1. Predicting Iron Ore Sinter Strength Using X-ray Diffraction Analysis
Nathan A.S. Webster1, Mark I. Pownceby1, Natalie Ware2 and Rachel Pattel1
1 CSIRO Mineral Resources, Private Bag 10, Clayton South, VIC, 3169, Australia 2 CSIRO Mineral Resources, PO Box 883, Kenmore, QLD, 4069, Australia
CSIRO Mineral Resources
Here, a preliminary assessment is made of the potential for partial least squares regression (PLSR) analysis to be used as a method for prediction of sinter Tumble Index (TI), a measure of sinter strength, from powder XRD data.
Background
Iron ore sinter, a composite material comprised of iron ore fines bonded by a matrix of complex Ca-rich ferrite phases (Fig. 1[1]), is a major component of blast furnace feed material.
Partial least squares regression (PLSR) analysis has been shown by König et al. to enable prediction of sinter basicity (CaO:SiO2 ratio) and Fe2+ content – the latter in particular being a key marker used for process control – from powder XRD data without the need for rigorous mineralogical analysis (e.g. time consuming Rietveld-based quantitative phase analysis, QPA) by the operator.[2,3] As well, the traditional wet chemical method for Fe2+ determination takes a few hours to complete, and more rapid feedback for the purpose of process control is desirable.
High reducibility, high mechanical strength and low reduction degradation are direct indicators of sinter quality; the ability to determine sinter strength, for example, rapidly through XRD (without the need for time consuming physical property testing) could be a powerful capability for industry.
Tumble Index (TI) is the industry standard method for assessing iron ore sinter strength, and here a preliminary assessment is made of the potential for the determination of iron ore sinter strength directly from XRD data.
Results
Figure 2 summarises the outcomes of the PLSR analysis. Trends in measured TI variation are followed by the calculated TI values. This is despite significant variations in mineralogy even for samples with similar measured TI values (Figure 2c). a
CaCO3 flux a b b
Ore
~1300°C
Pore
Coke
Ore
SFCA and SFCA-I phases, other Ca-rich ferrites, calcium silicates, glass, Fe3O4/Fe2O3
Ultra-fine coating b
Figure 1: Schematic showing a) a typical mixture of iron ore fines, flux and coke; and b) typical iron ore sinter product.[1] b c
Methods
Laboratory-based compact sinter tests were carried out using the method described by Clout and Manuel.[4] The -1mm size fraction of five natural iron ores were fluxed to 2.4 basicity using limestone, then compacted into Ni pots using a press to form 4 g tablets. Samples were heated in a tube furnace in a pO2 = 5  10-3 atm under a standard heating profile. Fired compacts were then tumbled in a modified bond abrasion tester for 8 min, sieved and the relative compact TI recorded as the percentage of material retained above 2 mm.
Tumbled compact samples were then crushed using a percussion mortar to < 500 μm, and micronized in ethanol for 4 min g-1. Powder XRD data were collected over the range 5° ≤ 2θ ≤ 80° on micronized samples using a PANalytical MPD fitted with Co tube, post-diffraction graphite monochromator and an X’Celerator detector used in 1D mode (active length of 2.122° 2θ).
PLSR analysis was carried out using PANalytical HighScore Plus V4.1. The calibration model was “trained” using 32 samples/datasets, and tested on 22 different samples/datasets. Background subtraction was applied to account for background variation with Fe content.
Figure 2: (a) and (b) Plots of calculated vs measured tumble index (TI) values for a series of compact sinter test samples. In (b) R2 = (c) XRD patterns collected for sinter samples with TI = 80.5 (lower) and 80.3 (upper).
Conclusion
PLSR analysis of XRD data shows promise as a method for rapid determination of iron ore sinter strength and, therefore, process control. Further testing is required on a greater number of compact sinter test samples and, ideally, pot-grate sinter samples to further assess the robustness of the method, and to establish how rapid the XRD data collection may be (i.e. how small a 2θ range may be used? What level of counting statistics are required?).
FOR FURTHER INFORMATION
REFERENCES
[1] M. Sasaki, Y. Hida, Tetsu-To-Hagane, 1982, 68,
[2] U. König et al., Powder Diffraction, 2014, 29, S78-83.
[3] U. König, N. Norberg, Iron Ore 2015 Proceedings,
[4] J.M.F. Clout, J.R. Manuel, Powder Technology, 2003, 130,
Dr Nathan Webster e p w