Abstract
In this paper, an on-line parameter identification algorithm to iteratively compute the numerical values of inertia and load torque is proposed. Since inertia and load torque are strongly coupled variables due to the degenerate-rank problem, it is hard to estimate relatively accurate values for them in the cases such as when load torque variation presents or one cannot obtain a relatively accurate priori knowledge of inertia. This paper eliminates this problem and realizes ideal online inertia identification regardless of load condition and initial error. The algorithm in this paper integrates a full-order Kalman Observer and Recursive Least Squares, and introduces adaptive controllers to enhance the robustness. It has a better performance when iteratively computing load torque and moment of inertia. Theoretical sensitivity analysis of the proposed algorithm is conducted. Compared to traditional methods, the validity of the proposed algorithm is proved by simulation and experiment results.
Highlights
The permanent magnet synchronous motor (PMSM) has several advantages such as high torque density, high precision, and high efficiency
In [23], an algorithm which integrates a Kalman Observer (KO) and recursive least squares (RLS) was proposed, where the RLS estimator is used for inertia identification and the KO is mainly used to obtain the knowledge of load torque, but the authors in [23] mentioned that it’s hard to choose appropriate values of the system noise matrix and measurement noise matrix in the KO
When V(k)VT (k ) in Table 2 are large, TL and ωprovided by KO, which are used as the inputs of RLS, subsequently causing a larger e(n) in (12), and in this situation, from (18), λ becomes smaller to give less weight to these unprecise historic data, which can speed up the process of the inertia identification and adapt to the load torque variation
Summary
The permanent magnet synchronous motor (PMSM) has several advantages such as high torque density, high precision, and high efficiency. In [23], an algorithm which integrates a Kalman Observer (KO) and recursive least squares (RLS) was proposed, where the RLS estimator is used for inertia identification and the KO is mainly used to obtain the knowledge of load torque, but the authors in [23] mentioned that it’s hard to choose appropriate values of the system noise matrix and measurement noise matrix in the KO. The contribution of this research is that while nowadays the control performance of most most servo products, such as Yaskawa, Fuji and Panasonic, is very good (the bandwidth of their speed loop and current loop is getting higher and higher), their auto-tuning function cannot operate with a sine-wave load, which is common for the robotic manipulator arms: they cannot identify inertia under a time-varying load-torque, which is exactly what is needed in the field of robotics.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
