22nd Foss Grain Network Meeting DETERMINATION OFGOOD MILLING HUMAN CONSUMPTION USING HUMAN USING CONSUMPTION QUALITY OF FOR QUALITY WHITEMAIZE Quality is our passion G T HE RAIN S Dr Corinda Corinda Erasmus Dr OUTHERN L Dr Paul Paul Williams Dr Technical Technical experts: ABORATORY March 2016 March Wiana Louw A FRICAN NPC NIT OUTLINE OF THE PRESETATION     C CALIBRATION P C W   I NTRODUCTION ROCESS ONCLUSIONS ALIBRATION A M M HY FRICA AIZE AIZE DEVELOP PROCESSING PRODUCTION FOLLOWED MODEL AND A AND IN M S CONSUMPTION OUTH ROLL ILLING TO A DEVELOP - FRICA OUT I NDEX IN PROCESS S OUTH THE PRODUCTION IN SOUTH AFRICA

Ha T 4.000.000 1.000.000 2.000.000 3.000.000 OTAL 0 PRODUCTION 2003/04 RSA

2004/05

2005/06 AREA

2006/07 Season

2007/08 UTILIZED (11

2008/09

2009/10 SEASONS

2010/11 FOR

2011/12 MAIZE

2012/13 )

2013/14 Total Yellow White MAIZE PRODUCTION IN SOUTH AFRICA

Tons 10.000.000 11.000.000 12.000.000 13.000.000 14.000.000 15.000.000 4.000.000 5.000.000 6.000.000 7.000.000 8.000.000 9.000.000 2.000.000 3.000.000 MAIZE 2003/04

2004/05 (11 2005/06 PRODUCTION

2006/07 SEASONS Season 2007/08

2008/09

2009/10 ) IN 2010/11 RSA 2011/12

2012/13

2013/14 Total Yellow White MAIZE PRODUCTION IN SOUTH AFRICA

t/ha 2,00 3,00 4,00 5,00 6,00

2003/04

2004/05 MAIZE

2005/06 (11

2006/07 SEASONS YIELD

Season 2007/08

2008/09 IN 2009/10 ) RSA

2010/11

2011/12

2012/13

2013/14 Total Yellow White MAIZE PRODUCTION IN SOUTH AFRICA

Thousand Ton 1000 1500 2000 2500 3000 500 0 04/05 MAIZE (10 05/06 MARKETING 06/07 IMPORTS Imports 07/08 08/09 Exports AND SEASONS 09/10 EXPORTS 10/11 ) 11/12 12/13 13/14 MAIZE PRODUCTION IN SOUTH AFRICA

Thousand Ton 4600 4800 3200 3400 3600 3800 4000 4200 4400 04/05 (10 05/06 MAIZE MARKETING 06/07 07/08 CONSUMPTION Human 08/09 SEASONS Animal 09/10 10/11 ) 11/12 12/13 13/14 MAIN MAIZE PRODUCTION REGIONS WHITE MAIZE PROCESSED 2014/2015 SEASON Animal/industrial 1.469.002 25% T OTAL WHITE Bio MAIZE 0% - fuel PROCESSED FOR LOCAL MARKET : 5 862 438 Gristing 32.141 1% TONS 4.361.295 Human 74% WHITE MAIZE PRODUCTS Maize Meal 1.399.297 Rice, , 64,2% 83.907 3,9% W HITE M - Jul '15 AIZE - Dec '15 P RODUCTS Total Total White - Maize Products Maize Maize Chop 637.754 29,3% OtherProducts 57.327 2,6% : 2 178 285 t 2851782 : WHITE MAIZE PRODUCTS 1.142.583 81,7% Super W HITE - Jul '15 M AIZE - Dec '15 M - EAL Total Total White Other Meal 23.080 1,6% Maize Meal Maize 233.634 Special 16,7% : 1 399 297 t 2973991 MILLING INDEX CALIBRATION – WHY?    in a short of space time measuring the milling quality of large numbers of samples A need R soaking) extracted from whole maize using dry milling (no steeping or It is anindex on based therelative ofyield milled products unsuitable forfuture production products and could thereforeresult in specific cultivars being the maize industry as it affects theyield of high quality Milling Index (MI) of white maize cultivars is important for milling performance of maize for thedry milling Milling The Index (MI) provides an indication of the expected EQUEST for a non FROM THE - destructive method capable of MAIZE MILLING INDUSTRY : industry CORRELATION – COMMERCIAL MAIZE MILL VS LABORATORY SCALE MILL   laboratory scale mills commercial and Milled on both trials White maize cultivar  calibration reference method for scale mill as using laboratory continue with study correlation to Acceptable WHAT IS MILLING PERFORMANCE (OR MILLING INDEX – MI)?

