Yves Mollet

This paper deals with the two-dimensional (2D) finite-element (FE) and magnetic-equivalent-circuit (MEC) modelling of an existing 10 kW brushed wound-field DC machine, taking magnetic cross-saturation into account. The 2D FE model allows to thoroughly investigate the demagnetizing effect of the armature current on the pole flux and the cross-saturation effect on the field and armature-winding flux linkage. A minimum-topology MEC, allowing for these effects is proposed. Some of the reluctances are fitted stepwise using the FE model. It is shown that the simple MEC generates comparable results also at no-load and load conditions, but with highly reduced computational costs. All numerical results are validated experimentally.

ABSTRACT This paper deals with an on-line method for turn-to-turn short-circuit fault detection in low-voltage permanent-magnet synchronous machine drives. Due to the closed-loop control, the fault effects are reflected in the voltage. Therefore, an appropriate diagnostic index is proposed, which is derived from the positive- and negative-sequences of the voltage references. These sequences are obtained in the time domain via adaptive filters, which require only a few calculations. To increase the sensitivity to the fault, the algorithm is only applied to a part of the voltage references, i.e. the output of the proportional-integral controllers. Further, the cumulative-sum algorithm is introduced to cope with changes of small magnitudes. This algorithm allows a change of a fault index to be detected and can be used as a decision system. The resulting fault detection scheme is computationally cheap and can be embedded in the control unit. Simulations and experimental results validate the proposed method in steady state and the performances under non-stationary operating conditions are also investigated.

ABSTRACT This paper deals with an on-line method for turn-to-turn short-circuit fault detection in low-voltage permanent-magnet synchronous machine drives. Due to the closed-loop control, the fault effects are reflected in the voltage. Therefore, an appropriate diagnostic index is proposed, which is derived from the positive- and negative-sequences of the voltage references. These sequences are obtained in the time domain via adaptive filters, which require only a few calculations. To increase the sensitivity to the fault, the algorithm is only applied to a part of the voltage references, i.e. the output of the proportional-integral controllers. Further, the cumulative-sum algorithm is introduced to cope with changes of small magnitudes. This algorithm allows a change of a fault index to be detected and can be used as a decision system. The resulting fault detection scheme is computationally cheap and can be embedded in the control unit. Simulations and experimental results validate the proposed method in steady state and the performances under non-stationary operating conditions are also investigated.

ABSTRACT This paper focuses on the implementation of LCL filter active damping in the control of an inverter for a three-phase grid emulator application. The studies about LCL damping in the literature focus on the case of an inverter injecting current into the grid while a grid emulator application is rather about establishing correct voltage waveforms and emulating the grid for any kind of load connected to it, which brings a different perspective. After having analyzed the stability of the LCL filter taking into account possible load variations, three damping methods are compared and assessed with respect to those concerns.

ABSTRACT This paper presents a new on-line open-phase and current-sensor fault detection and isolation method for permanent-magnet synchronous generator (PMSG) drives with neutral point voltage control. Simple signal processing implementations for sensor outage and open-phase faults detection are combined with a cumulative-sum (CUSUM) algorithm capable of detecting smaller sensor output errors (offset or gain variation). The algorithm uses output signals from eight current sensors (two on each phase and two on the neutral wire). Experimental tests have been done on a 2kW-PMSG at different speeds and braking torques where different sensor-fault combinations have been tested with and without open-phase fault. Finally a test has been done while the shaft speed and the torque were varying in order to test the ability of the algorithm to work in transient conditions. Results show that the algorithm allows 3-out-of-6 redundancy in case of sensor outage and 5-out-of-6 redundancy in case of sensor offset or gain change with globally limited computational cost. Combinations of two outages and one offset or gain error can be isolated. In case of open-phase fault the algorithm can continue to detect sensor faults if at least one sensor is working on each remaining phase and on the neutral. A short fault detection time is obtained in case of sensor outage with a transitional detection by the CUSUM algorithm followed by the final detection by the signal processing algorithm.

This paper presents a comparison of soft and hard
chopping on an 8/6 SRM in terms of noise, vibration and
harshness. Transient-state measurements are used to plot speedfrequency
signatures of current, vibration and acoustic noise of
the SRM for different load torques. With this technique speedrelated
frequencies can be distinguished from resonance
frequencies, and therefore, more information can be extracted
from the plots. The results show that hard chopping increases the
loudness of acoustic noise compared to soft chopping, with a
frequency shift to higher values. This however leads to an
attenuation of the vibration and noise amplitudes at the most
critical resonance of the SRM, situated at a relatively low
frequency.

