Research of Quality Prediction Based on Extreme Learning Machine

Aiming at the problems of poor stability and generalization performance in quality prediction based on extreme learning machine (ELM), this paper presents an improved method of ELM. It is named as the DAE-P-ELM algorithm, which integrates denoising autoencoder (DAE) with principal component analysis (PCA). First, in order to reflect the characteristics and intrinsic relationship of the modeling data as much as possible, DAE technology is introduced to reconstruct the input data. Therefore, output weight sufficiently containing the input data information is obtained, which is used as input weight of the ELM. Then, the PCA technology is used to reduce the dimension of the hidden layer output matrix to avoid the multicollinear problem in calculation of output weight matrix, which solves the problem of poor stability of the model due to too many hidden layer nodes. Finally, the method is applied to the Tennessee Eastman (TE) process. The simulation results show that the content of components G and H predicted by this method is basically consistent with the real value, which proves that the proposed method has a good prediction effect.