Dimension reduction method of independent component analysis for process monitoring based on minimum mean square error

► We give a mathematical proof of dimension reduction in ICA based on minimum mean square error. ► The proposed dimension reduction approach considers the contribution of ICs and retains more information of process. ► Combining with KDE approach, a process monitoring method was proposed to monitor the variations of process in two spaces. ► The accuracy of proposed dimension reduction method was verified by the experiment and the performance of proposed monitoring method was demonstrated by the TE process. ► A ball mill pulverizing process was introduced and the proposed monitoring method was applied to the real industrial process.