Abstract: Process monitoring based on multivariate statistical projection analysis (MSPA) has successfully been applied to chemical processes. However, measured data are often dealt with as stationary information in industrial processes, which is not a case in fact. In this study, a new process monitoring method based on blind signal extraction and k-nearest classifier is presented, which dose not demand stationary measured data.In order to verify the effectiveness and feasibility of this method, the new process monitoring method was applied to a simple AR(1) process and a double-effect evaporator. The simulation results showed that the process monitoring method presented in this paper had fewer false alarms and missing alarms than that based on independent component analysis (ICA) and fewer false alarms than that based on multivariate statistical process control (MSPC).Therefore, the process monitoring method presented in this paper is more effective and better than conventional process monitoring methods.