An　innovative　method　based　on　support　vector　machines　is　presented　to　diagnose　the　fault　of　analog　circuit.　Firstly,　in　order　to　get　enough　fault　samples,　the　circuit　program　is　compiled　in　MATLAB　software　to　obtain　expressions　of　output　signals.　Secondly,　fault　samples　are　sent　into　Support　Vector　Machines　to　train　Support　Vector　Machines.　Thirdly,　the　test　samples　are　classified　by　trained　Support　Vector　Machines.　Finally,　an　example　of　analog　circuit　fault　diagnosis　is　provided.　The　result　shows　that　this　method　has　the　advantages　of　simple　algorithm,　high　efficiency,　high　accuracy,　great　capability　in　generalization　and　classification.　©2009　IEEE.