1984_brokensha_a trail to compare

8/7/2019 1984_Brokensha_A Trail to Compare

1/6

Proceedings of The South Af ic an Sugar Technologists' Association - une 1984 9

A TRIAL TO COMPARE CORE SAMPLING WITt-I THE FULLWIDTH HATCH METHOD OF CANE SAMlPLlNG

By M. A. BROKENSHASugar Industry Central Board

Keywords: Cane sampling, Core sampler, Hatch samplerAbstract

Core sampling using the single barrel overhead inclined corerhas been compared with the full width hatch metho d of canesampling. Core sampling precision taking 3 cores per vehicleload was found to be considerably p oorer than hatch sam plingand there was evidence of bias when extraneous matter waspresent. However, the magnitude of the differential bias wasgenerally small. Capital and operating costs for the core samplerare higher than for the full width hatch sampler.

IntroductionIn the South African sugar industry, sampling of preparedcane via the full width hatch sample r1 s the official metho d ofsampling for cane payment. The full width hatch sampler ispositioned at the end of the cane ca me r immediately prior tothe cane feed into the extraction plant and as such, is situatedsome distance downstream from the cane receival point at themill. As a consequence, mill yard cane han dling procedures areconstrained to a considerable extent because individual groweridentity of the cane loads has to be maintained up to the canesampling point.The new Felixton I1 sugar mill incorporates features in thecane yard off-loading and cane preparation systems that aredesigned to give optim um efficiency and cost savings. However,

there was concern that these features would give rise to inter-mixing of adjacent cane loads on the carriers to the extent thatit would be difficult to differentiate between the cane from dif-ferent suppliers as it passed the full width hatch sampler. Coresampling permits sam pling of the cane in the delivery vehicleand therefore intermixing of cane after offloading is not of con-cern with this method.Accordingly, in June 1982, the Council of the Sou th AfricanSugar Association (SASA) agreed that the Central Board CaneTesting Service (CTS) should co nduct a n extensive exam ina-tion of core samp ling with the v iew to its possible implen tationat the new Felixton I1 sugar mill. This investigation was con-ducted during the 1982183 season under the direction of a SASAappointed Technical Sub-committee.

Equipment .Core sampling equipmentCore sampler: The Felixton I1 mill is designed to handle canewhich is delivered in Hilo road vehicles and SAR trucks -both are large capacity vehicles each carrying more than 18tons cane. For these transport modes, the overhead inclinedform of core sampler was considered to be most suitable anda single barrel sampler, similar in design to that described byAshe,' w as constructed for the trial. However, unlike the m odeldescribed by Ashe in which the barrel lifts up before dischargingthe sample, the later model discharges directly into a screwconveyor without first lifting the discharge end into position.This permits the travelling camage system to be located at fourpoints (via locating bearings) instead o f only two a s in the caseof the earlier model, making for greater stability and smoo theroperation du ring coring.

The overhead corer is inclined a t an angle of 45" to the hori-zontal and from the retracted ,position the rotating barrel(500 rpm), fitted with a crow n cutter, moves dow n a slide fora total travel of 5 metres. Speed of travel is 10 met redm inuteand d epth of penetration into the cane bed is to within 400 mmof the vehicle's m ain floor level. 'To increase cane carrying ca-pacity there is a recessed section in the centre of the vehicleand cane in this section was ina.ccessible to the corer. Afterretraction, the cane core (k 13 kg,) contained within the barrelis eiected via a piston into a screw conveyor which transfersthesample to thk cane preparation equipment.Initially, the crown cutter was made of heavy gauge (7 m mthickness) 200 mm diameter tubing containing 16 large, rather

coarse teeth (45 m m long) which were splayed with a 3 m mset. It failed to cut cleanly and tended to sever the billets witha mangling, twisting action. Trash tended to be dragged intothe sample from the surrounding cane. The crown cutter wasreplaced with a unit consisting of a rolled band saw (3 m mthick) containing much smaller an d finer teeth (12 m m pitch;12 mm deep) with no set. Its cutting action was found to- beconsiderably better, bu t twisted severance of billets an d ballingof trash was still detected.Hippo mill: The particle size of the sample taken by the corerhad to be reduced for efficient sub-sampling. For this purpose,a H ippo mill (Model 69) was posilioned at the e nd of the screwconveyor leading from the corer discharge point.

