MIMO Soft-sensor Model for Nutrient Contents of Compound Fertilizer

Abstract

In compound fertilizer production, there are several quality variables need to be monitored and controlled simultaneously. Existing instruments and sensors now cannot measure these variables on-line. So soft sensor technique becomes an indispensable method to implement on-line real-time quality control. In this paper, a new method of multi-inputs and multi-outputs (MIMO) soft sensor, which is constructed with improved FasBack neuron-fuzzy system is proposed for these interaction variables. FasBack is very suitable for clustering and non-linear modeling. Levenberg-Marquardt algorithm is used to train some parameters in the model, in contrast to the primary FasBack, the improved FasBack possesses quicker and better convergence. Based on practical process data, the proposed improved FasBack is applied to build compound fertilizer soft-sensor model. Simulation results indicate that the proposed model is precise and efficient and it is possible to realize the on-line quality control for compound fertilizer