J. Soc. CosmeticChemists20:3-15 (1969) ¸

1969Societyof Cosmetic Chemists of GreatBrttain

Drop testingof plasticcontainers E. K. CLARKE*

Presentedat the symposiumon "Technical developments in cosmeticspackaging",organisedby the Societyof CosmeticChemists of Great Britain, at Harrogate, Yorks., on 27th March 1968.

Synopsis--This paper discussessome ways of conducting drop tests, and comments on a few results

obtain

ed.

INTRODUCTION

Until recently almost all plastic containersfor toiletries and cosmetics were made from various gradesof polyethylene. These containers were virtually unbreakable.In the eyesof the customertherefore,"plastic bottles"

has come to mean "unbreakable

bottles".

Recently,however,there has beenconsiderable useof polystyrenejars and talc containers.Increasingnumbersof PVC bottlesare alsoappearing. Comparedwith polyethylene,all these containersare relatively fragile. Although such containersare usually strongerthan glass,breakagemay upset the customerbecauseit is unexpected. It is not suggestedthat the containersneedto be unbreakable,but that they are made reasonablyrobust so that breakageis not common.The degreeof strengthneededto minimisecustomercomplaintscan only be establishedby marketingcontainersof knownstrength.It is the determination of this strengthwith which this paperis concerned. Drop testingfilled containershasbeena favouredtechniquewith many *Beecham Products (U.K.), M•idenhe•d, Berks. $

JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS

people,probablybecauseof its apparentrealism,and the lack of apparatus or other facilities

needed.

There appearto be two basicways of conductingdrop tests:Cumulativedrop method

In this method, each bottle is dropped a number of times, each time from a greater height until a break occurs.The height at which the bottle breaksis recorded.After, say, 50 bottleshave beenbroken,the resultsare shown on arithmetical probability paper. The cumulative percentage brokenup to and includingeachheight is plotted againsteachheight. A straight line drawn through the resultswill showwith fair acccuracythe level at which 50% break and with lessaccuracythe level at which say,

10% break.TableI showsthe resultsobtainedon 50 smallPVC bottles. Table

I

The bottles were droppedfrom 0.Sin. If they did not break they were droppedfrom 0.6m, and so on, increasingby 0.Sin at a time until they failed. Only the height at which the bottles failed

Drop hr.

0.3111

was recorded.

0.6m

0.9m

1.2m

1

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x x

x

Totals

Cumulative%

3

6%

10

14

13

26%

54%

80%

10

100%

It is obvious that this test makes the assumption that the bottles are

undamaged by dropping,until they are droppedfrom a heightsufficientto break them. For bottles made of perfectly elastic, brittle materials this might be believable,sincethe matehalwill recoverafter dropping,unless the elasticlimit hasbeenpassed,whenit will break.In practice,observation will show that with toughenedpolystyrene and somePVC compoundsat

DROP TESTING OF PLASTIC CONTAINERS

5

least, dropping from heightsinsufficientto break them will causedefects to appear.Suchdefectsare bruisingof the surface,dents,and stressmarks. It is felt that this uncertainty outweighsthe merits of the technique. The principalmerit is that a result is obtainedfor eachbottle. Other techniqueswhich only break a proportionof the bottles, do not, on the face of it, make suchgooduseof small numbersof samples.I prefer the "staircase method", describedbelow, for small samples. Single drop method

When usinga singledrop method a variety of statisticaltechniquescan be used. Two of these demonstratetwo basically different approaches. Staircase

method

This is a techniquedesignedto concentratethe testing around the level at which 50}/0 of the samplesfail. To conduct a drop test by this method one choosesan initial drop height and an increment.Thus, one may decide to start at lm, and move in stepsof «m. This choiceis usually influenced by past experienceof the type of containerbeing tested.The test is conducted using the following rule:If a samplefails, drop thenext sampleat that height lessthe increment; if a sample doesnot fail, drop the next sample at that height plus the increment.

Table II is built up as follows:Table

2.1m

x

1.8m

0

1.Sm 1.2m

II

0

x

x

0

0

x

x

X X

0.9m

X 0

0.6m

0 X

0

X x

0

0

X

0

0

x X

0 0

X x

0 0

0 0

0

The entries are made from left to right, in order as the test proceeds,

O----pass; X -•failure

In orderto calculatethe level at which 50%/0of the samplesfail, the results are tabulated

as follows.

