Garry R. Kenny, Mitchell G. Roe
Novapak 97, 1997

Abstract

A sensor has been developed for identification of postconsumer plastic containers formed of PEN and PEN/PET resin blends. The prototype sensorwas installed in a plastics sorting module currently commercially employed using othersensor types. Tests of the system were conducted to determine the separation capabilitiesof the new sensor as installed in the separation module. Based on the test resultsobtained, a 2nd generation sensor was developed and installed in the separationmodule. The separation capabilities of the 2nd generation system were testingused a Design of Experiment methodology to maximize the information obtained from the testprotocol. This paper presents background information on automated plastics separation, anda summary of the identification and separation accuracy test results.

Background

Plastic Bottle Separation

Development of systems to automatically sort post consumer plastic bottles began in1989 (1). The first commercial systems to separate PVC from PET utilizingx-rays were introduced in 1991 (2,3) . Also in 1991, Magnetic SeparationSystems (MSS) introduced a multi-resin system for separation on PVC, clear and coloredPET, natural HDPE and PP, and pigmented HDPE plastic bottles using absorption of x-ray,visible and near infrared electromagnetic radiation. Recent MSS developments have alsobeen reported (4,5).

The use of automated plastic bottle separation equipment by processors of recycledplastics has increased substantially since that time. As of December 1997 approximately150 process lines or separation modules are operating world wide to automatically separateplastic bottles by resin type or color. Currently there are a number of suppliers offeringseparation systems (6,7,8, 9, 10) or stand alone sensors (11, 12, 13, 14) --although the operating equipment base has been provided by a limited number of suppliers.

From a technical standpoint it is desirable to develop and apply a single sensor typeto identification and separation of all plastic resin types and colors. However,experience in plants shows that bottles thrown away by consumers are often collected bydisinterested parties, stored in available or convenient lots, and baled in availablebalers (often those also used for municipal waste). As such, the bottles typically containcontamination, become distorted, have surface contamination added, and present challengesto any sensor system, which without actual operating experience, is unlikely to beanticipated. We have found that resin specific sensors in general provide betterseparation that any single sensor currently available.

Hence, in 1996 MSS began development of a sensor to identify PEN and PEN/PET blendswith support from AMOCO Chemical and Shell Chemical. It was recognized by AMOCO and Shellthat recycling of PET bottles was a large and viable industry, and that introduction of

PEN containers could potentially impact the end uses of the recycle PET unless the PENcould subsequently be separated from the recycle stream. Also, since PEN is more expensiveto manufacture, it was considered possible that it would be desirable to recover PENcontainers separate from PET due to their potential higher value.

The first and second generation sensors was installed in an MSS provided commercialseparation module designated the Binary BottleSort® (Fig. 1,2). Testing of the 1stgeneration system was conducted at the Wellman recycling facility in Johnsonville S.C.Wellman provided the equipment and support to expose the PEN test containers to thetypical recycle conditions of baling, debaling, and materials handling. The test PENcontainers were also provided with simulated labels using colored tape.

First Generation Sensor

During the testing of the first generation sensor somedeficiencies in separation performance were noted, specifically with regard to high PENcontent containers and small containers. Further, during the course of the first testphase, additional separation requirements were developed. Improvements in separationperformance were addressed by a redesign of both the sensor software and hardware.Separation efficiency was tested for a variety of PEN containers including: juicecontainers (5 % or less PEN content), carbonated soft drink bottles (25 %), wide mouthjars (40 %), and medical containers (100%).

The first area of improvement in the sensor performance determined was indifferentiating between containers with a high PEN content and very dirty PET containers.This was addressed by utilizing additional wavelengths for identification of dirty PETbottles. Also, a smart hardware system was developed and implemented which automaticallyincreased the gain of channels that produced small signals. This increased the dynamicrange of the channels and improved the identification of high PEN content containers.

The second difficulty encountered during testing of the 1st generationsensor was that many of the misidentified PEN containers were either small in size or hadonly small areas not covered by labels. Test results showed that the sensor was verysensitive to the exact manner in which bottles passed over sensor element. This wasaddressed by doubling the number of sensor elements, thereby increasing the resolution bya factor of four.

Another conclusion from the Phase 1 testing was that an improvement in purity of theejected fraction might be obtained by initiating air ejection pulses only at bottle areasidentified as PEN, as opposed to ejecting the entire object. This was determined by notingthat the 1st generation sensor algorithm sometimes identifies side-by-sidecontainers as a single object. Ejecting only PEN areas could reduce the chance ofaccidentally ejecting a PET container along with the PEN container.

Experience in development and operation of plastic separation systems shows that properfeeding of the material is at least equally important as the sensing system used. This wasagain confirmed during the Phase 1 testing at the Wellman facility. It was noted thatsurge feeding of the separation module was increasing the amount of PET which was carriedover with the ejected PEN containers. The feed system was improved prior to thecommencement of the Phase 2 testing. The Coca Cola Company provided additional support fordevelopment and testing of the 2nd generation sensor.

