45. The Ultimate in “Downskilling”: Infrared Spectroscopy in Automobile Scrapyards
Patrick Hendra and Peter Mucci
A few years ago the Ford Motor Co approached the Prototype Group in the Engineering Faculty and asked if it could come up with a non-sampling spectroscopic method for analysing the plastics in scrapped automobiles. In scrapyards there are basically two levels of technology. In low-tech yards the cars are piled up and then dismantled by customers who want bits – doors, engines, bumpers, etc. When the hulks have ceased to be of interest to the punters, they are further stripped to produce heaps of steel, aluminium, plastics and the rest (upholstery, glass etc.). Crushers then produce a marketable “lump” of steel, the aluminium is valuable, but the rest is pretty near valueless. In fact, it may cost the yard owner money to dispose of it.
The alternative – high-tech yards – remove the engines and transmissions and then smash the rest into small pieces (~ 5x3cm in size). Magnetic and other crude separations produce steel and other outputs but the plastics, upholstery and small metal parts are then usually buried in the ground. Since they are inevitably horrible mixtures contaminated with oil and hydraulic fluid, this can be very environmentally unfriendly and costly.
46. In Europe, an EU directive has been agreed requiring manufacturers to re-use materials to certain degrees, ever increasing over the years, and Ford were concerned with this problem. If plastic bits were removed from cars, could these be categorised and sorted so that the recovered material would be marketable, hence making the whole dismantling process more financially viable? The Prototype Group therefore asked us to work with them and Fords to come up with a fast, gorilla-proof, analytical procedure.
47. We tried various hairbrained ideas and then, in collaboration with Dr. John Graham, found the answer – mid i.r. specular reflectance. As explained above in para. 19 the mid-infrared reflection spectrum of a plastic is very different from the absorption one. Further, the most specific data comes from optically opaque specimens such as carbon loaded polymers – just the kind of rubbish we were being asked to sort. We recorded specular reflection spectra of several polymers typical of the motor industry, all loaded with carbon or ground slate. A few examples are given in Fig.1.
Figure 1: Specular reflection spectra of typical plastics
Quite clearly the spectra, as spectra, are dreadful but they are different. Our first intention was to use the Kramers Kronig algorithm to derive the absorption component and then use this for search purposes, but the KK process is too slow. We returned to the differences between the reflectance spectra – why not create a library of these and use them as a base for a library search? The problem here was the level of background. We quickly found that examples of the ‘same’ polymer – say polypropylene used in bumpers, heater casings, or air cleaner covers, gave similar spectra but very different backgrounds. A simple way of overcoming this problem is to differentiate. Let me explain.
48. The first differential of a spectrum with respect to cm-1 makes all absorption bands look bisignate.
49. The background is a signal fairly constant in absorbance as the wavenumber is traversed, hence it has very little gradient, i.e. dA/dcm-1 is small.
As a result, we have
become very similar
50. If it turns you on to see your spectra looking like such, a second differential will return the peaks (but not the background), i.e. we have
We chose to stop at the first differential. The process takes milliseconds so there is no real penalty in time. The derivative reflectance spectra of various specimens of the same polymer look fairly similar to absorption ones – similar enough that they can be used to generate a database. See Fig.2.
Figure 2: First derivative spectra equivalent to those in Fig.1
51. Initial results were very encouraging – view a specimen of a piece of automobile for about six seconds – differentiate – library search – analysis, a total time of about eight seconds.
Two problems remained – samples had to be prepared for analysis i.e. they had to fit into the accessory on the spectrometer – hardly an operation appropriate to a scrapyard.
The second problem was grease, dirt and a generally non-spectroscopic environment. Ruggedisation was critical. To tackle the first problem, a special accessory was designed which placed the sampling point above the instrument lid and the viewed plane horizontal and downwards facing – see Fig.3.
Figure 3: Upward facing spectral reflection accessory
Obviously, inclusion of a window and perhaps a gentle flow of air would remove any possibility of contamination of the optics with dust, oil, or grease. It also made the construction of an all-enveloping case over the spectroscopic system easy. With careful design, the optical system was very efficient with an excellent signal reaching the detector.
52. Now sampling – bits of old car are rarely flat, almost never clean, and may be coated with paints. Plastic bumpers are a good example. Exposed unpainted ones are often polypropylene or polycarbonate filled with short strands of glass fibre, rubber modifiers, plus ground up slate and perhaps some talc, and carbon black.
53. Painted bumpers can be of polyester filled with long glass fibre, as used in boat hulls, or polybutylene terepthalate or even a polymer ‘alloy’ based on a variety of typically impact resistant compounds, all covered in several layers of paint (which itself will be inhomogeneous – primer, paint, overglaze layer, etc.).
On top of all this, a bumper is large, unwieldy and far from flat. To solve the sampling problem a simple rotary cutter can be incorporated in the complete machine or be mounted nearby on a bench. This shaves off a millimetre or two of the surface, removing greases and of course any paint, and producing a small area of flat surface. Experience, however, has shown that only about 10% of analysis needs surface preparation.
Figure 4: Reflection spectrum obtained from a Ford Escort (1982) front bumber. Lower, 1st derivative.
Analysis: Polycarbonate/polybutylene tereplthalate blend.
54. The instrument works, see Fig.4 – the process was patented by Ford and a commercial machine duly produced by the Prototype Group. To our delight, the machine caught the eye of the Science Museum (Britain’s collection of things scientific and technological, based in a large building near Central London’s Imperial College and the Royal Albert Hall). The Museum was creating a new “Challenge of Materials” gallery and hence our beast had found instant fame! The time required for an unskilled operator to examine a lump of rubbish with infrared and see the analysis displayed on a VDU is less than 8 seconds.
55. The obvious next question is cost – is it worth it? Perhaps the question doesn’t need answering because somehow the plastics have to be re-used to satisfy the EU regulations, but in the end economics always rears its ugly head. However, the machines are getting cheaper and a complete system is now around the price of a luxury car.
Polypropylene, P.V.C., polyethylene and polystyrene all cost around $1000 a ton as virgin material but nylon, polyesters and polycarbonates are two or three times more valuable. Recovered scrap is of variable value – poor quality, dirty, poorly sorted material has almost no value and can even be costly to dump. At the other extreme, because some retailers want to be GREEN, some scrap is in short supply!! Even though it is expensive to clean up and re-use.
56. The story is then clear – recovering large bits of cheap plastics is worthwhile and even medium sized lumps of valuable ones. It is not worth removing small widgets of anything, but bumpers, headlights, tailgates, air cleaners, dashboards, and the myriad of larger pieces, are worth the effort of dismantling and identification. Just think, a bumper weighing 10kg, even at only $150 a ton, is worth $1.50. Removal and sorting need take no more than a minute – sounds viable to me.
REF: Int.J. Vib. Spect.,[www.irdg.org/ijvs] 1, 4, 45-56 (1997)