm. A CIIine at 725.66 nm can be detected in the PVC spectra, which can be used for PVC identification.

SA96219,dod

Version: AJ31.05. 1996



BE 7717- Draft synthesis report .

,

7

2000 }“’’’’’’’’’’’’’’’’’’’’’’’’’’’”

‘H13PE bl

o

llllll!l!!tf,,,\,,,/!,l,,/,,l ,,,,

710

715

720

725

730 k [rim]

735

740

.!- L-ffl

745

75(3

Figure 4: Specrra ofsatnptes PVC blue atld HDPEblue. Tke PVC bhiespectrlitnis shified in the vertical direction by 200 counts for clarl~ of presewation {w= 1X2Q0 d> t~e[ = 1 ~} ‘i~~ = 10 VS, 1 pre-burst, i’ burst tneasm-ed). Figure 4 shows, that it is possible to identify PVC samples by measurement of the spectral range of the line Cl 725.66 nrn with [he IPDA.

1,0

J . . . . . . . . . -+-- ---- -

. . . . . . . . . . . . . . . ~---- ---

J..l . . . . . . . . . . . . . -. . . . . . . . . . . . . . . -- -T.

..:

J

G 0,2

0,0

cTi/)pex.Sattmann> SA9G219.dod

Version: A/31,05.1996



BE 77 Ii - Draft synthesis report

s

Nevertheless. the [PDA ~]~{ds more spectra} infornmtiofl thun necessary: measurement of the emission intensity at [he spectral position of the line is sufficient. For this reason. the Cl-line is detected with J photomultiplier tube behicd a monochromator. A reference signal is detected with a PIN photodiode ,behind an interference filter. Figure 5 shows, that the normalised Cl-signal for PVC samples is significantly higher than those for the other materials. The polymers are mainly composed of the elements carbon and hydrogen. Table 2 shows, that the polymers have different stoichiometric ratios of C- and H-atoms.

Formula

~ Polymer It

1

Pvc PE PET

{-CHCl-CH2-)n (-cH2-cE12-)n

] C/H-ratio ~

;~ 1.25

PP

(-o-cH~-cH*-o-co-c6H4 -co-)n (-CH(CH3)-CH2-)n

Ps

(CH(C6HJ-CHz-)n

1

0.5

9

Table 2: Chemical composition and s~oictiiometry ofpolyners. Therefore it should be possible to identify polymers or groups of polymers by measuring the C- and H- concentration. Detection of the spectral region presented in Fig. 3 enables the simultaneous measurement of the intensities of the lines H 486.13 nm and C 247.86 nrn, when the second order is not suppressed by filters as usual.

0,40 m 0,35 fxi % 0,30 --1; 0,25 al $0,20 : -0,15 0,10

FigL;re 6: Evulualion of WY-rcuiofor HDPE, PET and F’P samples, cj Tab. 1 (W= 2x150 m~, tde[ = 0.8 ps, ?i7)f = 10 ps, no pre-burst, ! burst averaged). SA96219. dod

—.

.—



_—

Version: fV31.05.1996

BE 7717- Draft synthesis report

9

.

Results of the quantitative evaluation of the C/H-intensity ratio for PE, PET and PP samples are showri in Fig. 6. The measured intensity ratios me displayed for the samples; error bars indicate the standard deviation of 50 measurements. PVC samples are not shown, since they are akeady identified by the chlorine Iine emission. Obviously the stoichiometric ratio can not clearly be determined by the measured intensity ratio. The reason may be line overlaps - the broad H-line is easily disturbed by additional lines, e.g. from colour materials -, or material dependent plasma conditions, since the excitation energies of the C- and H-line are different. Nevertheless, as expected from the stoichiometric C/H-ratio, the intensity ratio of the PET sample is higher than for the PE and PP samples. Since the error bars over~ap, further features are necessary to discriminate PET from PE and PP. PE and PP can not be distinguished by the C/H line ratio due to their identical stoichiometric ratio. 12000 10000

I

/Ti 323.452 nm

~ 8000 c 3 g 6000 L > .— : 4000 z —

I II

L

1111

~

i

II

~~i~

2000 . ----- ------ . . . . . .