Background - Naphthalate-Based Copolymers and Blends

Amoco Chemical Company recently completed startup of a 27,000metric tons per year plant to produce commercial quantities of 2,6 DMN (the PEN monomer)..Shell markets polyethylene naphthalate copolymers derived from naphthalene-basedpolyesters under the HiPERTUFTM Resins tradename. High tere PEN copolymers,which are made up of primarily PET with 2,6 DMN as a comonomer, meet market hot fillrequirements up to 95 0C at a much more economical cost than the PENhomopolymer.

Due to its higher cost, broad use in the near future of the PEN homopolymer for foodand beverage packaging is not anticipated in the U.S. On April 4, 1996 the FDA published aregulation permitting PEN homo-polymer for use in direct contact with aqueous foods,nonalcoholic beverages, and beverages containing up to 8% alcohol. This is the first newpolyester food contact regulation in the U.S. in over a decade. One major market,specialty beer packaging, has indicated interest in the PEN homopolymer. It is reportedthat Anheuser-Busch Company is currently testing this new plastic. Other applications forthe homopolymer will likely be small niche markets requiring sterilization such as medicalpackaging.

The latest commercial-scale addition to the polyester family. according to the study,PEN reportedly provides superior performance over other polyester resins, most notably,polyethylene terephthalate (PET). In particular, PEN is said to boost electrical,mechanical, and thermal properties-in some cases by as much as four to five times that ofPET. The report also predicts that the U.S. market demand for PEN will reach 11.9 millionlbs in the year 2000 and 33.8 million lbs in 2000 - up from 1996 levels of only 2.3million lbs.

Although PEN was first synthesized in 1945, it has been available commercially onlysince 1990. Until recently, interest in this polymer outside of specialized applicationshas been limited by the expense and availability of a key component: 2,6-napbthalenedicarboxylic acid.

At present, Amoco Chemical Co. is the sole commercial producer of NDC in the UnitedStates. having recently begun production at its Decatur, Ala., facility Other players inboth U.S. and overseas are expected to enter the market in the future, however.This holds the potential of reducing PEN polymer costs, and, hence, expanding marketopportunities.

The major target markets for PEN resin include bottles, films, and fibers. In 1996,film was reportedly the only market with commercial applications - bottles andfibers were still in the developmental stage. One of the anticipated new uses is inhot-fill containers, such as jellies and jams, where PEN's filling temperatures surpassthe 85'C limit of ordinary PET.

Identification and Separation Efficiency Testing

Description of Field Tests

The field testing focused on verifying the performance of the new features added to the2nd generation sensor. The most important of these features is the enhanceddecision-making software added to facilitate detection of a wide variety of PEN containersthat are expected to be ultimately found in the PET recycling stream. This software hasbeen specifically designed to efficiently detect PEN containers which have beencontaminated with levels of dirt typically seen on PET containers originating in curbsidecollection programs in the U.S.

This new software and hardware were also designed to do a better job of estimating thenaphthalate level of a PEN/PET container. The sorting machine could be set to eject eitherall PEN, 25% or higher PEN, or only 100% PEN containers.

For the original trials, PEN containers were prepared by labeling them with coloredmasking tape and then baling them with post-consumer PET. For the more recent trials, thePEN containers were labeled with colored duct tape and then processed in an actual MRFalong with other collected recyclables.

As a results, these containers ended up being significantly more crushed and muchdirtier than those used in the original trials. In addition, the use of duct tape ensuredthat none of the labels of the PEN containers were torn off. This represents a worst caseas far as automated sorting is concerned since some of the containers ended up having noclear areas at all. Finally, in both trials, some of the containers were furthercontaminated with food residues to measure the effect on the sorting accuracy.

As before, the field testing setup at Wellman consisted of a series of conveyorsconfigured as a recirculating loop. Two conveyors, lit with black lights to make it easierto spot the PEN, were used to take away the ejected and passed bottles. The feed rate wascontrolled by placing a known weight of bottles in the loop and measuring the time for atypical container to complete one lap around the loop.

In a typical run, a quantity of post-consumer PET was placed in the loop along with apredetermined number of PEN containers. Sorting mistakes for PET and each type of PEN werethen separately counted for a period of ten minutes and recorded. The run was thenrepeated two or three times. When necessary, PEN containers were changed out between 10minute runs to ensure that a representative selection of containers was used.

Comparison of 1st and 2nd Generation Sensors

Because of significant differences in bottle preparation and feed system between thefirst and second set of trials, a deliberate attempt was made to repeat some of the testsfrom the first trial as accurately as possible. This is the only way to directly comparethe 1st and 2nd generation sorting machines.

