Real-time reconciliation of mineral processing plant data using bilinear material balance equations coupled to empirical dynamic models

On-line material balance in mineral processing plants are extremely useful to upgrade the quality of the information, delivered by the sensors, which is used for manual or automatic control and real-time optimization strategies. A method for on-line reconciliation of flowrates and compositions of complex mineral processing plants is proposed. Measured process variables are filtered and unmeasured ones estimated by minimizing a criterion which contains deviations between filtered and measured values and deviations to process model equations. The model consists of two parts: steady-state mass conservation equations around the flowsheet nodes and dynamic transfer functions and pure delays representing the stream dynamic behaviour. The reconciliation method is tested on a simulated dynamic flotation plant. In comparison to on-line steady-state reconciliation methods, the dynamic models significantly improve the quality of the estimated process variables, because they avoid to have to compensate for process dynamics by averaging techniques which distort the signal phases and amplitudes.