The Hammerstein model structure identification method of an exhaust gas oxygen（EGO） sensor is studied.Combination of hyperbolic tangent with polynomial is selected as static non-linear function

and linear dynamic part of the model is chosen respectively as auto-regression with exogenous variable（ARX） model

output error（OE） model and Box-Jenkins（BJ） model.The cross-criterion method is used for parameter estimation and orders selection

and the models are checked by simulation comparison.Results show that final output error（FOE） criterion and final prediction error（FPE） criterion are suitable for choosing order with estimation data

while the former is more reliable than the latter.Both OE model and BJ model with three orders based on prediction error method can be used for modeling linear dynamic part of the Hammerstein model for EGO sensor.