Permanent magnet synchronous motor is widely used in industrial control because of its high efficiency and high dynamic performance, but the traditional direct torque control method is easy to cause torque pulsation and magnetic chain fluctuations, especially in low speed operation. In this paper, the fuzzy tree model is applied to the direct torque control of permanent magnet synchronous motor. By introducing the intelligent control method based on the fusion of fuzzy logic and decision tree, we adjust the switch decision rules, optimize the voltage vector selection, and improve the dynamic response ability and steady-state control accuracy of the system. Compared with the traditional direct torque control, the proposed method can adapt to different load conditions in real time, effectively inhibit the torque pulsation and magnetic chain fluctuations, reduce the energy loss, and improve the operation efficiency of the motor. Study using public data set, the different power levels of permanent magnet synchronous motor experimental validation, experiment shows that the fuzzy tree model significantly reduced the torque pulse (error by 40%), magnetic chain fluctuation (to 65%) and energy loss (power loss by 44%), and improve the dynamic response (response time by 25%) and efficiency (8%). Energy loss is greatly reduced with motor power loss decreasing from 4500 W to 2500 W up to 44%; in loss type 1, energy loss decreases to below 1000 W. For the dynamic performance, the response time was shortened by 15% to 25%, and the response time was reduced from 10 ms to 8 ms. In addition, the motor efficiency is generally increased by about 8%, showing higher operational stability and adaptability. The results show that the direct torque control method based on fuzzy tree model provides an efficient and reliable solution to optimize the performance of permanent magnet synchronous motors, especially in high power scenarios. The innovation of this paper is to use the fuzzy tree instead of the traditional switch table, and significantly improve the control performance through the adaptive rule adjustment.