On-line soft sensor based on RPCA and LSSVR for mill load parameters

Accurate on-line measurement of ball mill load (ML) affects production capacity and energy efficiency of the grinding process. An on-line soft sensor method based on recursive principal component analysis (RPCA) and on-line least square support vector regression (LSSVR) was proposed in this paper. At first, spectral principal components (PCs) at low, medium and high frequency bands of the vibration spectrum were extracted through PCA. Then, the extracted feature variables were used to construct LSSVR model. At last, when a new sample was given, the older PCA-LSSVR model was updated by RPCA and online-LSSVR algorithm recursively. Therefore, the innovate integration of the RPCA and online LSSVR makes the online soft sensor for ML parameters soft sensor practical. A case study shows that the proposed approach has higher accuracy and better predictive performance than the other normal approaches.