Electric smelting furnaces typically consist of a refractory shell incorporating strategically placed water-cooled copper components for maintaining furnace integrity. Over time, both the refractory lining and water-cooled copper components can undergo deterioration and wear, resulting in potential failure in the form of a furnace run-out, where furnace matte and/or slag leaves the furnace in an uncontrolled manner. With smelting occurring at extreme temperatures, large amounts of conventional thermocouples and new fibre optic technology (for copper) and resistance temperature detectors (for cooling water) are used to continuously measure the temperature of the furnace crucible lining with the aim of ensuring safe operation of the furnace. Measurements from these numerous sensors are typically recorded at high frequency, with data overload of operators and metallurgical staff alike commonly occurring.Monitoring of the furnace temperature measurements, which is pivotal to furnace integrity monitoring, has traditionally been accomplished through the use of alarm and trip limits set on individual temperature measurements of the copper coolers, their cooling water and refractory, with limits typically defined based on design criteria and not on current operating conditions. Due to the changes in furnace operating conditions and the sheer number of temperature measurements available on a furnace this often proves to be very ineffective, except when large upset conditions occur; and the warning is then often too late.When Mortimer Smelter, a platinum group metal (PGM) concentrate smelting operation located on the western limb of the Bushveld Complex, changed its cooler design to novel graphite-protected shallow-cooled composite copper coolers with graphite-protected deepcooled copper lintel coolers, the furnace was at some risk due to the new design. In order to confirm that maintenance could take place as planned, an alternative temperature monitoring strategy was required. Multivariate statistical data-based techniques, proven to be effective in circumventing the previously mentioned monitoring inadequacies through data compression, dimensionality reduction, and the handling of noise and correlation (Venkatasubramanian et al., 2003), were considered as the foundation of such a strategy. In this paper we discuss the development of and operating experience from a multivariate statistical data-based system for monitoring furnace temperature measurements. Further application of the system so developed, informing a decision to delay a matte endwall rebuild on the Polokwane Smelter, is also described.Furnace integrity monitoring using principal component analysis: an industrial case study by J.W.D. Groenewald*, L.R. Nelson*, R.J. Hundermark † , K. Phage*, R.L. Sakaran*, Q. van Rooyen*, and A. Cizek* Furnace temperature monitoring, the cornerstone of furnace integrity monitoring, has traditionally been accomplished using alarm and trip limits set on individual temperature measurements of the copper coolers and refractory, ...

The Polokwane Smelter operates a single sixin-line electric furnace, nominally rated at 68 MW. The smelter treats mainly Platreef, and UG2 concentrates from the eastern and northern limbs of the Bushveld Complex and smelts them to form slag, matte, and gas phases. The furnace operates with three matte tap-holes on the northern wall of the furnace and matte is tapped periodically from the furnace, with each tap being 25 to 35 minutes in duration. Two tap-holes are utilized per day, with one resting. A tap-hole is not utilized for more than two days in succession to ensure that it is not over-utilized and also to allow for any tapping channel refractory preventative maintenance.The matte tap-holes consist of copper cooling elements within which are refractories that contain the tapping channel. The hot face of the water-cooled copper tap-block is separated from the molten material by refractories. It is within, or on the surface of, the copper tap-blocks that the fibre optic technology was embedded. Due to the higher superheats of PGM matte compared to most other types of matte (Shaw et al., 2012;Nolet, 2014), the matte tap-holes are subjected to extreme process conditions which may render the furnace prone to failures. The wearing of these refractories introduces the risk of explosions should any matte come into contact with the copper block and then the cooling water channel.Fibre optic technology was installed strategically at the hot face and within the water-cooled copper tap-blocks and key watercooled copper coolers, with the intent to provide a more detailed and timely detection of temperature rises in the copper, refractory, and associated freeze-lining components on the hot face of the matte endwall, copper coolers, and the tap-block. Prior to the introduction of fibre optic technology, the methods of endwall condition monitoring were limited to the measurement of copper temperatures of the tap-block and cooper coolers by thermocouples installed at specific points, and the use of resistance temperature detectors (RTDs) that measure the change in cooling water temperature. The selection of fibre optics was made to provide a significantly higher density of sensors with an enhanced spatial resolution in comparison to the single conventional temperature measurements.Two types of fibre optic technology were installed in different copper tap-blocks and coolers, namely type A technology utilizing Bragg gratings and type B utilizing the Raman effect. The Bragg gratings reflect a wavelength of light that shifts in response to variations in Analysis and interpretation of fibre optic temperature data at the Polokwane Smelter by R.L. Sakaran, Q. van Rooyen, P.K. van Manen, and P.P. MukumbeTo help improve the monitoring of the matte tap-holes on the matte endwall to prevent failures, fibre optic temperature systems were installed on the six-in-line electric furnace at the Polokwane Smelter. This was done with the intent to provide more detailed and timely detection regarding the change in condition of the copper coole...