Application of variography to the control of species in material process streams: %Fe in an iron ore product - Synopsis

The use of a time series variogram as an aid in controlling the quality of a product was pioneered by Pierre M. Gy and is described. The ability to analyse and interpret covariance relationships time series data is possible using the variogram. The concept of stream heterogeneity allows sources of variability in process streams to be identified and isolated. Principally this is distributional heterogeneity that is due to a short range effect of segregation in the process stream related to size or composition distribution of the material on a local scale and due to a long-range change in the size and grade of the material stream. The variogram is simply a tool for representing these sources of variability in a way that is explicit, allowing identification of the source and providing important insights to temporal continuity between samples. This paper describes the chronologically related large-scale variability in an iron ore process stream (%Fe), and illustrates the control and mitigation of variability at various stages within the process using a variogram. Three main sources of variability are identified including the random variability V[0], the process variability V[1], and the variability due to cyclical phenomena V[cyclic], arising from human routine or mechanical defects. These sources of variability can be represented on the control chart that provides the plant superintendent with insights as to the sampling capability of the systems in place. Maintaining levels of variability within customer specification limits requires that production systems be controlled without overcorrecting for the trends in changing quality. This can be done only by representing the limits of control relative to the limits of specification on a control chart. Where the specification limits encroach on the control limits, one or the other, or both, may need to be changed; specification limits that constrain the control limits lead to reactive decision making that is expensive and stressful for plant operators.