Parameter identification is a critical factor influencing the dynamic performance of permanent magnet synchronous motor (PMSM)-driven servo systems. However, mechanical parameter identification encounters the following challenges: multiple parameters that are significantly mutually coupled and difficulties in direct measurement, especially with friction parameters. Currently, most scholars have navigated these challenges through partial parameter identification, either by ignoring Coulomb friction, neglecting the dynamic processes of Coulomb friction, or identifying only the parameters associated with the friction model. In this study, a scheme using a parallel-cascaded extended sliding-mode observer (PCESMO) is proposed to address these challenges. The PCESMO relies solely on phase currents and rotor position to identify parameters, including the moment of inertia, viscous damping, load torque, and parameters of a simplified LuGre model that considers the dynamic process of Coulomb friction during self-commission. Additionally, the PCESMO has low computational complexity and is easy to implement. The effectiveness of the PCESMO was validated through both simulations and experiments. Compared to other competitive methods, PCESMO-identified parameters result in optimal control effects.

To overcome the issue of poor absolute positioning accuracy owing to the low stiffness of serial robot arms, this study proposed a new dual-arm robot (DAR) through the optimisation of the configuration of traditional robots. The DAR adopts a novel kinematic structure comprising a pair of 3-degree of freedom (DOF) robots: the workpiece and tool arms. A kinematic modelling method based on a closed-loop virtual kinematic chain was proposed to solve the problem of synchronous control of both arms via one controller. The workspace of the DAR was calculated, and the singularity was analysed. Furthermore, the stiffness was evaluated through finite element analysis. The results demonstrate that the DAR exhibits significantly higher stiffness, thus enabling considerably better positioning accuracy. To verify the superiority of the proposed DAR, a machining platform was developed based on the DAR for grinding and polishing an aero-engine blade. Compared to a traditional serial 6-DOF robot, the proposed DAR significantly improved the contour accuracy, with a surface roughness Ra 0.3 v.s. 0.4 $\mu $ m. Furthermore, the profile deviation of the blade body and edge were $+$ 0.05 mm, -0.07 mm v.s. $+$ 0.075 mm, -0.07 mm, and $+$ 0.05 mm, -0.03 mm v.s. $+$ 0.05 mm, -0.05 mm, respectively. Note to Practitioners —This study aimed to enhance the machining accuracy of industrial robotic machining. A novel dual-arm robot with superior stiffness was designed and compared with a conventional serial 6-DOF robot. A prototyping DAR was developed as the foundation for a grinding platform’s subsequent design and construction. The platform was utilised in aero-engine blade grinding and polishing. The contour accuracy of the processed blade drastically improved compared with that of the traditional serial 6-DOF robot system. Experiments verified the proposed robot’s feasibility and superiority in improving the robotic machining accuracy. Furthermore, the proposed DAR exhibited good potential for application in other robotic machining scenarios, such as assembly, carving, etc.

An identification method is presented for the arctan two-parameter model, which describes the deadtime effect voltage error of PMSM drives' three-phase voltage source inverter (VSI) during self-commission. In this method, the motor maintains a standstill during the whole identification process. First, a disturbance observer is used to improve the tracking performance of the total voltage error. Then, a technique is used to extract the deadtime effect voltage error from the total voltage error without a voltage sensor during the standstill condition. Finally, the core algorithm (utilizing the bisection method) is developed to solve the coupling problem in the two-parameter identification process, such that the parameters in the model converge simultaneously. Furthermore, a sufficient condition for the convergence of the algorithm is given, and a mathematical proof is presented. Experimental results show that the deadtime effect voltage error estimated by our method is closer to the measured voltage error than other state-of-the-art methods proposed in the literature.