0 tllllll 111[111 300

310

320

lfl, ,,, 111111, ,,il ,,, IH 330

340

350

360

370

h [rim]

Figure 7: Spectral region of Ti 323.45 nm of not-dyed and red PVC ( W = 1 x2(XI mJ, ‘&/ = i P& l~nf = }0 ps, i pre-burst, 1 burst averaged). The possibility to identify additives in polymers may be demonstrated by the detection of titanium. Ti02 is often used in coioured plastics. The titani~m emission Iines can be used for identification of pieces coloured with Ti02. Figure 7 displays the spectra of not-dyed and red PVC (PVC no, PVC rd, cf. Tab. 1). The difference in the spectra is obvious. SeveraI peaks in the spectrum of [he red PVC can be assigned to the emission of Ti. The Ti-line at 323.45 nm is indicated in Fig. 7.

I

I

,i

i I

SA96219.dod

Version: A/31.05.1996

!1

—.—

,,

~’

-

‘-

BE 7717- Draft synthesk re~fi 10

For the automated identification of samples by their spectral features, regions-ofinterest (ROIS) are selected. The regions are spectral lines or emission bands. The intensities of about 10 ROIS are determined and fed into evaluation algorithms. For example, mu]tivariate statistical analysis (MVA) is used for evaluation [10,11].

Tabie 3: Results with multivariate sfa~isticai analysis for the reference samples, C$ Tab. 1 (all samples iisled in Tab. ], W = .2x 1~() inJ, rdei = (1.8 p, tinl = lo pS no pre-burst, 1 burst measured, values in %). Material

Correct dentified

PE

Misiidentified

84.6

PET

15,,4

98.0

PP

2.0

85.0

Pvc

99.6

15,0 0.4

91.9

$. I

Average

I

The aIgorithm has been trained with 50 data sets per reference sampIe (1000 data sets in total) and tested with other 50 data sets per sample. The results are listed in Tab. 3. The values represent the amount of test data sets of the samples of a material, that have been attributed to the correct material or to a wrong material, relative to all test data ‘ sets of a material. The ,relatively high mis-identification values of PE and PP are due to mixing of these polyolefines. The mis-identifications of PET and PVC are mixing of these two materials. Averaged over the materials, 8.1 % of the data sets are misidentified.

I

I ,

Table 4: Results ~vith a neural R t?hVOrk, Samples and parameters as ip Tab. 3 (values in %). Material

Correct identified

Misidentified

PE

Not identified

65.0

PET

28.6

72.0

6.4 0.0

PP

28.0

72.5

4.5

Pvc

23.0

99.2

0.4

Average

77.2

2.8

t

I

0.4 20.0

Better resuIts are achieved by neural networks (NN), which can reject ambiguous data sets. Results achieved with a feedforward network trained by backpropagation and two subnetworks are listed in Tab. 4. The network comprises 10 input celIs, 11 hidden ceils SA96219.dod version: JV31.05,1996



.13E 7717- Drafl synthesis report

11

and 4 output cdl.s. The hidden layer is split into two groups: 8 cells for PE, PET and PP and 3 cells for PVC. The network is [rained and tested with the same data as the MVA. The amount of incorrectly sorted particles is reduced from about 8% with MVA to 2.8%, which improves the purity of the sorted fractions. 20% of all particles have not been identified due to ambiguous spectral data and have been sorted out. To test the identification accuracy on waste bottles, 10 PE, 10 PET and 10 PVC bottles have been measured with the laboratory setup and classified with a neurai network. Each bottle is measured 20 times. The network topology is: 8 input nodes, 1 hidden layer with 11 nodes, and 3 output nodes. The network is trained with the data of 5 bottles of each material and tested for the other 5 bottles. The results are presented in Tab. 5. ln average, 20% are sorted out as above. The amount of COL-RC[ idcntifir.’! ?mttIes relative to all identified bottles is = 95.0%, the amount of PVC in the PE and PET fraction is O% and 7.670, respectively.

