Abstract-The aim of this paper is to study the feasibility of fault detection and diagnosis in a three-phase inverter feeding an induction motor. The proposed approach is a sensor-based technique using the mains current measurement. A localization domain made with seven patterns is built with the stator Concordia mean current vector. One is dedicated to the healthy domain and the last six are to each inverter switch. A probabilistic approach for the definition of the boundaries increases the robustness of the method against the uncertainties due to measurements and to the PWM. In high-power equipment where it is crucial to detect and diagnose the inverter faulty switch, a simple algorithm compares the patterns and generates a Boolean indicating the faulty device. In low-power applications (less than 1 kW) where only fault detection is required, a radial basis function (RBF) evolving architecture neural network is used to build the healthy operation area. Simulated experimental results on 0.3-and 1.5-kW induction motor drives show the feasibility of the proposed approach.

Recent feature selection scores using pairwise constraints (must-link and cannot-link) have shown better performances than the unsupervised methods and comparable to the supervised ones. However, these scores use only the pairwise constraints and ignore the available information brought by the unlabeled data. Moreover, these constraint scores strongly depend on the given must-link and cannot-link subsets built by the user. In this paper, we address these problems and propose a new semi-supervised constraint score that uses both pairwise constraints and local properties of the unlabeled data. Experimental results show that this new score is less sensitive to the given constraints than the previous scores while providing similar performances.

No abstract

No abstract

Abstract-The aim of this paper is to study the feasibility of fault detection and diagnosis in a three-phase inverter feeding an induction motor. The proposed approach is a sensor-based technique using the mains current measurement. A localization domain made with seven patterns is built with the stator Concordia mean current vector. One is dedicated to the healthy domain and the last six are to each inverter switch. A probabilistic approach for the definition of the boundaries increases the robustness of the method against the uncertainties due to measurements and to the PWM. In high-power equipment where it is crucial to detect and diagnose the inverter faulty switch, a simple algorithm compares the patterns and generates a Boolean indicating the faulty device. In low-power applications (less than 1 kW) where only fault detection is required, a radial basis function (RBF) evolving architecture neural network is used to build the healthy operation area. Simulated experimental results on 0.3-and 1.5-kW induction motor drives show the feasibility of the proposed approach.