 It is a test of suitability

 It attempts to simulate true yield and quality of products in a mill (in the case of a dry mill) using a suitable laboratory or pilot plant mill

 Products may include flaking grits, super maize meal, , ,

etc. PERFORMANCE  Yield and quality can be defined by various means including:

 performance repeatability

 cleanness of fractions MILLING

 maximum extraction of a target product (e.g polenta –

yellow or Special maize meal - white) MAIZE  market feedback re product eating quality SAMPLES USED FOR CALIBRATION    validation set fourth season’s samplesused a as developing the calibration and the Three season’s samplesused for seasons Trials repeated over 4 production parameters for milling quality and other quality three localities with three replications 35 white maize cultivars planted in ANALYSES PERFORMED ON THE SAMPLES Roff Colour Near Infrared Transmittance(NIT) 1000 100 Kernel Stress Breakage (%) susceptibility Kernel size TestWeight Moisture (5 Moisture Moisture Starch Protein Moisture Milling Index Roff Mill Kernel Cracks (%) - milling Hunter - - - - Milling Index Oven method NIR Datatec (% above 10mm sieve,above 8mm sieve,below 8 mm sieve) Mass (g) – Mass Hectolitre(kg/hl) mass fractions - Lab Colorflex done done in duplicate onNIR) 130 ° C every 15 th sample tosample NIR check LABORATORY SCALE MAIZE MILL

B1 Meal B2 Meal B3 Meal

Mill 1 Mill Mill 3 Mill B1 Grits 2 Mill B2 Grits B3 Grits

B3 Chop

Roff Roff B1 Chop B2 Chop Roff

Total extraction: (B1 Meal + B2 Meal + B3 Meal + B3 Grits) as a % of Whole maize

Each meal has different levels of starch, protein and fibre as well as different particle size, colour and cooking/eating quality BENEFIT TO THE MAIZE MILLING INDUSTRY 3,9 mil tons of white humanconsumption 3,9 mil tons of white humanconsumption maize processed for maizeprocessed for R 696 687 950 = 950 687 696 R 75% Maize meal 70% Maize meal extraction @ 25% Chop @ extraction @ 30% Chop @ R3500/ton R7000/ton R3500/ton R7000/ton 70% Extraction 75% Extraction EUR 41 371 018 371 41 EUR R R 19 507 262 600 R R 20900 638 500 R R 3 483 439 750 R R 4 180 127 700  Different NIT calibrations were tested using whole maize:

 Milling Index calculated on a dry base

 Milling Index adjusted to 14% moisture base

 Grit yield – 14% moisture base

 Total extraction – 14% moisture base

 Grit yield and total extraction are expressed as a mass % while Milling Index calculations are

MILLING INDEX CALIBRATIONS INDEX MILLING dimensionless index numbers (mass fraction ratios)  Comparative tests to evaluate efficiency of the NIT calibration:

 Same samples on the same instrument before and after a 12 month storage period

 Same samples on two different FOSS Infratec instruments

 Correlations of these samples with the laboratory mill

reference method

THE DATA THE ADDITIONAL TESTS TO EVALUATE ERRORS IN IN ERRORS EVALUATE TO TESTS ADDITIONAL  Effect of kernel size variation on Milling Index measurements

 To determine whether size classification of maize before milling can improve milling yield repeatability and process control

 Samples were classified and sieved according to different sieve sizes:

KERNEL SIZE VARIATION SIZE KERNEL 10 mm –  9 mm

 8 mm

THE DATA DATA THE ADDITIONAL TESTS TO EVALUATE ERRORS IN IN ERRORS EVALUATE TO TESTS ADDITIONAL

• Digital Image of maize kernels in Grayscale Inverted image - optimum contrast IMAGE ANALYSIS IMAGE

