The invention relates to a fixed-time predetermined performance recurrent neural network mechanical arm control method. The method comprises the steps that (1) a mathematical model of a mechanical armdriven by a direct-current motor is established, and an actuator model with an unknown nonlinear dead zone is established; (2) system reference output is determined, and a performance function, whicha tracking error needs to meet, is designed; (3) a fixed-time predetermined performance recurrent neural network controller, a neural network weight update law and a fixed-time differentiator are designed so that system output can track a reference output trajectory in fixed time, and meanwhile the system tracking error is limited within a performance boundary range designated in advance; and (4)stability analysis is performed on a control system, and controller parameters are determined according to the stability analysis result. Through the method, fixed-time predetermined performance trajectory tracking can be realized, so that tracking time is shortened, control precision is improved, and transient performance and steady performance of the system in the control process are guaranteed.