Mechanical resonance is a common problem in drive systems with elastic coupling. On-line adaptive notch filter is widely used to make systems stable and the key of this method is to identify natural torsional frequency from a speed feedback signal. However, because of common adoption of digital control and expansion of system bandwidth, oscillation frequency of the system is more likely to deviate from natural torsional frequency to a higher one. When oscillation frequency is shifted, the enabled notch filter with erroneous notch frequency causes an oscillation with a lower frequency and even makes resonance more severe. In order to explain this phenomenon, the classical two-mass model based classification of resonances is checked at first. Then, by taking digital control, current loop delay, and saturation nonlinearity into consideration, an improved digital mechanical resonance model is proposed and a criterion for oscillation frequency deviation is finally obtained. Furthermore, a more widely applicable and robust notch filter tuning strategy with no oscillation rebound is presented. In the end, the validity of aforementioned analysis and strategy is verified by experimental results.

Because they are economical, universal servos are chosen for most industrial applications. However, tuning them accurately poses some challenges. Accurate electrical and mechanical parameters are essential for model-based high-performance servo controller design. In motion control, the variation of inertia is much more significant than that of other parameters. Thus, inertia identification is a key to effective on-line controller parameter self-tuning, but most conventional inertia identification methods cannot be well applied in some complicated situations, such as those caused by irregularly and slowly varying speed. Additionally, traditional control parameter tuning theory displays a deep understanding of the relationship between controller gain and inertia, but lacks an analysis of the maximum bandwidth of the system. In this paper, based on a widely accepted structure which is composed of inertia identification and on-line controller self-tuning, several simple but useful modifications are proposed. Firstly, to reduce the noise from encoder quantization error and thus to improve the accuracy of inertia identification, a motor acceleration calculation method featuring an inconstant period is proposed. Then, the scope of the application is extended to make it suitable for position control by redesigning the inertia updating time of the conventional method. In addition, to guarantee the stability of servo systems, the upper constraint of the expected maximum system bandwidth is derived by taking the controller saturation nonlinearity and hardware capacity into consideration. Finally, a modified moderate systematic on-line servo controller parameter self-tuning method via variable-period inertia identification is presented. The validity, effectiveness and advantages of proposed strategies are verified by several experimental results.

Applying high-frequency power devices, such as gallium nitride (GaN) semiconductor devices, to the motor drive can significantly reduce system volume, losses, audible noise and increase system power density. In this paper, an LC filter designed as a sine wave filter is introduced into the servo drive system based on the GaN power device. An undamped method with variable delay time is proposed to solve the electrical resonance issue caused by LC filter and stator inductor. Detailed stability analysis is conducted, and rules on parameter tuning are provided. With additional proper time delay, the system can be stable without any extra sensors or power losses. Besides, this method is suitable for resonance suppression at variable switching frequency. Moreover, a wait-free and phase-continuous spread spectrum frequency modulation is applied to solve the issue of electromagnetic interference (EMI). The field-programmable gate array (FPGA) is used to achieve high control bandwidth when implementing the control algorithm. Finally, the ideal sinusoidal drive and low conducted EMI of permanent magnet synchronous motor (PMSM) is realized. Experiments on a 100 kHz GaN inverter with LC filter verify the validity of the proposed design and method.

For mechanical resonance suppression, the bi-quad filter is a powerful tool and has been widely used in industrial applications. Compared with the notch filter, the bi-quad filter has one more parameter for notch depth design, which is more suitable for servo systems with nature damping coefficients and limited resonant gains. Usually, the bi-quad filter tuning has three steps, including identification, discretization, and switching process. However, in industrial applications, these three steps still have room for improvement. In this article, three targeted modification methods are proposed. First, an identification method based on the twins-point search is proposed to find out all of the available potential elastic characteristics accurately and quickly, and get the resonant frequency, the rejection bandwidth, and the notch depth for filter design. Then, inspired by the discretization distortion correction of the notch filter, a parameter mapping Tustin method is used to achieve accurate discretization. Furthermore, considering the bi-quad filter's settling time, a “warm-up” buffer strategy prior to the filter switch-on is proposed to smooth the switching process. In the end, the validity and effectiveness of the proposed modification strategies are verified by experimental results.

Deadbeat predictive control (DBPC) requires accurate model parameters in its application, therefore it is difficult to use DBPC for surface mounted permanent magnet synchronous motor (SPMSM) with uncertain motor parameters. As a solution, this paper proposes a strategy to obtain the needed parameters, and then uses them to achieve DBPC on SPMSMs whose parameters are unknown. A detailed investigation on parameter sensitivity of DBPC is presented, and attention is paid to different influences between the incremental inductance and the apparent inductance. On this basis, parameter configurations in the DBPC are suggested. And novel current-controller-free self-commissioning schemes are proposed to identify the stator resistance together with the stator inductances under different magnetic saturation levels. Compared with existing commissioning schemes that do not depend on current controllers, solutions for commissioning voltage auto-tuning is provided. So the methods are able to automatically decide the amplitudes of the injected voltage for a given motor, and achieve a controllable current feedback during the commissioning. This brings significant convenience for the inductance identification at a specific saturation level. Moreover, the inverter nonlinearity compensation is not needed but still, identification accuracy can be guaranteed. The feasibility and effectiveness of the proposed method are confirmed by the achieved DBPC and the corresponding current tracking performances on two different SPMSMs.