Three-level Inverter Fault Detection and Diagnosis Using Current-based Statistical Analysis

Mehdi Baghli (1) , Claude Delpha (2) , Demba Diallo (3) , Abdelhamid Hallouche (1)

1 Université Djilali Liabès [Sidi-Bel-Abbès]
2 L2S - Laboratoire des signaux et systèmes
3 GeePs - Laboratoire Génie électrique et électronique de Paris

This paper is an incremental contribution to the monitoring of power converters. It deals with open switch Fault Detection and Diagnosis (FDD) in three-level Neutral Point Clamped (NPC) inverter for electrical drives. The approach is based on the already available phase current measurements. The fault features that are evaluated in the time domain are the first four statistical moments. Three methods of FDD have been evaluated. The first one analyses the time domain evolution of the fault features. With this method, the fault is detected with a probability of detection higher than 80% when the operating speed is higher than a tier of the nominal speed and the SNR ≥ 40dB. However, the performances are degraded for low speed and for lower SNR. To improve the performances, we use the Cumulative Sum (CUSUM) algorithm designed for mean and variance variations. The optimality has been verified with the Kolmogorov-Smirnov test of the selected features Probability Density Functions (PDF). The fault is detected efficiently with both CUSUM indicators with almost 100% of detection probability even for the smallest considered fault duration of 100s.

Keywords: Three-level Neutral Point Clamped (NPC) inverter; Open Switch Fault (OSF); Intermittent faults effect; Fault Detection and Diagnosis (FDD); Cumulated Sum (CUSUM) analysis

Domains: Signal and Image processing

Mehdi Baghli, Claude Delpha, Demba Diallo, Abdelhamid Hallouche. Three-level Inverter Fault Detection and Diagnosis Using Current-based Statistical Analysis. Prognostics and Systems Health Management Conference (PHM-Chongqing 2018), Oct 2018, Chongqing, China. ⟨10.1109/phm-chongqing.2018.00123⟩. ⟨hal-01895228⟩