Samenvatting
PH.D. Thesis - Fabien GABRIEL - 2012August 29, 2012Title : Low Speed Sensorless Control of Electrical Motors
Summary :
In some applications, such as vehicle propulsion, the permanent-magnet (PM) machines
present important benefits compared to other machine types : compactness, high-torque
production, no rotor electrical circuit and therefore simple control methods. Their optimal
control requires however the knowledge of the rotor position. This position can be measured
by dedicated position sensors coupled on the rotor. But they introduce a risk of failure,
require space and have a cost. Therefore, intensive research is performed in order to remove
these sensors through the development of estimation methods, referred to as position/
motion-sensorless or self-sensing methods.
Without additional sensor, the estimation methods are based the current and voltage
measurements used for the normal control operations also. A promising strategy for low
speed or standstill operations makes use of high-frequency signal-injection in order to identify
the orientation of a magnetic anisotropy related to the rotor position. The thesis focuses on
these anisotropy-based solutions. These anisotropic properties are particularly pronounced in
permanent-magnet synchronous machines (PMSM) and in more traditional synchronous
machines with salient poles. In PMSM, they are mainly due to the magnetic saturation effects
of the PM in the iron.
The experiments are performed on a specific PMSM type : a 3 kW BrushLess DC (BLDC) in-
wheel motor intended for the automotive research. This type of motor is particularly
challenging due to several non-ideal characteristics:
1) The magnetic anisotropy that can be shifted with respect to the real position of the rotor.
This is due to the presence of spatial harmonics in real machines that create disturbing
position-estimation oscillations and to the contribution of the normal-operating currents in
the magnetic anisotropy, leading to misalignment in the rotor-position estimation. The
harmonics are generally neglected in self-sensing methods but they are highly pronounced
in the our BLDC motor. An analytical analysis is proposed in this thesis.
2) Another problem concerns the voltage-source inverter (VSI) supplying the motor. For cost
and reliability reasons, the output voltage is often not directly measured. In that case, self-
sensing operations rely on the command voltage sent to the VSI. The behavior of the VSI is
however not perfectly linear. In particular, the relation between the expected output voltage
and the command voltage is very nonlinear when one phase current crosses zero due to a
phenomenon known as zero-clamping. Moreover, the current ripples (that result of the
pulse-width modulation (PWM) technique used for most of the VSI) complicate any attempt
to compensate for the zero-crossing nonlinearity. Solutions to detect and to prevent zero-
crossing is proposed in this thesis.
3) Most of the self-sensing methods based on the high-frequency signal-injection assume
that the machine behaves as a purely integrator through high-frequency variations of the
magnetic field when injecting high-frequency voltages. This assumption is however not
valid in the considered BLDC motor due to significant resistive effects (mainly due to eddy
currents). Moreover, this assumption must be reassessed when the rotation speed
increases (high speed operations). The impact on the position estimation are addressed
and a specific solution that completely removes the issue is proposed in the thesis.
| Datum prijs | 17 apr 2013 |
|---|---|
| Originele taal | English |
| Begeleider | Philippe Lataire (Promotor), Jacques Tiberghien (Jury), Rik Pintelon (Jury), Johan Deconinck (Jury), Royal Military Academy, Brussels Xavier Neyt (Promotor), Marc Acheroy (Jury), Frederik De Belie (Jury) & Ralph Kennel (Jury) |