Soft-Sensor Model of Mine Water Inrush Sources based on PCA-CRHJ Network

Abstract

The soft-sensor model for identifying the source of mine water inrush and its performance are studied experimentally. In this novel framework, principal component analysis (PCA) has been used to decompose the original water quality historical data into several sub principal component data which can effectively extract the characteristics of multivariate time series. After that, cycle reservoir with hierarchical jumps (CRHJ) model is established by training and verifying the principal component data. Through the simulation and verification of the historical water inrush data of two measured mines, the experimental results show that the PCA-CRHJ model shows the best performance compared with echo state network (ESN) and cycle reservoir with jumps (CRJ) models generally. The hybrid model, with accurate and stable results, is highly effective for mine water inrush source discrimination and can flexibly be applied in other mine regions.