In this paper

a multi-model soft sensor method based on Dempster-Shafer theory (DS) and Gaussian process regression (GPR) was proposed. Firstly

GPR was used to build the sub-models of the proposed soft sensor after clustering training dataset. Secondly

the initial weightings were designed based on membership functions and output posteriori probabilities of GPR based sub-models

respectively. And the initial weightings were fused using the combination rule of DS. Finally

the weighted sum of sub-models with the fused weightings was used to output predictive means and uncertainty. The proposed method was validated on simulation data of a penicillin fermentation process and industrial data of an erythromycin fermentation process. For comparisons

single model-based soft sensor and traditional multi-model soft sensor were also studied. Simulations showed that the proposed method had better predictive accuracy and lower predictive uncertainty.