In order to overcome coupling effect of aero-engine's control loop, a general predictive decoupling controller based on coactive adaptive network-based fuzzy inference system (CANFIS) is designed for aero-engine distributed control system with long uncertain time-delay. The input and output data of engine nonlinear model are used as sample data to train CANFIS. The front parameters are fixed after training, and the latter parameters can be adjusted to make CANFIS approach actual system online. The network is taken as a predictive model, which can be used as a substitute for the feedback amending part of general predictive controller. The simulation results show that the control system has a fine performance of tracking and decoupling capability to different reference signals such as step and ramp, it can also follow the reference track stably when time-delay is changed, which proves that the controller is of insensibility to time-delay and of strong robustness.