Table 5: Results with new-al networtifor waste botdes of HDPE, PET and PVC (W= 2x150 nd, tdel = 1 p.k, ~iti = 10 j.M, no pre-burst, I burs~ measured, va[aes irz %).

I

Material PE

,

PET Pvc

I

Correct 1 Misidentified identified

Average

?

r?ot 1 identified

76

6

18

67

0

33

85 76

I

6

14

I

9 20

I 3.2 Laser and pIasma light guiding equipment - autofocus system The main specifications of the opticaI system are the focussing range and the focussing time. The former depends on the maximum dimensions of the plastic bottles to be processed, which is = 60 mm for non-pressed bottles, indicating that the focussing range should be & 30 mm around a central position. The accuracy for this range was chosen to be f 1 mm, taken into account the depth of focus of the hTd:YAG beam. The latter is the time allowed for the autofocussing action and is related to the speed of the conveyor belt. The initial target was set at 0.3 s, with the aim to reduce it at a later stage to 60-100 ms, or even fu@er, especial] y in the case that the focussing range”is also reduced by means of a bottle pressing unit incorporated in the handling device. After considering and testing various solutions for the specifications for the three modules of the optical system, the components of the final set-up, shown in Fig. 8, are the following:

I

SA96219 .dod

I I

I

I-.

Versiox A/31.05. 1996

I i 1 !

1 / i

I

BE 7717- Draft synthesis rep[jrt

,

12

.\

SA96219. do&

Version: A/3 1.05.1996

. BE ‘7717 - Draft synt$esis report. t, ..

13

c) The design of the plasma collectiori optics is performed using a commercially available ray-tracing software, It is a two mirror system which collects the light emitted by the plasma on the tip of a fiber delivering the light to the spectrometer. M2 is a spherical concave mirror wi[h a focal length of 75 mm (R = 150 mm), diameter 100 mm, thickness 15- 17 mm, surface accuracy ?J4 at 633 nm and a central hole of’ 16.1 mm dia. at an angle of 20° to the optical axk, so that the laser beams can pass. The coating is UV-enhanced A!+ MgF9. M 1 is a flat mirror with a diameter of 50 mm, thickness 1042.5 mm and similar surface accuracy and coating as Ml. The main operation of the autofocussing action is perforned by means of a PCmicrocomputer and it can be shortly described with the following electronic signal sequence: Position Sensitive Detector (PSD) pre-amplifier board s signal conditioning board with power supply ~ interface cards ( e PC-microcomputer) s stepper m.omr controller = stepper motors. The 25 rnW laser diode is modulated by a TTL-pulse generated by the PSD processing board, so that the PSD and diode laser are synchronizeal. The softw’are that runs on the PC-microcomputer determines the stepper motor positions for both the lens F2 and the fiber tip holder as a function of the PSD signal, according to calibration files stored in the computer memory. These calibration files are generated during a procedure involving stationary experinmnts with plastic samples of various colors, by means of special software utilities. The functionality of the autofocus system was tested with various polymer samples Iisteal in Tab. 1. As a general conclusion it was proved that the system was performing adequately with ail the samples or bottles having various colors or irregular surfaces. This was achieved by a final improvement realised by a correction procedure based on the measured intensity of the light scattered by the object surface. This correction proved insufficient only in the case of transparent samples and bottles, where the scattering from the plastic specimen surface is very weak. Attempts to create irregularities on the surfaces of the bottles in the pressing uni~ of the handling device with the aim to increase the scattering of the diode laser beam proved insufficient for the PSD to function adequately in the latter case.

3.3 HandIing device Several systems were considered in order to serialize the plastic bottles. The finally chosen design of the handling and sorting equipment includes three belt conveyors of variable speed and a pressing unit, cf. Fig. 9. The belt conveyor no. 1 is a sloping carrier (length 2055 mm, width 440 mm) with a band made from two textile sheets plastified and covered with PVC with transversal profiles designed to catch bottles from a hopper placed at the bottom of the transporter.