Initially, problems were experienced with the Hippo mill.Th e orifice plate frequently choked; strong up-draught windageexisted and finely prepared cane tended to build up in largeam oun ts on the internal walls of the machine. On the otherhand, the degree of preparatioil was excellent. Followingnumerous trials with various types and sizes of screens, areasonable comp romise was obtained by removing the screenand repo sitioning the washboard between the 3 a nd 5 o'clockpositions. The base of the drum was fitted with a solid platecontaining one large, 150 mm diameter aperture.Sub-sampler: From the H ippo mill the cane passes thou gh atwo stage sub-sampler of similar design to that described in th eLaboratory Manual for South African Sugar Factories. Thesampler used in the trial required careful tuning to give a re-presentative sub-sample of the stratified core sample. Stratifi-cation of the sample commences at the corer since the coresample is usually taken diagonally across the bundle of canegiving rise to a stratified sequencle of butts, middle-stalk andtops within the barrel. This stratification is retained throughthe Hippo mill and hence through the sub-sampling systemand even into th e sample receiving bin. To overcom e this re-quires a far greater degree of tuning than is necessary withprepared cane taken via the full width hatch - n this lattercase the cane has been thoroughly prepared and very well mixed.

Both flip-flaps were set to operate at maxim um cut frequency(2 seconds accepted/2 seconds reject). Proper bulking up of thefeed to the second flip-flap was achieved by reducing the speedof rotation of the screw conveyor from 50 rpm to 25 rpm andby reducing the pitch from 300 nnm to 150 mm . It was alsonecessary to mix thoroughly the contents of the sample priorto hand sub-sampling. This was time consuming and arduo usand in practice it would be necessary to increase the numberof autom atic flip-flap sub-samplers.


2/6

zedings of T he South African Sugar Technologists' Association - une 1984Reject material conveyor: The reject material from the sub-sampling station was pneumatically conveyed back to the ve-hicle via a large blower.Sample shredder: The cane preparation out of the Hippo millwas still too coarse for analysis (somewha t poorer than the millprepared cane) and accordingly, was given a further shreddingin the sample shredd eri used for the cane sampled via the fullwidth hatch. After shredding, the cane from the two streams(core and hatch ) were indistinguishable in terms of preparation .Full width hatch sampler and ancilliary equipment.

The full width hatch s ampler used in this trial is the standarddesign sliding gate type used routinely for cane testing opera-tions and is described in the Laboratory Manual for SouthAfrican Sugar Factories. The two stage sub-sampler and sampleshredder are also routinely used standard units as described inthe same manual.Vehicle marshalling system.

A traffic light system was designed and installed to controlthe marshalling an d m ovem ent of vehicles at the core station.Vehicle positioning was controlled with the aid of guide rails,road h um ps and beacons. The driver was made to alight fromthe cab whilst the corer was being operated (red robot signal).Laboratory analytical equipment.

All items of laboratory equipme nt used in the trial were thestandard items used routinely in the CTS for direct analysis ofcane as described in the Laboratory Manual for South AfricanSugar Factories. ProcedureThe core sampler was installed at the Darnall sugar mill.A relatively high proportion of its cane crush is delivered inHilo vehicles but there are no rail deliveries. The absence ofrail deliveries on which to test the corer was not considered ashortcoming as the orientation of the cane bundles within aHilo is more varied than in a railway truck and if the coresampler can cope satisfactorily with a Hilo then it is reasonableto expect that its operation with the railway truck would alsobe acceptable. Other factors in favour of conducting the trialat the Darnall mill at that time, were the extended crushingseason at th is mill (well int o February, 1983) which perm ittedmore time for the trial and that extraneous matter measure-ments, which were needed for the trial, were conducted on aroutine basis.The trial was planned to run round the clock and two personsper 8 hou r shift, were employed to attend to the core samplingoperations while a few additional items (notably extra ovensfor cane moisture determ ination) were added to the cane testing

