Monitoring, control and real-time optimisation of complex metallurgical plants require that strategic variables be directly measured or inferred from other process variable measurements. The sensors to be implemented in the plant must be carefully selected in such a way that all the strategic variables can be estimated with sufficient precision and at reasonable cost. The objective of the study is to develop and illustrate methods which allow an optimal selection and placement of the sensors into the plant. The optimality is defined with respect to process observability (minimal or redundant), cost of investment and maintenance of the sensors, and expected improvement of the process performance. The method is based on a model of mass and energy conservation around the plant flowsheet. The model, combined to cost and accuracy data for the potential sensors to be implemented, is used to determine the best compromise between the instrumentation costs and the anticipated benefits. The design technique is illustrated for a generic mineral separation plant. (C) 2003 Elsevier Science B.V. All rights reserved.