I

cTippex.Sattmann> SA96219,dod

I

1

Version: .W31.05.1996

;1

i

* . ●

BE 7717- Draft synthesis report

14

Although the speed is variable from 15 to 60 rdmin, the transport speed’ should be around 40 rnlmin. Lateral sheets ensure that the bottles will not fall off the be][ Conve. yors.

@I’lu%,nlcI:

1

i

/3

Figure 9: ,Front view of the individwlisation and transport unit. The belt conveyor no. 2 has a transport speed superior to the previous one and adjusted to 50 rnlmin. The pressing unit ensures that the surface position of the bottles will not vary greatly from one to another. This reduces the range that the autofocus system has to cover. Bottles are pressed progressively by pressing chain: supported by a structure of laminated profiles. In order to avoid that a bottle rizes up after pressing, a roll is located 10 cm from the end of the upper pressing chain and 2 cm over the lower pressing chain. This pressing roll throws the bottles to the sorting belt conveyor by means of a set of gravity rolls and a protective case.

I

A photocell is placed besides the analysing point to detect The bottles on their way to the sorting belt conveyor. The sorting belt conveyor is a horizontal carrier whose band is divided into 16 equal areas by transversal PVC profiles with a height of 60 mm. Therefore each bottle is identified with a position on the belt. A magnetic metallic part is placed in the side of each profile. An induction detector mounted under the left hand side of the conveyor de[ects the metal parts of the transversal profiles. When a bottle reaches the right position. the nozzle in that position blows, sorting the bot~le out of the stream.

Four electro~;alves have been placed on the carrier in order to sort out four different types of plastic bo~tles. Each electrovalve has an air nozzle with two air outlets. The direction of the blow’in: devices ha.; been optimized with the intention to cover the complete space in which the plastic specimens can be located. A fifth fraction is formed by the specimens that are no[ blown out by the nozzles. This conveyor has its own programmable switchboxd with the automate tind the pro:ramme that governs the sepw-ation system. The [ransport speed should be adjusted LO 60 m/min. -: TiJ)r}ex, s:ltLmann>s.A96? Ig. tioC’

V(:rsiOn:

fWl, OL.19YG I

< BE 7717- Draft synthesis report

15





4 Conclusions The sorting of waste bottles into pure polymers fractions of FE, PP, PET and PVC with laser-induced breakdown spectroscopy has been investigated. Element and molecule st~ctures were detected in spectra of laser-induced plasmas. Classification algorithms based on multivariate statistical analysis and neural networks were tested.

I

1

I

*

The achievable identification accurcy depends on the kinds of polymers and their additi ves. Typical results on recycled and virgin po] ymer samples achieved under laboratory conditions are 90-95% for PE and PP and > 99% for PET and PVC, where 20?z0 of all measured samples could got be identified and have been sorted out. For bottles pieces, about 95% have been classified correct] y. The measuring times are e 1.00 ps, the evaluation times in the order of milliseconds, enabling sorting rates of 10 per second and more. Autofocussing to the varying geometry of the bottles is a challenging problem. AH autofocus system based on triangulation operates reliable with coloured samples, but not with transparent ones. Future work should aim to avoid the necessity of an autofocus system by proper handling of the bottles and usage of a focussing lens with a long focal length. A modular handling and sorting equipment for singularization, pressing to & 10 mm surface position range and sorting of 3 bottles per second was constructed and put into ‘ operation. The handling and sorting equipment is successfully combined with a PC for laser triggering and generation of sorting signals. Singuiarization and transport of about 10 bottles/s is assessed as feasible. Nevertheless, in order to avoid that two bott~es arrive at the identification point at the same time, the handling device has to be carefulIy adjusted. In summary, after completion of the fundamental researc~ project, the possibility of laser-based identification of plastics is realistic. Further work has to concentrate on technical oriented development.

Acknowledgments This work was funded by the European Community in the 13rite-EuRam prograrnm, project no. BE-7717, contract no. BRE2-CT93-0589.