Five types of PEN containers were used and are summarized below

Four selected tests from the previous trials were repeated. Eachconsisted of three 10 minute runs. The results are summarized below.

Based on these results, the 2nd generation sensor workssignificantly better than the 1st generation sensor. Using the ALL PEN sortsetting, all of the sorting accuracy’s have improved to 90 percent or higher. Sortingbetween different naphthalate levels also shows substantial improvement. The PET loss intothe ejected stream increased partly because a greater number of PEN containers were beingejected.

Another interesting result to emerge came from comparing the sorting accuracy when massfed to that when the bottles are fed in a singulated manner (one at a time). The 1stgeneration sensor performed quite well on singulated bottles, but the performance droppedoff significantly when the bottles were mass fed. The 2nd generation sensordisplays no such drop off. Sorting accuracy’s were the same for singulated bottles asthey were when bottles were fed with PET at 2500 lbs/hr. This is a direct result of thesoftware and hardware changes that are embodied in the 2nd generation sensor.

Second Generation Sensor Trial Results

The more recent trials using the 2nd generation PEN sensor were created by astatistician using the Design of Experiments methodology. The variables of interest werePET type (curbside, demand deposit, or curbside deposit), feed rate (2000, 2500, or 3000lbs/hr), weight percentage of PEN containers (3%, 9%, or 18%), and the sort setting (ALLPEN, 25% PEN, or 100 % PEN). The response factors of interest were PEN removal efficiencyand PET loss.

These criteria were used to generate a full factorial (which has all combinations ofall variables) consisting of 81 separate trials. From this, a fractional factorialconsisting of 9 representative trials was chosen for investigation. One of these trialswas designated as a control and was repeated several times under three differentconditions. The full statistical analysis of the results is ongoing and will be reportedat a future time. The discussion here will be limited to the control.

The control trial is described as follows:

Run time: 40 minutes as 4 separate 10 minute runs
Feed rate: 2500 lbs/hr
Sort setting: ALL PEN
PET type: curbside
PEN:        11.2 % CSD (carbonated soft drink, 25 %naphthalate content)
                       3.2 % HG (one liter narrow neck, 5 % naphthalate content)
                       3.2 % NWM (wide mouth jar, 40 % naphthalate content)
                       0.4 % SNB (one liter narrow neck, 100 % naphthalate content)

The control trial was run a total of seven times, including twice usingfood-contaminated containers and once with most of the labels removed from the PEN.

The results of all four runs of a typical trial are shown below.

The results are quite good, considering that all of the PENcontainers have full labels and are dirty. The HG and NWM containers have the largestlabels, limited their sorting accuracy and making for a great deal of variability. Thesmall amount of SNB present (only two bottles in the loop) also created a large amount ofvariability in the runs. The large PET loss results primarily from the large total numberof PEN containers (135) that were in the loop.

The total naphthalate content of the incoming bottle stream was 4.64 %, based on thenaphthalate levels of the individual PEN containers present. The naphthalate content ofthe passed stream was reduced to 0.56 %. The ejected stream had a naphthalate content of18.68 %. The ejected stream averaged 1 PET container for every 5 PEN containers present,which is good for an automated sorting machine of this type.

 The results of all seven control trials are summarized below.

The above results are quite interesting. The presence of foodcontamination reduced all of the system’s sorting accuracy’s, but by only asmall amount. In the previous trials, food contamination caused a much more substantialdrop in the performance of the 1st generation sensor. The systems sortingaccuracy’s on PEN containers with small labels are remarkable.

The 2nd generation PEN sensors’ performance appears to have only beenlimited by the fact that a PET bottle will sometimes cover up one of the PEN containers.This suggests that, in a realistic commercial application where some labels have come off,the sensor will perform extremely well.

Conclusions

An extensive set of trials has been conducted to evaluate theperformance of the MSS 2nd generation PEN sensor. Based on the results of thesetrials, the 2nd generation PEN sensor is clearly superior to the 1stgeneration PEN sensor, both in sorting accuracy and in differentiating between containerswith different naphthalate levels. The performance on dirty containers andfood-contaminated containers was also much improved. Further improvements are anticipatedin both the system software and hardware.

References

2. Rutgers University Center for Plastics Research
3. Govoni Simbianca
4. National Recovery technologies, Inc.
5. High Speed Plastic Bottle Identification - Industry Review and New Developments inPlastics Sorting - IdentiPlast Conference, Brussels, Belgum, October,1997, G. Kenny, M.Roe, R. Bruner.
6. New Challenges and Developments in Plastic Sorting - R97 Conference, Geneva,Switzerland, February 1997, G. Kenny and C. Crow.
7. Binder+CO AG
8. SRC Vision
9. Heckert Umwelttechnik GmbH
10. I.R.S. Limited
11. Sydel
12. ASOMA Instruments
13. Massen Machine Vision Systems GmbH
14. Optex
15. Umwelttech Analytik und Anlagen GmbH