Recently

independent component analysis(ICA)has attracted wide attention in process monitoring.This article proposed a novel dimension reduction method of ICA for process monitoring

on the basis of minimum mean square error(MSE)

and the mathematical proof was presented.The order of the independent components(ICs)was ranked according to their importance estimated by the MSE

and correct latent variables were selected to improve the performance of process monitoring based on ICA.The proposed dimension reduction method was applied to fault detection in a process simulation and the data of pulverizing system in power plant under fault condition.The results verified that this method could decrease the miss detection rate and enhance the detection reliability.