First

it is determined

whether

there were fewer

JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS

passes,or failures. Table III

is constructedusing whichevereventswere

the fewer. Table

III

f

fh

zf

[I•(fh)]

H----ho q-d

H ho

= •-

d

=

___•

50% drop hr. lowest height at which a pass or failure occurred depending on which

f h

= = + • =

Increment

is used to construct

the table.

used.

Number of occurrences at each height. The heights are numbered, 0, 1, 2, etc. from ho upwards. -• is used when failures are used to construct table.

+• when passesare used.

Formulae are given to calculatethe standard deviation, one is:-

Where

s=1.62d Z(•2)[ ZI] -[-0.029

s is the standard

de•ation.

From this it would be possibleto calculatethe level at which, say, of the sampleswould fail assumingthat the resultswere a nomal distribution.

(s)

H• = Hs0•l.• •

It is suggestedthat this type of calculationis a misuseof statistics. The staircasetechniquespecificallyforcesthe resultsto lie closeto the mean, and is an attempt to reducethe spread.To usesuchresultsto estimatethe normal spread must be grossly inefficient. However, when only small numbersof samplesare available,this methodis probably one of the best ways of estimatingthe 50% level. Probit

test

In my view, an estimateof the spreadis useful,and for this reasonand alsobecauseno calculationis required,a Probit test (1) is preferred. A minimum of 50 samplesare divided into five groupsof ten. If there is

DROP

TESTING

OF PLASTIC

CONTAINERS

no priorexperience with the particularsampletype beingtested,a further fivesamplesaretaken.Thissetof five samples is usedto estimatea suitable starting height, by droppingeachoncefrom a differentheight. The first set of ten samplesis nowtestedat the chosenstartingheight,eachsample beingdroppedoncefromthesameheight.Thenumberof failuresisrecorded. The next set of ten is now droppedfrom a height 25% higher or lower, depending on whethertherewerelessthan $ failuresor morerespectively, from the first ten.

After examiningthe resultsof droppingthe first two setsof ten, judgment is usedto determinethe drop heightsto be usedfor the next three sets.The aim is to drop eachset from a different height, and to cover as wide a range of heightsas is possiblewithout producingeither ten failures or ten passesin any set of ten; suchresultsare uselessfor the subsequent graphicaltreatment. The resultsarenowexpressed aspercentages and plottedon arithmetical

probabilitypaper,and the bestline drawnthroughthem. The 50ø/0level can now be read off the graph to a moderate accuracy,probably about

+10%. Estimatesof the percentage of failuresto be expectedat other heightscan alsobe read off the graph,althoughthe accuracyis steadily reducedas either end of the spreadis approached.It will be noted that no calculations

are needed.

An exampleis given below. The resultsare recordedin the order in which they occurred. The first set droppedat 2.4 m. It gavet3failures. Sincethis was more than 5, the next height was chosenas 25ø/0lower than the first. Sinceonly 3 failed, a set was droppedat the intermediateheight of 2.1m. These results suggestedthat dropping from 2.7m would give a

fairly highrate of failure;7 failureswerefound.The last setweretherefore droppedfrom3m in orderto spreadthe graphout further. Table

IV

Drop height

Failures

2.4m 1 .Sin 2.1m 2.7m 3ra

This examplewas chosenbecauseit gavea fairly goodgraph (Fig. 1). However, even resultswhich are very scatteredstill allow of a reasonable estimateof the 50% level. The valuesaway from this level are, of course,

impossible to estimatesensibly(Fig. 2).

JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS

3.3

3-0

2.7

2'4

o

0-1

I

2

5

10

30

50

Per cent

70

90

98

99

failure

Figure 1 PVC bottles: 10 dropped at each height.

3'0

2'7

u* 2'4

E 2'1

.?

121

1',5

1'2

0'9

O. I

2

I0

$0

Per cent

50

70

90

98

failure

Figure 2 Effect of different lines drawn through scattered results

99.99

99

DROP

TESTING

OF

PLASTIC

CONTAINERS

COMPARISON BETWEEN CUMULATIVE AND SINGLE DROP

A technicaldata sheet (2) dealing with Vinatex/Dorlyl PVC bottle compoundsshowsone comparisonbetweendropping50 bottlesby a cumulative method, and a test where 50 bottles were dropped from each height. A 50% level of 2.7m was found by the cumulative method, and of 2.9m by the singledrop method. This seemsa very small difference,but only representsone experiment. We have not donemuch work on this point, but two experimentscan be quoted.

1. Bottles of a flattened flask shapewere used, weight 13g, madefrom ICI's PVC PCO 638. The results obtained were lower than can be achieved

with this compoundbecausethe bottles were producedon a machinerun which causedthe PVC to burn slightly.

