By kernel principal component analysis(KPCA)

a kernel principal model was built to extract the system's non-linear redundant information

and then reconstruct the data in the input space.The residual error produced by the reconstructed data and original sample data was detected by the sequential probability ratio text(SPRT) to timely notice the fault symptoms

and to identify the fault causes by a comprehensive analysis on the system.The diagnosis simulation for the condenser system of a 600 MW power plant was conducted.The results showed that

this KPCA method can accurately identify the fault and build fault diagnosis model with a certain accuracy and practibility.