Abstract: A chemical process has usually a larger number of measured variables which are often driven by fewer unmeasurable latent variables (or sources).Extracting such latent variables and monitoring them will improve the process-monitoring performance.However, lots of existing process monitoring methods are based on the assumption that measured variables are subjected to independent and identical distribution (iid), which is not always valid in a chemical process.In this paper, an extended multivariate statistical control method based on independent component analysis with temporal structure is proposed to overcome the limitation of the assumption of iid.For investigating the application of this method to a continuous-stirred-tank-reactor process (CSTR),fault-detection performance was evaluated and compared with that of the conventional independent component analysis (CICA).The simulated results demonstrated the advantages (e.g. decreasing the number of false alarm and missed alarm) of ICA with temporal structure over the CICA approach under the assumption of iid, which is still widely used.