laboratory. The target was to inspect three or four Hilo loadsper hou r with the core; these consignments would also be testedvia the CTS routine using the hatch sampler. Selection of Hilovehicles for core testing was essentially random, although thecomputer was used to m onitor the num ber of tests in each testcategory (burnt, trashed, variety, load configuration, etc.) sothat corrective action could be taken if the number of tests ina particular category were too low. Standard core samplingprocedure was to take 3 cores along the central longitudinalaxis of the vehicle.Extraneous m atter and ash O/o cane (soil) tests were conduc tedby the milling staff in accordance with the procedures pre-scribed in the Laboratory Manual for South African SugarFactories.

Precision and bias.The two desired properties of cane test results obtained bya cane sam pling system are tha t the results should be reasonablyprecise or repeatable and that they should not be substantiallybiased relative to the true value.

Of particular importance is that whatever biases exist, theyshould be consistent o r should differ only slightly for differenttypes or categories of cane.Testing for bias: As a check on the hatch sampler, 100 testswere done to compare the "static" an d "dynamic" modes ofopera tion of the hatch. This involvecl captu ring a single hatchfall-out in the routine ope rating mann er (this is the "dynam ic"sample). The cane elevator was immediately stopped and thehatch d oor fully opened a nd held in th is position w hile a wholeslat load of cane was carefully fed into the open aperture sotha t the entire load was captured ("matic" sample). The lattertechnique is the best means of sampling that could be devisedand the results obtained for the static tests were taken as thestandard.

The only means for testing the core sample for bias in thisinvestigation was by comparing it with the hatch result. Tocheck the consistency of any b ias, deliveries were classified andexamined in terms ofvariety of cane,. ethod of infield loading,bundle configuration within the Hilo,burnt or trashed,age of cane at harvest,wearlperiod of use of core cutting head,overall time delay (burnt/cut to crush),time delay - ore sample to hatch sample,trash % cane,tops O/o cane, andtime of day of core sampling.

Testing for precision: With hatch sampling, it is possible totake two samples from a delivery which can be regarded asrandom and independent. This is because only a small portionof the consignment is not available to the samp ler due to theoverlap with the adjacent consignments. The method entailscollecting alternate hatch fallouts separately and is describedby Buchanan and Brokensha.'Such a procedure is impracticable ,with core sampling underthe con ditions of this trial because a relatively large portion ofthe consignment in the Hilo is inaccessible to the corer and toobtain two effectively random and independent samples, itwould be necessary to go to the extreme of unloading and re-loading the delivery vehicle before taking the second sample.

Sub-sampling efficiency at the corer: Due to the unmixed andtherefore stratified nature of the samp le material yielded by thecorer, it is essential that the sub-sampling procedures be effi-ciently performed. To ensure that adequ ate efficiency was beingattained, tests were conducted wherein 3 cores were capturedand processed via the normal routine i.e. Hippo mill prepa-ration followed by au tom ated sub-sampling with two flip-flaps,hand sub-sampling incorporating thorough mixing, coning-and-quartering, and a final particle size reduction in the sampleshredder. Three further cores were captured from the samevehicle (immediately adjacent to the original 3 coring posi-tions). Each of these cores was processed separately throughthe Jeffco-cutter-grinder (which yields a very finely preparedmaterial) and each of these three portions of Jeffco preparedcane was thoroughly mixed and sub-sampled to a small amountvia coning-and-quartering. These 3 end-amounts (one from eachcore) were then mixed together to form the sample which wasanalysed for comparison with the routine result.Core versus hatch test comparisons for individual growers: ,Ledger accounts were compiled for indiv idual g rowers, reflect-

,ing core versus hatch test comparisons for individual consign- ,ments and in respect of weekly and to-date averages.


3/6

Proceedings of Th e South African Sugar Technologists' Association - uneGrab sampling performance: In the course of this trial it wasfelt that it would be useful to determine the grab sampler pre-cision and accordingly, a limited number of tests were carriedout wherein Hilos were sampled and tested using core versusgrab versus hatch comparisons.

