dr eugene donskoi - project leader, process …m-n.marketing/downloads/conferences/ecic2016...12...

“MINERAL4/RECOGNITION4: A UNIVERSAL OPTICAL IMAGE ANALYSIS

PACKAGE FOR IRON ORE, SINTER AND COKE CHARACTERIZATION”

Dr Eugene Donskoi - Project Leader, Process Modelling - Downstream Processing Performance

12 September 2016

Carbon Steel Futures group/Resources, Community and Environment Program/ CSIRO Mineral Resources

• Textural Characterisation

Chemical assay

Mineral composition

Textural composition

Texture – microstructure which relates both the spatial distribution of different minerals and porosity, and the spatial distribution of different grains and crystals

H = hematite (martite), HH = hydrohematite,vG = vitreous goethite, oG = ochreous goethite,K = kaolinite, P = pore, E = epoxy resin

WHY ?

1. Ore texture is important for prediction of downstream processing

2. Particles within the same textural group have analogous properties

• Reflected Light Microscopy1. Capability to reliably identify different iron oxides

and hydroxides2. Higher resolution, faster and cheaper than QEM

SCAN and Raman spectrography

• Increased productivity (up to 10 times higher than manual point counting)

• Higher level of characterisation (quality and amount of information, consistency)

MINERAL4/RECOGNITION4: A UNIVERSAL OPTICAL IMAGE ANALYSIS PACKAGE FOR IRON ORE, SINTER AND COKE CHARACTERIZATION | Presenter - Dr. Eugene Donskoi

Examples of four hematitic particles exhibiting different textural types ranging from dense to highly porous


These different textures have different characteristics in terms of hardness, attrition

resistance, moisture absorption, and will behave differently during comminution and

beneficiation processes as well as granulation and sintering

Comparison of modelled and experimental TI values for runs included in the modelling

a) modelling without textural information (SD = 1.53, R-Sq = 75.6%

TI without textural information for runs

used in modelling

55

60

65

70

75

80

55 60 65 70 75 80

EXPERIMENT

MO

DE

LL

ING

TI with textural information for runs used in

modelling

55

60

65

70

75

80

55 60 65 70 75 80

EXPERIMENT

MO

DE

LL

ING

(b) modelling with textural information (SD = 1.13, R-Sq = 86.9% )

Not a single outlier has been removed (either from the modelling set or the verification set) !!!


CSIRO OPTICAL IMAGE ANALYSIS (OIA) PACKAGE (‘MINERAL4’ and ‘RECOGNITION4’)

• Initially created for iron ore and now successfully used for other minerals

• Allows automatic, semi-automatic and manual acquisition and comprehensive processing of sets of images

• Works for any particle size including lumps

• Allows users to build their own iron ore classification scheme and to perform automated iron ore and gangue texture classification

• Imports images in all major formats including output from QEMSCAN and MLA

• performs calculation of mineral composition, chemical assay, density, dimensional or textural characteristics for every particle section, liberation class and ore texture class or any particle group based on specific mineral, dimensional, textural or chemical criteria

• performs calculation of mineral liberation, iron liberation and mineral association characteristics for any group of particles


Parameters //Part Type Hardness

No. of

Part Freq

% Area

% Wt % Mean Fe Tot

Mean SG

Mean Mineral

Area (µm2)

Mean Shape Fact

Mean Elong

Shale-Kaolinitic Soft 52 0.43 0.36 0.18 10.02 2.11 3757 4.15 1.7

Microplaty hematite Medium 1016 8.39 8.37 8.05 69.05 4.15 4485.3 3.22 1.76

Dense hematite V.Hard 5143 42.45 43.87 48.93 69.2 4.81 4643.1 2.47 1.75

Vitr.-Ochr. goethite Soft 1097 9.05 8.66 6.66 58.42 3.32 4294.5 3.08 1.78

Martite-Goethite Hard 1022 8.44 8.7 9.34 67.96 4.64 4631.6 2.59 1.78

Martite-Goethite Medium 1158 9.56 9.83 9.41 67.19 4.13 4622.2 3.09 1.73

Goethite-Martite Soft 404 3.33 3.29 2.56 61.66 3.36 4429.8 3.28 1.72

Goethite-Martite Hard 858 7.08 6.97 6.52 63.53 4.04 4424.2 2.92 1.76

Ore texture group characterisation (example of output to Microsoft Word)

This tables includes the following parameters grouped by iron ore texture type: the number of particles, frequency %, area % and weight % of this ore texture group, total iron and specific gravity with coefficients of variation (standard deviation/mean), mean area with coefficient of variation, mean shape factor and elongation, percentages of each mineral within each class by area or by weight, and the percentage of pores by area.


