1 INTRODUCTION
Current sensors are widespread in electrical systems [1], [2]. They can be used only for monitoring but in the most cases they are needed for current regulation loops. Current regulations can be linear using Proportional Integral (PI) regulators [3] or based on direct control techniques, with hysteresis controllers [4]. In both cases, performances and robustness of the whole system depend strongly on measurement accuracy. However, several faults can deteriorate the quality of the output sensor. Faults can occur on the sensor itself, although continuous technology progress led to current sensor less sensitive to external disturbances sources, like EMC, and less affected by nonlinearities [5], [6]. Faults can also due to external circuit, like offset or gain error, bad communications or disconnections, especially in transport applications. Effects of these faults have been largely reported in literature, for DC and AC drives [7], [8], [9], for transport application [10], renewable energy systems with power converters like variable speed systems [11], [12], [13], [14]. The largely used approach for sensor FDI is based on residuals generated by comparison between measured sensor output and reconstructed sensor output, using other system sensors. The reconstruction of the considered sensor output is performed using estimators [15], or observers [12], [16], [17], [18]. Besides the usual comparison results between the two approaches, (less computation time for estimators, but less robustness to system parameters variations) both methods require a deep knowledge of the system under control, an accurate model and relatively long time to adapt and adjust the FDI algorithm for specific application.