Multi-model soft sensor based on Dempster-Shafer rule

There are disadvantages in traditional model methods for the soft sensor, such as low predictive accuracy, poor fusion ability and weak adaptability. In this paper, a multi-model soft sensor method is proposed based on Dempster-Shafer (D-S) rule. Firstly, the affinity propagation (AP) clustering method and the least squares support vector machine (LS-SVM) are used to establish multiple sub-models. Then, the multi-model output of the soft sensor is obtained through the fusion of the sub-models based on the weighting factor calculated by using D-S rules to improve the model prediction ability and fusion ability. The proposed method is used to build the soft sensor model of a nonlinear system and the ester rate. Simulation results and industry application indicate that the proposed method has better predictive performance and higher accuracy in comparison with the traditional soft sensor.