Part Type / Mineral

Keno-Magnetite

(Wt)

Hematite

(Wt)

Hydro-Hematite

(Wt)

Vitreous Goethite

(Wt)

Ochreous Goethite

(Wt) Kaolinite

(Wt)

Quartz

(Wt)

Pores %

(Area)

Shale-Kaolinitic 0.0876 1.83 0 5.31 9.65 80.12 2.99 23.08

Microplaty hematite 0.0971 97.92 0.0569 1.26 0.49 0.13 0.0405 18.11

Dense hematite 1.57 97.02 0.15 0.85 0.36 0.0493 0.0033 5.27

Vitr.-Ochr. goethite 0.053 1.39 0.0408 75.65 21.3 1.53 0.0262 17.56

Martite-Goethite 0.93 84.88 0.59 11.65 1.86 0.0573 0.026 6.24

Martite-Goethite 0.58 79.19 0.19 15.86 3.94 0.22 0.0265 14.84

Goethite-Martite 0.37 34.07 0.18 45.94 18 1.41 0.0403 21.28

Goethite-Martite 1.18 38.97 0.61 53.08 5.97 0.13 0.0493 10.9

Sample of Brockman ore with some addition of martite/kenomagnetite ore

corresponding mineral mapOptical photomicrograph


Sample of Marra Mamba ore

corresponding mineral mapOptical photomicrograph


Identification of martite and microplaty hematite using textural identification

Original reflected light photomicrograph Mineral map (martite - yellow, microplatyhematite - magenta)


Automated identification of primary and secondary hematite


Problem – the same mineral, so similar reflectivitySolution – segmentation by textural identification

primary hematite – light blue, secondary hematite – dark blue, magnetite –magenta, SFCA – green, glass – cyan, porosity and epoxy – yellow

Automated identification of Fibrous, Columnar SFCA and Dense SFCA(silicoferrite of calcium and aluminium)


Problem – the same mineral, so similar reflectivitySolution – segmentation by textural identification

primary hematite – light blue, secondary hematite – dark blue, magnetite – magenta, fibrous SFCA – light green, Columnar SFCA – cyan (relatively large elongated grains, prismatic structure), glass – dark green, porosity and epoxy – yellow

a)

Comparison of sinter phase quantification by manual point counting (PC) and automated optical image analysis (OI) techniques


0

10

20

30

40

50

60

70

80

90

100

Blend1 -PC

Blend1 -OI

Blend2 -PC

Blend2 -OI

Are

a %

glass

dicalcium silicate (larnite)

dense SFCA

prismatic SFCA

micro-platy SFCA

magnetite

secondary hematite

primary hematite

Automatic Identification of Inert Maceral Derived Components (IMDC) from Reactive Maceral Derived Components (RMDC) in Coke

Currently there are three methods used for IMDC identification :

1) bulk identification of IMDC2) small size porosity IMDC

identification3) Identification of “washed out” areas

Magenta – IMDC, blue – RMDC, yellow - porosity


Textural identification of IMDC based on identification of small size porosityMap of fine porosity

IMDC areas obtained based on map of fine porosity

Photomicrograph of coke

Porosity map

Final IMDC map from fine porosity method


Segmentation of IMDC boundary layersInitial photomicrograph of coke Identified IMDCs and RMDCs

IMDCs only with allocated boundary layers

The wall abundance (RMDC) in the IMDC boundary area is 7.4%, while the overall abundance of RMDC in the sample is 23.6%)


The change of RMDC abundance and thickness in IMDC boundary area depending on IMDC area.


20

30

40

50

60

70

80

1 10 100 1000

RM

DC

ab

un

da

nce

Area /10000 (µm2)

RMDC abundance in IMDC boundary

Coke1

Coke 2

Coke3

Coke 4

0

5

10

15

20

25

30

35

40

1 10 100 1000

RM

DC

th

ick

ne

ss

Area /10000 (µm2)

RMDC thickness in IMDC boundary

Coke1

Coke 2

Coke3

Coke 4

Correlations between parameters characterising parent coal composition, coke structure and measured coke properties.


6055504540

90

80

70

60

50

40

Porosity

Vit

rin

ite

12

11

10

9

8

76

5

4

3

2

1

Scatterplot of Vitrinite vs Porosity

0.00240.00220.00200.00180.00160.00140.00120.0010

80

70

60

50

40

Specific Neck Thickness

Vit

rin

ite

12 9

8

76

5

4

3

2

1

Scatterplot of Vitrinite vs Specific Neck Thickness

250002000015000100005000

65

60

55

50

45

IMDC Boundary Length

MC

S

Scatterplot of MCS vs IMDC Boundary Length

MCS - arithmetic mean coke size in mm. determined by hand sizing over 125.0, 106.0, 75.0, 50.0, 25.0 and 13.2 mm square mesh screensAREA_1 – average pore area, ModWT- modified wall thickness, ECD-equivalent circle diameter

