Coal Preparation » Fine Coal
This project was based on adapting the Particle Profiler system for ultrafine coal streams such as flotation circuits and thickener feed. The current Profiler system captured images of particles (0.1 - 2mm) settling in water and tracks them to measure the average size and speed of each particle. From this information, the particle density is calculated using a Reynolds Number - Drag coefficient calibration. This project extended the system to also measure shape factors and greyscale/colour values; with a goal of being able to identify different mineral phases (if possible).
The hardware was expanded to a microscope system and a custom-built optical cell to match the microscope optical requirements. The microscope provided magnifications x10-220 and pixel resolutions from 14um/px down to 0.7um/px. However, in practice it was found that 2.7-5um/px provided a reasonable combination of attributes. A lighting system was also added to allow frontlit particles to be evaluated for colour (previously only backlit silhouettes were measured).
The system was tested during development on chalco-pyrite tailings (for greyscale measurement) and a selected sink 1.80 sample (for mineral colour responses). The expanded output files included a Summary Histogram and a particle data file containing individual greyscale histogram data. This was later developed to include the option to choose greyscale or a colour channel from Red, Green or Blue. It was found that the binary chalco-pyrite / silica system produced a bi-modal distribution of greyscale (outputted as a Summary Histogram)-largely revolving around composite particles. The individual particle histograms were found to correlate with the visual “grade” of particle based on its mineral content. The same concept was applied to a Sink 1.80 sample that was comprised of kaolinite, quartz, siderite and hematite. In this case, the RGB histograms were used to study the colour metrics of the analysed mineral particles and used as a basis for modelling optically different minerals. Through this method it was possible to replicate up to 84% of the Summary histogram. It must be stated that the mineral phases that were optically detected and quantified did not match the previously measured mineral concentrations. However, it did detect elevated levels of three dominant minerals which aligned with the expected kaolinite, silica and siderite. As a first trial of the colour analysis, this was considered to be a good response though clearly more refinement is needed.
The Profiler system was then tested on CHPP samples of a flotation circuit (feed/product/reject) and thickener feed samples from a different site which corresponded to “Good” dewatering properties and “Bad” dewatering properties. The specific thickener operational issues were not identified, but rather based on operational experience of different coal seams.
It was found that the CHPP samples each had varying issues with agglomeration occurring during analysis with the Profiler system. Using a shape factor (Circularity >0.5), it was estimated that these samples had 22-50% agglomerated particles; which impacted the settling rate and density distribution. The optically measured size distribution also showed larger sizes than those from laser sizing. Overall, it was observed that the flotation samples appeared to be more greatly impacted than the thickener samples. In comparison to float/sink analysis, the flotation product did not reproduce results well in any mode. The flotation reject sample did appear to have similar results to float/sink results; with further improvements expected from calibration.
The thickener feed samples were selected for deeper analysis using greyscale and colour metrics. From a “Dominant Peak Analysis” of particle histograms, it was observed that the sample with “Good” dewatering properties had a relatively simple mineral profile consisting of a single dominant mineral type (83% of peak pixels), with only minor amounts of additional components (2-6% of peak pixels). The sample with “Bad” dewatering properties showed a relatively complex mineral profile consisting of up to 13 peaks of significance. The highest component was ~40%, with other lesser peaks varying 2-15%. Average RGB values were then able to be assigned to these “optically distinct” minerals; which could be used in future work for “tagging”. These rather remarkable results suggested that “Bad” dewatering properties may be due to an inability to flocculate the wide array of mineral properties contained in the sample. Whereas, the sample with “Good” dewatering properties would be easier to optimise thickener operations because of its simple mineralogy.
The Profiler proved to be quite susceptible to agglomerated particles, returning false density measurements that indicate more coal is present. This was seen as a major issue and several control methods were evaluated using a laboratory particle laser sizer to reduce particle agglomeration. It was shown that pump speed and reagent addition could partially negate some of the impacts and that ultrasonics further reduced particle top size. Efforts to emulate the ultrasonics and reagent additions prior to Profiler analysis had some success in reducing particle size distribution, but more focussed work is needed on sample pretreatment.
The technology developed in this project has shown a significant amount of information may be drawn from a settling particle. Even the capability to detect agglomerated particle and provide feedback for sample pre-treatment settings is considered to be highly valuable. The greyscale/colour metrics have been demonstrated to show great potential in terms of indicating particle “grade” (average greyscale/RGB) and mineralogical identification.