Aiming at the difficulty of effective control of rotary inverted pendulum system, based on neural network adaptive control method and backstepping iterative control technology, we proposed a neural network finite time adaptive control algorithm. The mathematical model of the pendulum system with rotational motion is established by considering the uncertainties and external disturbances inherent in the system, which affect its behavior due to uncertain factors, the system model has the very big uncertainty, using the neural networks to approximate the system, and then finite time constraint control algorithm is put forward, make the inverted pendulum system stabilization in the limited time. The designed finite time constraint control algorithm can ensure that the inverted pendulum state does not leave the constraint range during the controlled process, and at the same time stabilize in finite time, and give the convergence time. The algorithm's convergence was validated through Lyapunov stability theory, and its efficacy was proven via simulation experiments, ensuring the stable operation of the rotary pendulum system.