Adaptive neural network control for rotor’s stable suspension of Vernier-gimballing magnetically suspended flywheel

Abstract

Vernier-gimballing magnetically suspended flywheel with conical hybrid magnetic bearing can produce gyro moment by tilting the rotational rotor around a certain radial direction. When rotor is tilted, the nonlinear variation of conical magnetic bearing’s displacement stiffness and the coupling interference can result in not only poor stability of suspension control system but also precision’s degradation of gyro moment. To solve these two problems, the forces acting on tilted rotor are analyzed, and furthermore, a nonlinear model considering variation of displacement stiffness is constructed accordingly, then a radial basis function neural network is utilized to estimate the nonlinear variation of displacement stiffness and disturbance, and the adaptive controller with nonlinear variation compensation is presented based on the Lyapunov stability theory. Compared with cross-feedback method in tilting control and proportional–integral–derivative method in translation control, simulation researches are done, and the results indicate that the presented control method does well in estimating nonlinear variation and has better performances on suspension control when rotor is tilted.

PIDNN control for Vernier-gimballing magnetically suspended flywheel under nonlinear change of stiffness and disturbance

“…The upper and lower bounds are taken as LB = [0 0 0] and UB = [100 100 100], and the initial parameters and the optimization results are shown in Table 2. The performance of the RBF neural network control is cited in the paper presented in Tang et al 21…”

“…12 A feedback linearization is employed in combination with robust control techniques for the regulation of a single-axis test rig actuated by a multiple-pole MB; 13 however, the performance of this method will be reduced when the rotor tilts from the equilibrium point. 14 An adaptive backstepping control used to compute nonlinear control currents can improve control performance, 15 but it has the feature of over-parameterization, which will complicate the system and increase the dynamic order. Lin et al 16 researched a double-integral sliding mode control system to adjust and stabilize a highly nonlinear magnetically suspended rotor system to improve its robustness, but this method requires a large amount of computation and inevitable chatter exists in the sliding mode control.…”

“…The design process needs to be further improved. 21 Some literature has proved that the PIDNN method has the ability to solve nonlinear problems. Zribi et al 22 proposed an adaptive tuning method of PIDNN controllers for nonlinear processes, and the weights are updated by using an improved gradient descent method.…”

Stability analysis of adaptive control using fuzzy adapting rate neural emulator: Experimental validation on a thermal process

“…Although many researchers use neural networks in adaptive control, [14][15][16][17] one of the main issues with adaptive control is that the adaptive controller needs an online adjustment. 18 The major basic algorithm to update these parameters is the gradient backpropagation.…”

Suppression of high-frequency disturbance to satellite by Vernier-gimballing magnetically suspended flywheel

“…As for a 5 degree-of-freedom (5-DOF) MSFW, Tang et al, 2015 pointed out that its angular momentum vector can be tilted actively with respect to the spacecraft body by active magnetic bearings (MBs), and the tilting torque is usually more than 1Nm and larger than 0.1Nm by far. The MSFW with Vernier-gimballing ability is named as Vernier-gimballing MSFW, and the tested maximum torque introduced in this paper is up to 4.36Nm (Tang et al, 2017). As an actuator for the satellite attitude control in 3-DOFs, the principle of Vernier-gimballing MSFW to output torques is the same as that of the double flywheel or double gimbal variable speed control moment gyroscope (Fang et al, 2015; Peng et al, 2015), but the bandwidth of the Vernier-gimballing control of flywheel is 7.7Hz and is bigger than that of control moment gyroscope, which is generally 5Hz (Xu et al, 2015), so the Vernier-gimballing MSFW can respond to the command rapidly.…”