This work proposes a new method for the estimation of the load torque in mechatronics systems that are equipped with electrical drives and are characterized by repetitive mechanical cycles. The procedure examined in this research allows an enhancement of the control quality by using a feedforward controller based on the estimated load torque. A harmonic speed controller is also used for improving the velocity control. Its structure can also be integrated in the procedure for the identification of the torque. A second important contribution of this research is the utilization of the estimated load torque for the detection of failures in rolling bearings.
As the demanded load torque in many production machines is repetitive, and as such a periodic function of the shaft angle of the driven machine, it can be represented in terms of Fourier series. Hence, the estimation of the load torque can also be carried out through the calculation of the Fourier coefficients by using either a phenomenological approach or an empirical one. After sufficient online learning on the running mechatronics system at different operating speeds, sets of Fourier coefficients can be obtained by using a sliding-window method. The two methods allow the load torque estimation to be continuously conducted within the whole operational range of the repetitive mechanical system.
The estimated load torque is used for two purposes. First, it is utilized as a compensation signal in a feedforward control scheme to improve the quality of the velocity control by canceling speed oscillations originated from the angle-dependent load torque. Second, it is adapted in a diagnostic procedure for the detection of bearing faults. Since a faulty bearing causes changes on the load torque, its spectrum contains information related to the bearing failures. A diagnostic procedure is proposed that does not require any additional sensors than the usually installed in an electrical drive, and it is capable of detecting the two most commonly found types of bearing faults including single-point defects and generalized roughness fault.
Another approach examined in the frame of this work aims the enhancement of the control quality. This approach relies on the harmonic speed control. For this purpose, two structures for the speed control are proposed. One is known as harmonic speed control, where each harmonic in the spectrum of the speed error is regulated by a proportional-integral (PI) controller. As previously mentioned, the proposed control structure can also be used for the online calculation of Fourier coefficients in the aforementioned load torque estimation procedure.
The second scheme is called PI-R speed controller, where the conventional PI speed controller is augmented by resonant (R) parts to control specific harmonics of the speed error. As the proposed speed controller separately manipulates harmonics of the speed error, it ensures a good dynamic tracking response of the speed control loop.
An implementation of the control scheme without mechanical encoder, a so called sensorless control scheme, complements the proposed methods. In this way the estimation of the load torque, the detection of the bearing faults, and the harmonic speed control can be carried out without mechanical encoder. A signal injection technique and an enhanced voltage model are used to estimate the rotor position of the driven machine. The utilized machine is a permanent magnet synchronous machine (PMSM) that can be operated in the sensorless field-oriented control scheme in the low-speed as well as in the high-speed regions. A combined angle estimator which is based on the signal injection technique and on the enhanced voltage model was developed for obtaining a smooth speed transition between the two speed regions.