700000600000500000400000300000200000

66

65

64

63

62

61

60

59

AREA_1

AS

TM

Har

dn

ess

12

9

7

6

5

4

3

2

1

Scatterplot of ASTM Hardness vs AREA_1

0.0550.0500.0450.0400.0350.030

66

65

64

63

62

61

60

59

Ratio of ModWT and ECD

AS

TM

Har

dn

ess

12

9

7

6

5

4

3

2

1

Scatterplot of ASTM Hardness vs Ratio of ModWT and ECD

50454035302520

65

60

55

50

45

Percent of Nodes Area

MC

S

12

9

8

7

65

4

3

2 1

Scatterplot of MCS vs Percent of Nodes Area

Dependence between coke strength indices and cumulative amount of IMDC

ACARP Project C25048 - kickoff meeting | Eugene Donskoi

-1

-0.9

-0.8

-0.7

-0.6

-0.5

-0.4

-0.3

154 217 307 435 615 869 1229 1738 2458

Co

rre

lati

on

Co

eff

icie

nt

IMDC ECD (µm)

Correlation Coefficients between M40 and Cumulative amount of IMDC down to certain size fraction

-0.9

-0.8

-0.7

-0.6

-0.5

-0.4

-0.3

154 217 307 435 615 869 1229 1738 2458

Co

rre

lati

on

Co

eff

icie

nt

IMDC ECD (µm)

Correlation Coefficients between I40 and Cumulative amount of IMDC down to certain size fraction

Dependence of correlation coefficient of M40(AS1038) and I40 on cumulative amount of IMDC down to a specific size of IMDC expressed in equivalent circle diameter (ECD).

Large Object Characterization: Pellets, Lumps, Pebbles - Magnetite pellet heated to 620OC

Ø 12.7 mm, compound image made of 18 x 18 2 x 2 MosaiX imagesclear hematite presence in outer layers

Individual image:

2 x 2 MosaiX at magnification x100

magnetite – pinkhematite – whiteporosity – dark


Large Object Characterization: Pellets, Lumps, Pebbles -Magnetite pellet heated to 620OC

compound mineral map

clear hematite presence in outer layers

magnetitehematitefluxporosity


Magnetite pellet heated to 620OCMineral/porosity distributions

magnetite hematite porosity


Graph of average hematite, magnetite, and porosity abundances as a function of distance from pellet centre.


Conclusions• Iron ore textural information is important for prediction of downstream

processes.

• Textural identification in Mineral4 allows identification of different morphologies of the same mineral in iron ore and sinter.

• Textural identification allows differentiation of IMDC and RMDC in cokes and comprehensive characterisation of coke structure.

• Unique features and capabilities created in CSIRO’s Mineral4/Recognition4 software allow comprehensive and, in some cases, unique characterisation of important features relevant to ironmaking researchers and industry plant operators.


Thank you

Eugene Donskoi - Project Leader, Process Modelling - Prediction of Downstream Processing Performance

Phone: +61 7 3327 4158

Email: [email protected]

Carbon Steel Futures group/Resources, Community and Environment Program/CSIRO Mineral Resources

Optical Image Analysis Advantages and Drawbacks (compared with SEM methods)

ADVANTAGES

• Fast and cheap

• High resolution

• Reliable identification of minerals with close chemical compositions (like hematite, hydrohematite, magnetite)

• Porosity identification

DRAWBACKS

• No simultaneous chemical composition

• Possible misidentifications

• Problems with identification of non-opaque minerals

within epoxy blocks

COMMON PROBLEMS

• Representability of the studied sample

• Stereological error

• Touching particles or particle sections separated by cracks or pores.

Donskoi E., Manuel J. R., Austin P., Poliakov A., Peterson M. J., and Hapugoda S. Comparative Study of Iron Ore Characterisation

by Optical Image Analysis and QEMSCANTM”, Iron Ore 2011, Perth, WA, Australia, pp 213-222


Average presence of iron ore minerals in different layers of thin 1mm settled sample in epoxy polished block.

0

10

20

30

40

50

60

Hem

atite

Hydro

-H

em

atite

Vitr

eous

Goeth

ite

Ochre

ous

Goeth

ite

Kaolin

ite

Quart

z

Pore

s

% in

blo

ck l

ayer

Mineral

Layers Top to Bottom-150+106 µm fraction

Top

Middle

Bottom


Possible sectioning of non-liberated particles during block polishing

a)

b)

132

Side view

Plan view

a) sectioning of three similar two-mineral particles - side view; b) corresponding particle sections: – two particles appear as fully liberated and only the one in the middle is not liberated

Lin, D, Gomes C O, and Finch J A, 1995, Comparison of stereological correction procedures for liberation measurements, Transactions of the Institutionof Mining and Metallurgy, 104:C155-C161.

•Dependence from texture.•Iron ore much more complex


Unique feature - Automatic identification of non-opaque minerals based on relief

(a)

(b)

(c)

(a) Ore photomicrograph, (b) Image enhanced for quartz identification, (c) Coloured mineral map (Q = Quartz)

Q

Q

Q

Q


Applications of “Mineral4/Recognition4”OIA of iron ore sinter

Determination of coal maceral composition

OIA of manganese sinter

Manganese ore analysis


dr eugene donskoi - project leader, process …m-n.marketing/downloads/conferences/ecic2016...12...

Documents