Switched reluctance machines (SRMs) benefit from major advantages, such as cheap and robust construction and electrical fault tolerance, but also suffer from noise, vibration and harshness (NVH) issues. The present paper focuses on the NVH influence of the current hysteresis controller, which is generally used in SRM drives. The impact of changes in the controller sampling time is experimentally investigated on an 8/6 15 kW (30 kW peak) SRM. In practice the initial hysteresis controller, implemented on a dSPACE platform running at 10 kHz, is compared with an improved version running twenty times faster on external microcontrollers. Tests are performed in the form of run-ups, in order to study the drive in transient conditions and investigate a continuous speed range, while vibration, current and sound pressure waves are measured. Current waveforms are also simulated using AMESim software to compare with experimental measurements. Both controller implementations are modelled and both run-up and constant speed conditions are considered. To identify the frequency content related to the switching process, those simulation results are also analysed before and after having removed the current ripple due to switching through signal processing. Run-up results are plotted as waterfall diagrams, which both give a global overview of the frequency content for the considered speed range and allow distinguishing speed-order related harmonics and resonance frequencies. It is observed that the improved controller generates a shift of the spectrum towards higher frequencies with a global reduction of vibration and noise amplitude. Furthermore, the excitation of the dominating first ovalization mode of the SRM is considerably reduced.

The development of power converters has permitted the rise of switched reluctance machines (SRMs), due to their simple and cheap design and of their inherent fault tolerance. However, those machines suffer from torque ripple and noise and vibration issues, which constitute an obstacle to the extension of their application domain. This paper presents the influence of the bandwidth of a current-hysteresis controller on an 8/6 SRM in terms of noise, vibration and harshness. Simulations in AMESim environment and measurements on a 15 kW test bench are performed in transient state, based on continuous variation of the hysteresis bandwidth. Phase currents, acoustic noise and radial vibrations are measured without and with load, allowing for comparison of hysteresis-band-frequency plots. This transient state approach enables showing the evolution of a part of the frequency components with hysteresis bandwidth and distinguishing them from other components linked to speed or to structural behaviour of the motor. Results show that the current chopping by means of a hysteresis controller generates broadband frequencies in the phase currents that also appear in the vibration and sound pressure wave measurements. As the bandwidth increases, switching-related components are shifted towards lower frequencies, resulting in a higher excitation of the main resonance modes of the machine. Due to the higher current ripple in the case of hard chopping mode, the influence of the bandwidth is in general more important than in soft chopping mode. However, for the same bandwidth the ripple occurs at much lower frequencies in the latter case and therefore more attention has to be paid to possible interaction with the resonance frequencies.

ABSTRACT This paper presents a new on-line open-phase and current-sensor fault detection and isolation method for permanent-magnet synchronous generator (PMSG) drives with neutral point voltage control. Simple signal processing implementations for sensor outage and open-phase faults detection are combined with a cumulative-sum (CUSUM) algorithm capable of detecting smaller sensor output errors (offset or gain variation). The algorithm uses output signals from eight current sensors (two on each phase and two on the neutral wire). Experimental tests have been done on a 2kW-PMSG at different speeds and braking torques where different sensor-fault combinations have been tested with and without open-phase fault. Finally a test has been done while the shaft speed and the torque were varying in order to test the ability of the algorithm to work in transient conditions. Results show that the algorithm allows 3-out-of-6 redundancy in case of sensor outage and 5-out-of-6 redundancy in case of sensor offset or gain change with globally limited computational cost. Combinations of two outages and one offset or gain error can be isolated. In case of open-phase fault the algorithm can continue to detect sensor faults if at least one sensor is working on each remaining phase and on the neutral. A short fault detection time is obtained in case of sensor outage with a transitional detection by the CUSUM algorithm followed by the final detection by the signal processing algorithm.

A combined encoder and rotor-current-sensor fault detection and isolation algorithm for doubly-fed induction machines is presented. A rotor-position and current estimator based on air-gap power vector computation is used for residual generation. A cumulative-sum function permits detecting and isolating the investigated faults. The algorithm is experimentally tested on a 3 kW generator in steady and transient regimes. Sensor outages as well as current-sensor gain drifts and offsets are investigated. Transients do not significantly affect the algorithm, which also allows for a very short detection time for the most critical faults.