A new fault diagnosis approach for induction motor using negative selection algorithm and its real-time implementation on FPGA

Abstract

Condition monitoring of induction motors has become an important issue of researchers in recent years. The detection of broken rotor bar faults is one of the most difficult problems and many methods have been proposed for accurate detection of these faults. In recent years, some studies have been proposed to improve the diagnostic performance by combining different signal processing techniques. However, the proposed methods require high computational complexity. The contribution of this study is threefold. The first one is a new feature extraction method to distinguish different motor conditions by analyzing one phase of induction motor steady-state current. The phase space of the feature signal is constructed by using determined time delay and embedding dimension. The second contribution is to optimize the detectors of the negative selection algorithm by clonal selection. The proposed clonal selection algorithm minimizes the overlap between the detectors and maximizes the coverage of the anomalous data. Because the feature extraction method and test stage of the negative selection algorithm have low computational complexity, the last contribution is Field-Programmable-Gate-Array (FPGA) implementation for online detection of rotor related faults. The obtained results indicate that the proposed methodology demonstrates a high performance for diagnosis of rotor faults in induction motors.