(a) Cumulativemethod. 50 bottles were droppedgiving the results shownin TableI. The 50ø/0level foundfrom the graphwas0.87m. {b) Singledrop. 21 bottleswere droppedby a staircasemethod.The resultsare given in Table V.

Table

2.1m

x

1.8m

0

1 .Sin 1.2m 0.9m 0.6m

V

0 0

x

x

0

x

x x

x 0

0 x

0

x x

0

0

0

0

The 50% level is calculated as 1.33 m.

2.

Bottles used were from the same mould as above but were made from

Vybak VZ905.

(a) Cumulativemethod. 50 bottles were dropped as in experiment (1). The resultsare shownin Fig. $.

10

JOURNAL

OF THE SOCIETY

OF COSMETIC CHEMISTS

S•ngle drop

2.7

2.4

•Cumulaf•ve 0,9

0'1

,,, 2

I0

Per

50

cent

50

70

xdrop , ,, 90

98

99'99

fellure

Figure $ Comparison of single drop with cumulative drop.

(b) Singledrop, Probit method. 50 bottles were droppedin five sets of ten. The resultsare shownin Fig. 8. It can be seenthat the Cumulativemethod givesa 50% level of 1.17m, and the SingleDrop methoda 50% level of 1.56m.

TECHNIQUE

Droppingposition The bestresultswill be obtainedif the bottlesare droppedin the position which will give the most breakage.For many bottlesthis positionwill be on their bases, or on a corner of the base, because this is where the section

changes rapidly,andis usuallythinnedout. It is worthwhile, while experimenting with any new bottle, to find the most fragile positionfor drop testing.

DROP

TESTING

OF

PLASTIC

CONTAINERS

11

A variety of methodsare usedto try to make the bottles fall in a consistentmanner. In a techniquedescribed(2), the bottlesare droppeddown vertically mounted guide tubes. This method introducesthe danger that the bottles will bounce back and forth from the sides of the tubes.

Impact testersin which a table falls away from objectsplaced on it, are very useful.If the bottles are being droppedon to their basesit may be possibleto drop severalat once. If the bottles are to be droppedin any other attitude, jigs will be necessaryto hold them. Thesemust be carefully checkedto seethat no rotation is imparted to the bottles as the table falls away.

Some use a shaped rubber pad connectedto a vacuum source.This holds the bottle in the predeterminedattitude, and when the vacuum is released,the bottle should fall cleanly. Small bottles can be dropped by hand with reasonableresults. If the containeris held betweentwo fingersat a point vertically above the centre of gravity it shouldfall with very little rotation. However,it is difficult to drop them without impartingsomeslight twist. Dropping surface

It is considered important to drop the containersonto a smooth,massive surface.It is suggestedthat a steelplate at least 6 mm thick, embeddedin concrete,is ideal. Plain concreteis not recommended,partly becauseit is of variable smoothness,and partly becauseit will wear badly. A steel surfaceas describedwill test the containersmore rigorouslythan normal domesticfloors. However, it is consideredthat it is most important that consistentresults are obtained, since these can probably be correlated later with practicalexperienceof customercomplaints,etc. It might also be consideredthat a measureof exaggerationin a test is a useful safety factor. RESULTS

A few resultsobtainedin developinga PVC bottle for a wave set may be of interest.The bottle has a flattenedshape,90 cm3 capacityand weighing 13g. GradesPCO 634 and 747 are modified gradesof PVC. PCO 638 and VZ 903 are unmodified.It is interestingto note that there is a muchbigger difference between the modified and unmodified grades when tested at 10øC,than when tested at normal temperatures. The singlefiguresgivenin TableVI do not, of course,giveany indication

12

JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS Table

VI

Difference between bottles made from various grades of PVC. Bottles dropped on to bases. Grade

Drop height in metres 10øC

Room temperature

50% level ICI

PCO

634

ICI PCO 747 ICI

PCO

638

Vybak VZ 903

3.4 3.2 2.5 2.2

10% level

50% level

10% level

2.6 2.3 1,6 1.3

2.6 1.2 1.1

1.4 0.6 0.8

of their variability. There is as yet not sufficientdata to get a true picture of the variability of the testingsincethe resultshave beenobtainedduring production runs when conditionsare not yet consideredto be optimum. This meansmachine settingstend to differ between runs. However as an example, the figure of 2.2 m given as the 50% level for VZ 903 at room temperature, coversresultsvarying from 1.6 m to 2.6 m. For comparisonthe 50% level for a range of our standardglassbottles is from about 0.3 m to about 1 m dependingon the bottle. Very small glass bottles, of course have much more strength, and a shampoo minibottle has a 50% level of about 2.5 m. On the face of it the strength of the PVC bottles quoted seemsto be satisfactorycomparedwith glass.Whether this is sowill not be knownuntil moredata on customercomplaintsis available. After producingabout three million, only txvobottles have been returned with broken bodies.One of these appearsto have been broken by excessivesqueezing,the other by dropping. (Received.'31stJanua•7 1968) REFERENCES

(1) (2)

Shell Plastics CPL/PS/4. Vinatex/Dorlyl PVC bottle compounds(Vinatex Datasheet).

