A　robust　fault　detection　method　is　proposed　for　nonlinear　dynamical　systems　with　unknown　but　bounded　noises.　The　Takagi-Sugeno　(T-S)　fuzzy　neural　network　is　used　to　build　a　model　of　the　nonlinear　dynamical　system　when　the　system　is　fault-free,　taking　into　account　that　it　is　a　universal　approximator.　The　input　space　is　partitioned　by　means　of　a　fuzzy　clustering　algorithm　based　on　the　input　and　output　data　of　the　system.　Supposing　that　the　system　noise　and　approximation　error　are　unknown　but　bounded,　the　consequence　parameters　of　the　T-S　fuzzy　neural　network　are　determined　using　a　linear-in-parameter　set　membership　estimation　algorithm.　An　interval　guaranteed　to　contain　the　actual　output　of　the　fault-free　system　is　obtained　by　propagating　the　effect　of　model　uncertainty　to　the　model　output.　An　occurrence　of　the　fault　is　signaled　when　the　measured　output　crosses　the　computed　interval.　Simulation　results　show　the　effectiveness　of　the　proposed　method.　©　2013　AACC　American　Automatic　Control　Council.