Results and DiscussionPrecision.

The total number of deliveries tested by the two methodsover an 11 week period was 3 058.Th e variances of the core and hatch cane test value from allthese deliveries, and the derived variances of the sam pling er-rors are shown in Table 1.

TABLE 1Variance of cane test results and derived sampling error variance

Hatch CoreHatch Core Core-hatch "TRUE" sampling sampling/ test 1 test 1 diferences I variation I error I errorVariances of test results(3 058 deliveries)

I I

Derived estimatesof varianceI I

Note: Variance = (S.D.)'It is clear that the variances of core values exceed those ofthe hatch by a considerable margin for all the, cane test para-meters. For example, in the case of pol Oo cane, the core and

hatch figures are 1,48' and 1,08' respectively.

I

Pol Oo caneBrlx % caneMoisture %caneFIbre OocaneP u r ~ t yNon-polFIbrelpolratlo

The derived estimates of the variance of sampling errors forpol % cane for the core and hatch samples were 1, lo 2and 0,432,an d the estimat e of variance of the true values for all deliveriesis 0,99' (see Appendix I). Stepping up the num ber of cores from3 to 6 improved the precision (0,2 units pol O/o cane and 0,3units fibre O/o cane) but still left it well above the hatch level.

1,48'1,25'2.51'2,58'4,76'0,58'0,37'

1,08'0.92'1,93'1.95'3,80'0,37'0,25'

Initial work showed core precision to be much poorer thanthat shown in Table 1: An intensive campaign of special testsrevealed that the sub-sampling system design and operationwas not effective with the stratified nature of the core samples.The system was modified as described earlier.Confirmation that the sub-sampling system was operatingefficiently after the modifications, was provided by the pro-cedure described earlier in which tw o sets of 3 cores were takenadjacent to one another and one set processed via the normalroutine i.e. Hippo mill etc. while the second set was processedthrough th e Jeffco cutter-grinder, hand m ixed and sub- samp ledetc. in the m anner described.

1,18'1,03'2,08'2,23'3,87'0,50'0,31'

The precision found by the special sub-sampling routine(Jeffco cutter-grinder preparation) compared closely with thecorresponding estimate obtained with the data yielded via themodified Hippo/flip-flap sub-sampling system.It is concluded that the level of precision reflected in Table

1 is all that can be expected from the corer taking 3 cores pervehicle. The core samplers' principal limitation is the smallam oun t of cane which it captures. A 3 core sample taken froma 20 ton consignment would be approximately 40 kg whereasthe corresponding hatch sample would be 200 kg (8 fallouts at25 kg per fallout). Furtherm ore th e materi al sam pled by thehatch is well mixed whereas that samp led by the . corer isstratified.

0,99'0,82'1,69'1,65'3,32'0,41'0,23'

Efects of core sampling precision on the accuracy of growerspol % cane: weekly, monthly and seasonal averagesThe following table contains the estimated standard errorsof a grower's weekly, mon thly an d !ieasonal pol Oo cane avetagesobtained with hatch and core sampling, calculated at variouslevels of cane production. These estimates do not take intoaccoun t the possible effects of differential bias. It simply affordscom parison of th e effects of the (different levels of precision

associated with the two different sampling techniques. Coreprecision is taken at f ,l units of pol Oo cane and the hatchprecision at +0,43 units of pol % cane.TABLE :2

Impact of core sampling precisiorn on grower'gtest results

Tons consignmentscane per

0,43'0,42'0,93'1,03'1,83'0,29'0 1

I,lO20,94'1,86'1,98'3,41'0,4120,29'

Th e lesser precision rendered by t he core sam pler has a neg-ligible effect on seasonal da ta. However where only a few testsare involved, the average result obtained with the core sampleris subject to significantly more enror than that obtained withthe hatch.

season

Grab sampling precision.Arrangements were made to sample Hilos by three samplingtechniques: core, grab and hatch. Sampling precision estimatedfrom the core/grab-hatch differe nn:~or 268 Hilo consignmentsare given in Table 3. TABLE .3

