An Adaptive Soft Sensor for Mill Load Measurement Based on PCA and FasArt Neural Fuzzy Networks

Precise load measurement is important for the supervision of the pulverizing process in thermal power plant. This paper presents an adaptive soft sensor based on PCA and FasArt neural networks to achieve this purpose. PCA is firstly used to compress the input secondary variables and the dimension is reduced from 9 to 3 with little loss of information. Then FasArt model derive the knowledge from the training data and construct the relationships between the input secondary variables and target variable automatically. Experimental results show that the proposed model achieve a high accuracy. Moreover, the model has potential advantage of incremental learning capability.