This paper describes a real-time on-line process monitoring and fault isolation approach using PCA (Principal Component Analysis). It also presents the software implementation architecture using an OPC (OLE for Process Control) compliant framework, which enables a seamless integration with the real plant and DCS. The proposed approach and architecture are implemented to monitor a refinery process simulation that produces cyclohexane using benzene and hydrogen. The result shows that both sensor faults and process faults can be detected on-line and the dominating process variables may be isolated. ©2005 IEEE.