Week

Precision: core and grab samplersI I

Mon'h

Std. dev. of a consignmenttcst: 3 grabs I +0,76 1 + 1,831'01 % cane

Std. dev. of a consignmenttest: 3 cores I a 1.09 I 2 1,98

SEHatch

Fibre % cane

Space available for the grab saimpling tests at Darnall wasrestricted and the tests were conducted under adverse condi-tions. Nevertheless the pol precision of the grab method ismuch improved compared with the corer.Earlier work at Felixton sugar mill in 1979 (involving 5 500consignment inspections) produced the following data: stan-dard deviation of a consignment lest with grab sampling; pol

O/o cane + 0,54 and fibre O/o cane i 1,46.

!iE SE / SE SE / SECore Hatch Core Hatch Core

I I

Dlferential bias.To determ ine the extent of non-,ran dom sampling error, the(core-ha tch) differences were related to the other variables ormeasu reme nts which were detailed earlier (e.g. variety, me thodof loading etc.)A m ultiple regression analysis was don e for the variable s pol

% cane, fibre % cane, purity and fibre/pol. The R 2 (adjusted)values obtained were:Pol O/o cane . . . . . . . 0,O 13Fibre Oo cane . . . . . . 0,034Purity . . . . . . . . . 0,007Fibre/pol . . . . . . . . 0,026

The proportion of the variance accounted for by the factorsdescribing the cane (variety, method for loading, etc) is small.Clearly the differences are largely due to random sampling er-rors (or possibly som e other unmeasured factor).


4/6

12 Proceedings of T he Sou th African Sugar Technologists' Association - une 1984Hatch bias.

In this case the hatch was investigated for bias by carryingout 100 dynamic versus static tests. The results of these testsare shown in Table 4.Hatch bias tests

I I( Pol % cane I Fibre % caneMean of static tests 1 13,43 1 16,12Mean of dynamic tests 16,06Residual difference + 0,06Standard error of difference k 0.03 k 0.09

There was no statistically significant evidence that differenceswere affected by the level of extraneous matter present orwhether the cane was burnt o r trashed. These results were takenas confirmation t hat the hatch is not subject to significant bias.Core sampling bias.

This assessment is based o n the data obtained after the im-provements to the crown cutter and to the sub-sampling tech-nique. Comparison between the core and hatch sampling resultsfor 3 207 consignments a nd on a weekly basis, is shown in Table5.

TABLE 5Comparison of core and hatch sampling

1 I Pol % Cane 1 . Fibre O/o Cane( Difference I

Although the pol and fibre differences are statistically sig-nificant they are relatively small. The pol difference is fairlyconsis tent from week to week whereas for fibre, the consistencyis not quite as good. A classification of the results into burntand trashed cane (and ignoring consignments with a mixtureof bu rn t.an d trashed cane) gave results as indicated in Table6.

1 Difference

MeanStd.error

TABLE 6Core bias: burnt and trashed canedifferences (core-hatch)

I , I I I I

I Pol O/o cane ( Fibre O h cane ( No. of tests

3 207

TABLE 7

11,86

Burnt caneTrashed caneDifferential biasStandard error

Core bias - trash and tops % canedifferences (core-hatch)

11,95

The bias in pol % cane is relatively small. However, the fibredifference is substantial and statistically significant.In Table 7 the classification of the core-hatch differences intotrash and tops % cane is shown. Tests conducted prior to thosereported in Table 6 have also been included in Table 7.

- ,02-0,12-63k 0,OS

There is evidence of a differential bias in pol for both trashand tops Oo cane (see Table 7). A large bias is apparent in polfor trash levels greater than 10/o. However, this result is notbased on as many tests as for the lower levels of trash.

Trash O/o caneLess than 5%Between 5 & 10%Greater than 10%Tops % caneLess than 3%Between 3 & 6%

Further classification in terms of age of cane, harvest-to-crush time delay, infield loading method and Hilo loadingmethod are shown in Appendix 2. I[n these categories wherethere is a pol bias, it is negligible. The fibre bias in the twocategories age of cane and infield loading method are how ever,not so easily dismissed.