DISCUSSION

MR. W. G. BURDON:Could you tell uswhether you have any experience of changeof impact strength xvith time, particularly in relation to vibration during transport. Also any experiencewith the influence of the contents on impact strength. I believe it is well known that solids, for instance, will stand far greater heights but I wonder whether intermediate physical products, such as thick creams, will also influence the drop height.

DROP

TESTING

OF PLASTIC

13

CONTAINERS

THE LECTURER:I have no direct experienceof these points with PVC bottles. I have a lot of tests on bottles which were being continually developedwith changing machineconditions.They are now fairly well settled, and in the future I hope to have someinformation on how they will changeas the material ages. Vibration from travel tests is concernedwith the viscosity of the product. Because of the messynature of drop testing, I confinemy drop testing, in the main, to filling the container with water, but we have done work on the impact strength of shampoo sachets, with different viscosity shampoos,and there is no doubt that the more viscous the material

in the container, the more resistant the

container

is to

shock. I believe that a gelled product, however, is much worse, transmitting the shock straight through. I have not carried out any drop testing with bottles filled with viscousproducts.It may be interestingbut on the whole I believewater is one of the most stringent test fills. MR. D. A. DEAN: What is the accuracy of fill, as a + 5 cm3 variation on a 90 cm3 bottle could be significant?

THE LECTURER:We fill the bottles with the same volume as used in practice; we use a volumetric filler to fill all our drop test containers, with a standard fill - in this case 90 cm3.

MR. D. A. DEAN: I suggestthat customer reaction is more important than just detail of numbers of customer complaints. May I ask if thosecomplainantsinvolving PVC bottles have been questionedas to exact cause? Is there any correlation between customer reaction on glassversus PVC bottles - as related to breakage? THE LECTURER:As far as the complaints are concerned,as a company we have not made any detailed survey, as far as I am aware, of reactions to these particular bottles. As far as salesfiguresare concernedthe product is selling rather better than it was before we put it into a plastic bottle. Whether this has anything to do with PVC or the fact that it is merely plastic, or the fact that it is a different colour, I would not like to say. DR. K. G. JoaNsoN: Just a small supplementary point to Mr. Clarke's comments. This product was originally filled into a glassbottle, and we went on to a bottle almost identical with the one Mr. Clarke has illustrated, but in opaque HD polythene. The saleswere virtually unaffected, but one or two large stockists refused to handle the opaque bottle - they insistedthat we produce a transparent bottle and for that reason we went on to PVC in the virtually identical container. Since then there is no question that saleshave increasedvery considerably.We had no adverse reaction. We had four customercomplaintsof breakagewith the PVC bottle - we never had complaints of glassbottle breakages,although we know that they break.

DR. C. V. SOWTER:You have explained how to derive the height at which 50% of the bottles break, and I can see that the technical purposesare a good reason for doing this but a marketing man is not interested in the fact that only half of the bottles will break. He is more concernedwith knowing that, say 1% will break if dropped from 1 m, or something like this. Would you recommend a separate test to provide figuresof this sort or would you say that a 50% level of breakageis the only real sensible

measurement

to make?

14

JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS

T•E LECTURER:I would not say that the 50% level is the only realistic one that one can make, but it is probably the only practical one becauseof the quantity of bottles necessary.If I wanted to find a level at which 1% broke, I would use the technique describedhere as the probit techniquein order to estimate the spread,but if we think about 1%, it means that for every 100 bottles dropped, one will break at that height, and a single bottle breaking does not give a very good estimate. It means that 5,000 bottles must be dropped. In order to answer your question scientifically I would have to drop several thousand bottles by a probit method and plot the results.

