In this article, we propose a digital twin model (DTM) based on the nonlinear autoregressive model with exogenous (NARX) inputs model network to predict engine gas path parameters accurately. The DTM is combined by a model-driven model (MDM) and a data-driven model (DDM). To allocate the function of MDM and DDM, a pretreating fusion method is proposed for the first time, which is divided into three parts. First, all parameters are predicted by the MDM. Second, for the parameters with bad predictive effects, the DDM is employed to optimize them. Third, the parameters with good predictive effects and those optimized by DDM are fused to generate the DTM. The DDM is built by a two-stage NARX. Particularly, a NARX with a gate recurrent unit attention mapping function is used to improve the accuracy of the predicted parameters. The experimental results show that the maximum prediction error of the DTM is less than 5%. This implies that the fused DTM guarantees the prediction accuracy of each gas path parameter in the case of performance degradation.