Abstract: Aiming at the driving motor NVH problems caused by radial electromagnetic force of low orders, this paper gathers noise signals of a driving motor to analyze their order characteristics. An electromagnetic-structure-acoustic simulation model is also built and its accuracy is verified through experimental signals. An analytical model of radial electromagnetic force considering harmonic currents is established based on Maxwell tensor method. Besides, the radial electromagnetic force waves of the main time orders of are decomposed both on time and space dimensions. Combined with the transfer functions from each stator tooth of the motor to the noise measurement point, a noise prediction model at the measurement point is established based on the linear superposition method. The concentrated force of the radial electromagnetic force waves on each stator tooth is calculated by Cotes formula. By applying the formula to the noise prediction model, the contribution degree of various time and spatial orders of the radial electromagnetic force to the noise at the measuring point is obtained. Therefore, the harmonic current order that greatly affects the noise is figured out. Taking the noise pressure level at the measuring point as the optimization objective, the amplitude and phase of the harmonic currents are optimized through the genetic algorithm by considering the influence of the harmonic currents on torque pulses as the restriction condition. The optimized harmonic currents are injected into the electromagnetic-structure-acoustic simulation model to verify the optimization effect. The results indicate that, through the optimization of the harmonic currents, the motor vibration noise can be decreased by optimizing the radial electromagnetic force under the premise of controlling the torque pulse.