-0,09?c0,02

-0,05+0,47-,52t 0,07

Within the framework of all the tests conducted on the corefor bias assessm ent, there is statistically significant evidence ofdifferential bias but at magnitudes which are either small ornot based on large numbers of tests.Inspection of pol O/o cane averages for individual growers asdetermined by the core and hatch test methods are comparedin Appendix 3. It is seen that whilst the majority of growersexhibit differences between their core and hatch pol % cane to-date averages that are relatively small, there are a num ber w hohave differences that are rather high.

Pol % cane(Std error)

-0,04 (?0,03)-0,lO (?0,03)-0,45 (+O,I I)-0,05(+0,02)-0,20 (?0,04)

8072 276

Operational logistics of core sampling versus hatch sampling.Unde r the conditions of this trial the single-barrel corer tooksix minutes to capture 3 cores from one Hilo consignment.This delay was viewed with concern by the haulier in view ofits potential to increase the turn-around time of vehicles in thecane yard. By speeding up th e ope ration of the corer, it wasestimated that at best, a single-barrel corer manned by oneoperator would require at least 4 minutes to capture 3 coresfrom one Hilo consignment.It is understood that at Felixton 11. the capture of 3 cores

17,58

from a single SAR consignment would need io be completedwithin two minutes; outside of this 2 minu te interval, the chainof SAR trucks would be moving (or prone to movement). Itwas estimated that a triple-barrel corer should be able to com-plete the capture of 3 cores within 2 minutes; the core stationmay have to be manned by 2 persons. Two triple-barrel corerswould be needed for SAR consignments at Felixton: one foreach tandem.

Initially, two tandems will be operating at Felixton 11, andtogether will be capable of processing 30 Hilo consignmentsper hour, thus restricting sampling time to 2 minutes per ve-hicle. One triple-barrel corer would, theoretically, meet thisneed.In time, when 3 tandems are operating, a second core sam-pling station would be required for road transport, and a thirdstation would be required for SAR consignments.If hatch sampling is employed at Felixton 11, one hatch sam-pling installation would be needed at each tandem.

Fibre % cane(Std error)

+0,25 (+0,05)+0,31 (+0,06)+0,56 (+0,20).+0,31(k0,04)+0,25 (+0,08)

Costs: core sampler versus hatch sampler.

No. of tests

1 6711 375

1592 4 1 6 '730

17,26

At current prices, the cost of a single-barrel corer (installed)is approximately R160 000. It is estimated that a triple-barrelunit wou ld cost in the region of R250 000. Th e cost of a hatch

0,32t 0,04


5/6

- une 1984 13on is approximately R85 000. The costs of m a-ance (and replacements) of hatchng equipment, amou nts to R 10 000 per hatch samplingon, per a nnum . T he equivalent figures for single-barrel

Occurrence of mechanical failures was minimal over the 20Conclusion

The trial has shown th at the degree of imprecision associated

Standard deviation of a test

Bearing in mind tha t this imprecision has to be superim posednto the real variations in the quality of cane from one con-signment to the next (t-0,99 units of pol % cane and + 1,65units fibre Oo cane) an individual growers' test results wouldfollowing fluctuations around his m ean figure:

Hatch samplcr

Standard deviations

Pol O/o cane+ 0.43+ 1.10

At the higher imprecision level encountered with the corefor a single test, there is the risk that cane analysis may losecredence with growers even though the individuals' seasonalaverage will not be seriously affected.The poorer precision with the corer is because of the limitedquantity of cane captured during coring; it failed to gain any-where near as good a representation of the cane within a singleconsignment as compared w ith the hatch. It is possible to im -prove the precision by taking mo re cores over a wider coverage.However, an increase from three to six cores per vehicle pro-duced only marginal improvement and it is apparent that togain a meaningful improvement, the additional coring time,engineering effort and costs involved would probably beprohibitive.In terms of bias assessment, it is concluded that within theframework of the current tests, there is statistically significant

evidence of differential bias but at m agnitude s which are eithersmall or not based on large numbers of tests. As far as theapital and operating costs are concerned, it has been foundhat those for the core sampler are higher than the correspond-ing costs for the full width hatch sampler.

