In order to find the lowest fuel consumption rate in steady-state operation of a variable-cycle engine, and thus increase the flight range as well as the flight radius of the aircraft. In this paper, we design a multivariable control law for variable-cycle engines and adopt a deep reinforcement learning algorithm in machine learning to optimal the control law for variable-cycle engines in real time online, which results in a lower fuel consumption rate compared to the engine fuel consumption rate under conventional optimal algorithms.