This paper presents a new online statistical monitoring based on dynamic independent component analysis (DICA) to detect the Tennessee Eastman challenge process faults. The proposed method employs dynamic feature extraction approach to capture most of the inherent dynamic fault information. This leads to an efficient fault detection with superior performance compared to independent component analysis (ICA) approach in both detection rate and number of false alarms. A new statistic measure has been introduced to enhance the monitoring capabilities of ICA and DICA. An approach based on cumulative percent variance (CPV) has been incorporated to mechanize the selection of required number of independent components in both ICA and DICA online monitoring methods. To choose the best time-lag order for each fault dynamic model in the DICA augmented data matrix, a multivariate auto regressive exogenous (ARX) model structure has been adopted by validating the minimum Akaike’s information criterion (AIC) index. An online procedure based on a multi-class support vector machine (SVM) with Gaussian kernel function, being set by sub-optimal width parameters, is employed to classify and isolate each fault. The SVM uses one against all (OAA) algorithm for fault classification and sequential minimization optimization (SMO) to solve the classification problem. Performances of the developed process monitoring methods (ICA-SVM, DICASVM) are evaluated on the Tennessee Eastman challenge process (TE).