Multiple Bilinear Models Based Soft-Sensor for Rare Earth Cascade Extraction Processes

In rare earth cascade extraction processes, element component content (ECC) is an important quality measure to assess the control effect. In this paper, a multiple models-based soft-sensor is developed to predict ECC in order to solve the difficulty with on-line measurement. First, the subtraction clustering algorithm is used to locate the operating points and calculate the model number, then around the multiple operating points a set of multiple bilinear models are established. At every sample instant, the parameters of the multiple models are identified by the least square algorithm, and an optimal model is determined according a switching law. The application results show that the proposed soft-sensor is effective and the prediction accuracy is relatively high by comparing with data collected method from industries