Adaptive Variable-Frequency Speed Control For Ultrasonic Motor Based On RBFNN

Being a motor totally based on unique working mechanism, the ultrasonic motor (USM) is quite different from the traditional electromagnetic motors, it is operated on the base of the inverse piezoelectric effect of the piezoelectric ceramics and its driving torque is converted from the ultrasonic vibration instead of from the electromagnetic effect. The traveling wave type USM is analyzed in detail in this paper as follows.At first, the working mechanism of the USM is described. The unique construction of the traveling wave type USM is analyzed in detail, the principle of the production of the vibration, the formation of the standing wave, the combination of the traveling wave and the running rules of the particles on the surface of the stator are discussed.Then a practical driving circuit for the USM is designed. The USM is a typical mechanical and electrical composite, so its performance is affected directly by the quality of its driving circuit. A driving circuit for the motor is designed according to its peculiarity, where a signal whose frequency is adjustable is generated by the self excited oscillation, then the signal is shifted in phase, so four sets of square-wave signals that differ 90 degrees in succession are generated, which can be applied to drive two sets of push-pull main circuits. Four input signals of the motor are obtained through experiments. And the driving frequency and voltage is easy to regulate. So this circuit meets the requirement of the most experimentation very well.Next, the optimized speed control strategy is discussed. Up to the present, no accurate mathematical model capable of expressing both the dynamic and the steady characteristics of a USM has been derived, because of the resonant frequency float, the complicated coupling among the variables, high nonlinear characteristics and uncertainty of the parameters and so on. This paper proposes a new adaptive speed-control scheme. Two closed loops are constructed. The inner one is built as a mechanical resonant frequency compensator. The frequency is regulated in the other loop by the Radial Basis Function neural network (RBFNN) controller whose parameters are adjusted on-line by the use of another RBFNN, which can approximate the nonlinear input-output mappings of the driving system. With the proposed strategy, excellent flexibility as well as high precision and good robustness is obtained by experiments based on DSP.At last, on the base of the previous method, an improvement has been made on the adjustment of the parameters of the RBFNN by the use of hybrid hierarchical genetic algorithm. Because only one RBFNN is used in this method, the control for the USM is simpler and more efficient.