Great effect of three way catalytic convertor (TWC) performance on oxygen sensor output voltage has made the sensor (located after catalyst) as the main signal in almost all today’s TWC monitoring algorithms. In this paper output voltage of nonlinear oxygen sensor is estimated using a nonlinear autoregressive with exogenous inputs (NARX) model. The estimation uses ECU calculated exhaust gas flow and air fuel ratio from upstream lambda sensor and does not need any unconventional sensor measurement. Combined with simple structure applied to the NARX model, an algorithm applicable to real-time computations on vehicle is designed. Estimated voltage is compared to real measurement from a car during new European driving cycle (NEDC), showing good performance of the estimation algorithm. Moreover, by tests results it is shown that this model can monitor the effect of catalysts aging.