Fibre % canet- 1,041 ,98

Th e investigation has verified that, viewed solely from a caneampling stand point, hatch sampling is unquestionably m orecient, convenient and cheaper than core sampling.

Fibre % canek 1,95t ,58Hatch samplerCorc sampler

The efforts of members of the SASA Experiment Station,Sugar Milling Research Institute and the Cane Testing Servicein the practical application of the trial are greatfullyacknowledged.In writing this paper, the author has drawn freely from thereports prepared in the course of the trial by the SASA CaneSampling Technical Sub-com mittee.

Pol % cane+ 1.07? 1.48

REFERENCESI. SASTA Laboratorv Manual for South African Suaar Factories. 2nd edition.1977: 4.2.2.1.2. Ashe, G. G. (1979). A cane core sa mp l~ ngnstallation a t Umfolozi - ProcS .!fi Sug Techno1 Ass 53: 76-77.3. Buchanan, E. J. and Brokensha. M. A. (1974). The application o f directcane testing to the South African sugar industry - Proc in1 Soc Sug CaneTcc,hnol 15: 1456- 1469.

APPENDIX IDERIVATION O F SAMPI.ING VARIANCES

The observed ca ne test value for a delivery, obtained by a particular samplin gmethod, is postulated to be made up of the following (additive) components:I. the true content of the delivery;2. an overall average bias associated with thc particular sampling method;3. a non -random sampling error (N RE) asr;ociated with the particular categoryof cane being tested, and the sam pling rnethod em ployed (differential bias);4. the random sampling error (RE) associated with the sampling method.

T a k ~n ghe example of pol % cane, the variance (V) obtained from all 3 05 8deliveries tested by both the core and hatchi methods are:Vtt,.,,,.,,l V l , t , t + V,,tt t,,.,,,t,,- VItt ,I,; ,,,,,,= 1,08'Vi,,,,,,= V 11 + V,nl.,,.,,,,,+ VI,,-i,,,,,l= 1,411'

-V,,,,,,.l,:,,,.l,, Vwul,:,,ct,,+ V~L,I,;,,,I,,)(Vw1,, , , , , , I + Vnl.t,.,,,,,t)= 1318'Solving these equations gives

Vwnl il,,,,,l,i+ Vnl:il,:,,,t,, = 0-43'V,,,,,,,,,,, + Vnl: ,,,,,,, = 1,lO'V I U I I I = 0,99'

This me ans of estimation relies on the i~ssump tion hat neither method issuch that it causes variation to be reduced. (T he use of first expressed juice andJava Ratio would perhaps have been such a method).APPENDIX 2

DIFFERENTIAL BIAS ASSESSMENTI. Sort by age of cane

Differcnces (core:-hatch)I Pol % cane 1 Fibre O/o cane ( No. of tests

13 - 15 months16 - 18 monthsDiKercntial biasStandard error2. Sort by harvest to crush time delay

Differcnces (core-hatch)Hours delay I Pol % cane I Fibre % cane I No. of tests

Less than 24 hours25 - 48 hours49 - 72 hours -0.093. Sort by infield loading method

Differences (core-hatch)Loading Method 1 Pol % cane I Fibre % cane I No. of tests

Category ACategory BDifferential biasStandard errorCategory A = mechanically loaded by Funkey-Bell into tip trailer.Category B = hand loaded into bundles then mechanically loaded via sideloader into trailers.4. Sort by Hilo loading pattern

Differences (core-hatch)Loading pattern

Category ACategory BDifferential biasCategory A = all bundles lodged horizontally within the vehicle.Catcgory B = some bundles angled upwards within the vehicle.


6/6

14 Proceedings of The South African Sugar Technologists' Association - une 1984APPENDIX 3

TABLE O F COMPARISONSIndividual grower's to date pol % cane averages hatch versus core (includes 'only those with 4 0 or more dual tests)

Pol Oh caneGrower 1 c s Diference Standarderror

1984_brokensha_a trail to compare

Documents