MR. C. D. RENFREW:I could perhaps offer another explanation for the very low impact figure you have obtained with the cylindr'cal bottle. In early work on PVC bottles we found that its notch sensitivity was very important in impact testing and that when setting up moulds for blow-moulding, if the two faces of the mould were deliberately offset as little as 0.127 mm this had the same effect as removing ten parts of impact modifier, i.e. if you had a ten parts impact modified material this would reduce the impact strength of the bottle to the unimpact-modified state. I wonder if the mould was not properly aligned in your tests on the cylindrical bottle becauseit is a dramatically low figure. T•E LECTURER:This is very interesting. Mr. Renfrew, what is a satisfactory temperature for cold drop testing? I have done a certain amount of testing at 10 ø because it is at a low temperature that the difference between an impact modified grade and a non-modified grade really shows up. I have chosen 10ø as being not too cold to be unrealistic.

MR. C. D. RENFREW:I completely agreewith you. The effect is even more dramatic at 5 ø where the effect of the impact modifier allows it to give a very significant effect - it is perhaps unrealistic to test at 5 o MR. P. G. S•.TON: We use an injection-moulded polystyrene talc container with a welded bottom. Initial evaluation by drop tests at 1 m in the laboratory gave no failures at the weld, but in use the bottoms dropped off. Do you feel that this should have been predicted by drop tests? T•E LECTURER:It depends how you drop them. We had some problems with a similar container, and we investigated its strength not by drop testing, but by an impact test on the side of the container. We found that the basespopped off very neatly if they were not stuck on very well. I should think in a drop test you are holding the base fairly firmly while it takes the shock. We filled our bottle with talc and dropped it on one corner so that it dropped at roughly 45 o.

MR. D. R. CRABTREE:Surely dropping it on its base is the most critical situation becauseyou get the most hydraulic effect which is reduced by dropping it on the corner.

T•E LECTURER:I do not seewhy the hydraulic effect should differ with the point of impact. You are still getting a shock wave through the liquid; in fact, you are funnelling it into the corner if you drop it on to its corner. The problem with dropping it on to its corner is the difficulty to avoid imparting rotation and some of the force is lost as rotational energy on impact. By dropping onto the base you get a better

DROP

TESTING

OF

PLASTIC

CONTAINERS

15

shock. If you could drop them so that they hit firmly on the corner you would probably get a good result. MR. C. E. Hm•s: In relation to orientation of drop, I am absolutely convinced that there is a major differencebetweendropping onto the bottom and dropping onto the flat - I do not accept that dropping onto the bottom is the most severe condition. I am surprisedthat nobody has referred to dropping on the cap becausethe shoulders are so frequently a source of potential problem. I would like to suggest that when one drops a bottle which has a pronounced difference between the minor and major axis onto its large, flat face, one gets,what I believe is called in America, 'hydronamic hammer', and one imparts a shockto the liquid which encouragesit to flow violently at right angles to the plane of impact. Thus one will tend to knock out the face and knock off the shoulder,and one is therefore imparting a shockto the bottle which is diametrically opposedto the shock that it received when dropped onto its face. In relation to viscosity of the product I entirely endorse what you have said. I believe that it is unconformingto test with water, becauseone always suspectsthat one should test with the product, but whenever I have tested in strictly comparable series, between product and water, I always come to the conclusion that it was very reasonable

to use water.

I believe that 10ø is very reasonablefor testing, but I believe it is important to remember that bottles are frequently stored in quite a cold cellar and might be brought on to the production floor very cold. They can be well below 5 ø, and when first being filled in these conditionsthere is little doubt that they are probably in their

most sensitive

state.

'['}•,; Ll,:cTum,zu:Your comments are very useful indeed. Our bottles are closed with a long polythene top which would cushion the impact quite well, and it would therefore be unrealistic to drop them onto the top. Our factory storage is heated so we do not have that problem which you mention. Mm R. J. Fom.•t•s: I believe that in Scandinavia they use higher impact grades (luring the winter.

MR. L. G•T•D: [ think it is important to remember that some people may be packing liquids with a very much greater density than water, and that the load inside a bottle can be very high. I have experienced considerable problems of container breakage with a sugar syrup. The water drop tests did not reveal the problem but dropping with the product did. Admittedly we were not doing these in PVC bottles but the differ nee of the weight of the contents was a point that we had to note. MR. D. B. VI•c•: What part does the ullage play? I would rather have thought that the larger the ullage the more the liquid would be free to move, and the more the hammering effect at the time of impact. Yet you mention worse results with a gel?

TH• L•C•URF.R: My commentson the gel are based on very little experiencebut we have looked at a few containers with gels and they did seem remarkably fragile, and they do transmit shock rather well.

I have not studied ullage systematically on our container but when investigating sachel strength in the past, this was an important point and I seeno reasonwhy it is not here. I do not know, however, whether there is a critical level.