Year

2009

Degree Name

Doctor of Philosophy

Department

School of Engineering Physics, Faculty of Engineering

Abstract

Mineral processing and metal production techniques depend on the mineralogy of the feedstock fed into the processing plant. The ability to perform on-stream mineralogical characterisation of feedstock materials, or to monitor intermediate, product and waste streams would allow better process control and increased efficiency. On-line elemental analysers based on X-ray fluorescence and prompt gamma-ray neutron activation analysis are widely used, but existing mineralogical analysis methods rely on extracting and measuring small samples. This can introduce sampling errors and is time consuming, particularly if the sample must be removed to a laboratory for analysis. These methods are therefore ill-suited to process control applications. This thesis develops a new technique for monitoring the mineralogy of industrial process streams in real-time. The technique, called energy-dispersive X-ray diffraction (EDXRD), is well-suited to the application of on-stream mineralogical analysis of mineral slurries. An EDXRD analyser measures the energy spectrum of X-rays diffracted by a sample material at a fixed angle. This method uses much higher X-ray energies than the conventional X-ray diffraction technique, therefore greater depth penetration and is obtained with less reliance on sample preparation. This results in it being better suited to the application of on-line diffraction measurement. An extension to the EGSnrc Monte Carlo code was developed that enables X-ray diffraction to be modelled. Diffractive scattering from both crystalline and amorphous materials can be modelled, as well as materials containing both crystalline and amorphous components. It was shown that this method can be used to simulate the diffraction spectra of samples containing mixtures of different materials. The purpose for developing this extended code was to use it to aid in the design and development of EDXRD analysers. A laboratory prototype EDXRD analyser was designed and developed. The instrument was designed to measure a wide range of commercially important minerals in both dry powder and slurry form. Monte Carlo modelling was used extensively to optimise the design of the instrument and predict its performance. Comparisons between Monte Carlo modelled and experimental spectra obtained with the instrument showed good agreement, validating the method developed to simulate diffractive scattering. Quantitative mineral phase analysis was performed on two suites of materials in order to investigate the accuracy with which the mineral components could be determined with the EDXRD analyser. The first suite consisted of twenty samples, each containing six commercially important minerals. Regression analysis performed on the spectra showed that all six components could be quantified with accuracies of better that 1 wt%. The second suite contained seven minerals found in potash slurry. Good measurement accuracies were obtained for most of the components. The spectra of the samples in both suites were also modelling using Monte Carlo simulation in order to determine if simulated spectra can be used to predict the measurement accuracy of an EDXRD analyser. It was found that the analysis accuracies obtained from the modelled spectra agreed well with the experimental results. This showed that the measurement accuracy of an EDXRD analyser can be predicted using Monte Carlo simulation. A system for optimising the design of an EDXRD analyser was developed. The system uses performance data derived from Monte Carlo modelling for 1.7 million instrument designs and a computer code to find the optimal analyser design to measure a material of interest. The advantage of the system was demonstrated by redesigning the prototype analyser using the optimisation code. It was shown that the optimised instrument delivers significantly better performance than the prototype analyser. Finally, the methods and knowledge developed in the thesis were put to use in the design of a potash slurry analyser. The analyser was designed to measure potash slurry on-line for the purpose of process control. The design of the analyser was optimised using the optimisation code. The analysis accuracy of the analyser was predicted using Monte Carlo modelling, which showed that all mineral components of the slurry could be quantified with accuracies of better than 0.7 wt%. This result demonstrated that EDXRD has the potential to be a viable tool for the on-line analysis mineral slurries.

02Whole.pdf (6005 kB)

Share

COinS