In the present work, an improved statistical analysis, modeling and monitoring strategy is proposed for multimode processes with between-mode transitions. The subject of analysis is multi-source measurement data, with each source of data corresponding to one operation mode. The basic assumption is that the underlying correlations among the different modes are similar to a certain extent and a multimode common community can thus be enclosed by some common bases immune to the mode changes. By making an adequate projection of measurement space, the mode-common subspace is separated and can be represented by a robust statistical model. The remaining mode-specific subspace would be more specific to different operation modes. Moreover, a between-mode transition identification algorithm is designed, which can distinguish the normal transition behaviors from those abnormal disturbances. The proposed method provides a detailed insight into the inherent nature of multimode processes from both inter-mode and inner-mode viewpoints. More process information is captured which enhances one's understanding of the multimode problem. Its feasibility and performance are illustrated with a practical case. (C) 2010 Elsevier Ltd. All rights reserved.