PGM Conference

J Heyns, A Bogaers, Willem Roos and Johan Zietsman

Engineers working in the pyrometallurgical industry are increasingly leveraging computational modelling to gain better insight into high-temperature processes where limited experimental data is available. In the past, the higher cost associated with, for example, computational fluid dynamics implied that engineers could run a small number of simulations and, hopefully, improve their understanding of the system. Unfortunately, models in the pyrometallurgical industry are dependent on a large number of inputs that may be subject to substantial variations. In an ideal world, changes in the input parameters would result in a smooth, linear change in the model outputs, allowing one to consider a base case and by exploring a few simulations around the reference point gain a good understanding of the system behaviour. The question is then, in a world where sharp or non-linear changes are the norm instead of the exception, do computational models provide insight or do they merely result in a false sense of security? This paper investigates the potential of employing uncertainty quantification to better quantify the risks associated with the process. For demonstration purposes, the electromagnetics of a circular alternating current platinum group metal smelting furnace are considered, and the influence of variations in the design and operating parameters as well as material properties on the furnace resistance are computed. Furnace resistance is an important design metric, as an engineer needs to ensure there is sufficient energy available for the process, but that it does not exceed the upper limit and affect operational stability. In this paper, we show how the probabilistic philosophy of uncertainty quantification can help us gain insight and advance the pyrometallurgical industry. An important requirement of this approach is, however, to have a rich data set. For this reason, we discuss how recent technological advances have lowered compute costs and improved open-source software, providing engineers with a cost-effective solution to conduct large scale data mining using computational models.

T Makgoale, N Sweeten, H Weitz, J Zietsman

The platinum group metal (PGM) industry in South Africa is dealing with increasingly challenging ore bodies that are often not well suited to traditional matte-smelting furnaces. Engineers can use models to assess the impact of new concentrate blends on the furnace so that operators can prepare for the impacts or modify blending practices to achieve a desired goal. Predictive models based on thermochemistry are state-of-the-art for pyrometallurgical processes; however, it is important to understand the accuracy and limitations of such models before interpreting the results. For this reason, we compared model results to data available from literature to determine which databases are the most accurate and to what degree. FactSage 8.2 introduced the new FTsulf database for describing sulphide systems. In this paper we compare experimental results from literature to FTsulf, FTmisc, FSstel, and FToxid, all of which can also describe liquid matte solutions. FTsulf was found to be a significant improvement on the older databases. We believe that thermochemical modelling has a key role to play in the development of new technologies to ensure the continued success of the South African PGM industry.