Improvement of Sampling for Moisture Content Analysis in Iron Ore Pellet Feed – Variographic Characterisation and On-Line IR-Analysis

Accurate information of physical, chemical and metallurgical quality characteristics such as moisture content, iron grade and particle size are crucial for successful process control and for producing products within customer specifications. Correct moisture content of pellet feed is an important variable for successful balling of iron ore pellets and to minimize power consumption in the sintering process. Control of the filtering process in the pelletizing plant is therefore essential for both productivity and high product quality. The current measurement systems for moisture content at LKAB are based on two types of manual grab sampling, subsequently analyzed with rapid moisture determination methods. Previous studies of the measurement systems have indicated high sampling variability resulting in lowered possibilities for successful process control. The objective of the present study is to evaluate an alternative filter sampling approach and testing of on-line Infra-Red analysis (IR-analysis). On-line IR-analysis has the advantage of eliminating physical sample extraction and allowing for continuous monitoring of moisture content. However, this approach is sensitive to the surrounding environment, making careful calibration and control crucial for the quality of measurement results. Furthermore, as IR-sensors only target the surface of the conveyed pellet feed, several incorrect sampling errors need to be considered and counteracted to eliminate sampling bias and reduce the remaining measurement variability. Variographic characterization was applied to the two current grab sampling approaches of combined filtered material, to the alternative filter sampling method, and to the On-line IR-analysis. The results indicate that both the sampling of separate filters (using composite sampling and ISO-standardized moisture analysis), as well as a calibrated and well- maintained IR-sensor can improve the relative variability significantly, by lowering the total measurement system variance from 60% to 25%.