Binary detection and object measurements EFFECT OF MAIZE SIZE CLASSIFICATION ON NIT MEASURMENTS

NIT MI 100.0 110.0 120.0 130.0 70.0 80.0 90.0 >8mm Effect Maize of Kernel Size on NIT(mm) MI Maize Maize classificationsize (round hole sieves) >9mm >10mm Mix MI Sample 8 MI Sample 7 MISample 6 MI Sample 5 MISample 4 MI Sample 3 MISample 2 MI Sample 1 MI Sample MOISTURE OF THE WHOLE MAIZE KERNELS DURING SCANNING ON THE NIT

 Three sets of samples conditioned to a different moisture level namely 11%, 15% and 18%

 OBJECTIVE: spectral scans of samples with same

MOISTURE hardness but different moisture levels to test the – interaction between MI readings and moisture

levels

THE DATA DATA THE ADDITIONAL TESTS TO EVALUATE ERRORS IN IN ERRORS EVALUATE TO TESTS ADDITIONAL MI CLASSES USED IN INDUSTRY AS MEASURE OF MILLING QUALITY OF MAIZE:

MILLING INDEX MEASUREMENT MILLING QUALITY

<60 Bad milling quality

60 – 80 Acceptable milling quality

80 – 100 Good milling quality

100 – 120 Excellent milling quality

INDUSTRY STANDARD INDUSTRY >120 Exceptional milling quality MILLING AS USED AN INDEXCLASSES (MI) ADDITIONAL TESTS TO EVALUATE ERRORS IN THE DATA – MOISTURE on the air relative humidity sample may fall into a different class Note: Moisture content may influence MI to the effect that a - MI is variable depending I RANGE REPRESENTING SAMPLES G NDEXES ROUPS OF : SELECTED M ILLING A ADDITIONAL TESTS TO EVALUATE ERRORS IN THE DATA – MOISTURE THE S 2D THREE CORE PLOT DISTINCT FOR MOISTURE THE FIRST TWO GROUPS PRINCIPAL COMPONENTS OF THE RAW NIT SPECTRA SHOWING CONCLUSION:

 Sample moisture strongly interacts with actual Milling Index readings on the NIT

 Higher moisture values will give lower Milling Index values on the same sample

MOISTURE  Sensitivity is high, moisture variations of as low as – 1% showed significant variations in NIT readings

 Discussions - practical solution required

THE DATA DATA THE ADDITIONAL TESTS TO EVALUATE ERRORS IN IN ERRORS EVALUATE TO TESTS ADDITIONAL 3D PCA PLOT OF THE PRINCIPAL COMPONENTS OF THE SPECTRA OF THE SAMPLES STORED FOR 1 YEAR. SAMPLES DO NOT SHOW TWO SEPARATE GROUPS INDICATING RELATIVELY FEW CHANGES

DURING STORAGE. A FEW OUTLIER SAMPLES CAN BE SEEN

STORAGE

– THE DATA DATA THE

Blue – Year 1

ADDITIONAL TESTS TO EVALUATE ERRORS IN IN ERRORS EVALUATE TO TESTS ADDITIONAL Red – Year 2  RESULTS:

 Inverse linear relationship observed for moisture - can possibly be incorporated in the instrument calibration models

 This will improve general measurement and calibration precision for MI on the NIT

 Kernel size variations did not have a noticeable interaction with NIT Milling Index readings

 Calibrations can be developed for different

applications such as grits for breakfast MILLING INDEX PROJECT INDEX MILLING cereals or polenta PRIMARY OBJECTIVE:

 Roll-out of newly developed calibration model for use in the industry ADDITIONAL OBJECTIVES:

ROLL OUT ROLL  Collect outlier samples from industry – analyse to improve robustness of calibration

 Validate model on mixed samples

 Compare with different laboratory scale MILLING INDEX INDEX MILLING mill (Industry in-kind contribution) With gratitude to: The Maize Trust for financial support FOSS (Rhine Ruhr, South Africa) for assistance SAGL staff for milling and other quality testing

Thank you for your attention! ACKNOWLEDGEMENTS