Background:
Brushless DC motors are highly efficient motors due to its high torque to weight ratio, compact
design, high speed operating capability and higher power density. Conventional Hall sensor based rotor position sensing is
affected by the heating, vibration, interference and noise.
Objective:
The innovative, cost effective and easily implementable sensorless techniques are essential in order to achieve
high efficiency, reduced current and reduced torque pulsations. Further, a delay free, high load fast startup is also
important issue.
Methods:
In this paper an extensive review of various techniques based on the detection of freewheeling diode current,
phase back EMF zero crossoing point detection, back EMF integration method and third harmonic back EMF was done.
The study and effect of various PWM strategies on back EMF detection was studied. Later on the sensorless schemes
based on flux linkage estimation and flux linkage increment were introduced. The load torque observers, unknown input
observers, sliding mode observers, L∞-induced observers, H ∞ - deconvolution filter for back EMF estimation were
also reviewed. As the brushless DC motors have no back EMF at starting and for back EMF based commutation a
minimum speed is required for sufficient back EMF. Therefore various strategies of open and close-loop reduced current
startup have been studied to achieve effective commutation without reverse torque. Initial position detection (IPD)
schemes, which are mostly based on saliency and current response to inductance variation, is effective where reverse
torque is strictly prohibited. A detailed review of these initial position detection techniques (IPD) has also been presented.
Results:
The detailed mathematical and graphical analysis has been presented here in order to understand the working of
the state-of-art sensorless techniques.
Conclusion:
The back EMF detection using direct and indirect methods of terminal voltage filtering have the problem of
delay and attenuation, PWM noise, freewheeling diode spikes and disturbance in detected back EMFs is a drawback. The
parameter detuning, underestimation and overestimation, offset problem, system noise and observer gain variation etc.
limit the applicability of observer based technique. Therefore, a more robust and precise position estimation scheme is
essential.
UAVs in rail damage image diagnostics supported by deep-learning networks