A multivariable neural network adaptive control is provided for a turbofan engine.The nonlinear system is compensated dynamically by a 3-layer neural network.An on-line weight-tuning algorithm with simplified Hebbian rule and e-modification is worked out to guarantee tracking performance as well as bounded weights without off-line training.The precise model of the engine is not necessary for design of control system.The multivariable control system has satisfactory performance for the full flight envelope with good adaptation and robustness. Simulation results show satisfactory dynamic and static performance within the whole flight envelope and good effectiveness of reducing interaction in the multivariable system with significant coupling.