The work aims to design a cutting control system based on STM32 to improve the cutting precision and speed of corrugated board shears. The working principle of corrugated paper cutting was introduced. A control system was built with STM32 controller as the core and the hardware structure diagram was given. Due to the nonlinear characteristics of the control process, it was difficult for the conventional PID controller to achieve the ideal control effect. Based on radial basis function (RBF) neural network, an equal length cut-off controller of corrugated board was designed. At the same time, the stochastic gradient descent method was used to adjust the network weight in real time and a learning rate attenuation method was given. The feasibility and the effectiveness of the method were verified by simulation and experiment. The experimental results showed that the REF neural network control system had good speed following characteristics and anti-interference performance. Corrugated board length fluctuation was not large, the maximum average deviation was only 0.2 mm. Corrugated board cut was level, and there was no folding, pulling and other conditions. The system runs stably. The control system can improve the cutting precision of corrugated board and meet the requirements of related industries.