Nonlinear model predictive control (NMPC) is a strategy suitable for dealing with highly complex, nonlinear, uncertain, and constrained dynamics involved in aircraft engine control problems. Because of the complexity of the algorithm and the real-time performance of the predictive model, it has thus far been infeasible to implement model predictive control in the realtime control system of aircraft engine. In most nonlinear model predictive control, nonlinear interior point methods (IPM) are used to calculate the optimal solution, which iterate to the optimal solution based on the Jacobian and Hessian matrix. Most nonlinear IPM solver, such as MATLAB fmincon and IPOPT, cannot calculate the Jacobian and Hessian matrix precisely and quickly, instead of using numerical differentiation to calculate the Jacobian matrix and BFGS method to approach to the Hessian matrix. From what has been discussed above, we will 1) improve the real-time performance of predictive model by replacing the time-consuming component level model (CLM) with a neural network model, which is trained based on the data of component level model, 2) precisely calculate the Jacobian and Hessian matrix using automatic differentiation, and propose a group of algorithms to make NMPC strategy quicker, which include making use of the structure of predictive model, and the integrity of weighted sums of Hessian matrix in IPM. Finally, considering input and output constraints, the fast NMPC strategy is compared with normal NMPC. Simulation results with mean time of 19.3% – 27.9% of normal NMPC on different platforms, verify that the fast NMPC proposed can improve the real-time performance during the process of acceleration, deceleration for aircraft engine.