Principal component analysis（ PCA） is a typical data dimension reduction multivariate statistical method

which has been used in fault diagnosis. PCA is applied in bridge deflection sensor fault diagnosis. Introduce theory of PCA

and study fault detection method based on PCA and fault diagnosis method based on contribution rate. Calculate squared prediction error（ SPE） and Hotelling T2statistics

when statistics exceeds threshold

judge there are sensor faults in system

and by using SPE contribution figure judge fault source. By simulation

the practicability of PCA in fault diagnosis is proved

but the result also show that PCA is not so sensitive to small fault.