This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT 1 An Approach on MCSA-Based Fault Detection Using Independent Component Analysis and Neural Networks Juan Enrique Garcia-Bracamonte, Member, IEEE, Juan Manuel Ramirez-Cortes , Senior Member, IEEE, Jose de Jesus Rangel-Magdaleno, Senior Member, IEEE, Pilar Gomez-Gil, Senior Member, IEEE , Hayde Peregrina-Barreto, Member, IEEE, and Vicente Alarcon-Aquino, Senior Member, IEEE Abstract—This paper presents a novel approach on motor current signature analysis (MCSA) for broken bar fault detection of induction motors (IMs), using as input the current signal measured from one of the three motor phases. Independent component analysis (ICA) is used over the Fourier-domain spectral signals obtained from the input and its autocorrelation function. The standard deviation of spectral components within a region of interest (ROI) of an ICA signal output was found to exhibit substantial differences between damaged and healthy motors. Separation of the ROI in one, two, and three sectors leads to an improved extraction of feature vectors, which are further fed into a neural network for classification purposes. The assessment of the proposed method is carried out through several experiments using two damage levels (broken bar and half broken bar) and two load motor conditions (50% and 75%), with a classification accuracy ranging from 90% to 99%. The contribution of this paper lies in a new technique of signal processing for ICA-based feature extraction in a 3-D feature space for IM fault diagnosis. Index Terms—Broken bar, fault detection, independent com- ponent analysis (ICA), induction motor (IM), motor current signature analysis (MCSA), neural network (NN). I. I NTRODUCTION I NDUCTION motors (IMs) are a mainstay in the industrial world; however, like any other electromechanical machines, they are susceptible to many types of faults in field applica- tions. In the past decades, fault diagnostics and condition mon- itoring in rotary machinery have been a challenging research topic in the scientific community. Specifically, profound efforts have been devoted to IM fault diagnosis due to the economic and technical consequences, if the problems are not timely attended. Squirrel-cage IMs are among the most common Manuscript received July 24, 2018; revised January 17, 2019; accepted February 1, 2019. This work was supported by the National Institute of Astrophysics, Optics, and Electronics, Mexico. The Associate Editor coor- dinating the review process was Edoardo Fiorucci. (Corresponding author: Juan Manuel Ramirez-Cortes.) J. E. Garcia-Bracamonte was with the Electronics Department, National Institute of Astrophysics, Optics and Electronics, Puebla 72840, Mexico. J. M. Ramirez-Cortes and J. de Jesus Rangel-Magdaleno are with the Elec- tronics Department, National Institute of Astrophysics, Optics and Electronics, Puebla 72840, Mexico (e-mail: jmram@inaoep.mx). P. Gomez-Gil and H. Peregrina-Barreto are with the Computer Science Department, National Institute of Astrophysics, Optics and Electronics, Puebla 72840, Mexico. V. Alarcon-Aquino is with the Department of Computer Systems and Electronic Engineering, University of the Americas, Puebla 72810, Mexico. Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TIM.2019.2900143 rotary machines used in industry, representing around 85% of power consumption in industrial plants. Among several types of faults which can appear in IMs, broken bars represent a very common problem, particularly in heavy duty systems. Many methodologies and techniques aiming on fault detection have been published in the past decades. Most techniques perform fault detection using current signal analysis [1], [2], vibration analysis [3]–[6], or a combination of both [7]. Many of these proposed techniques use the fast Fourier transform (FFT) to analyze the frequency spectrum in search of the spectral characteristics of the faults. Once a feature extraction process is implemented to represent the motor conditions, a classifica- tion technique should be incorporated into the system. Neural networks (NNs) classifiers have been proved to provide a good approach in terms of classification rate, accuracy, and appropriateness for hardware implementation. For example, in [8], a classifier for fault detection of IMs based on Cascade NN is proposed. They used the stator current on time domain to extract features with principal component analysis, and those features feed the cascade NN to detect stator winding interturn short, rotor eccentricity or both. The work proposed in [9] used NNs on features extracted from vibration signals on time domain or alternative on the frequency domain. The methodology detects healthy motor, inner race fault, and outer race fault. In previous works, it is shown that the spectrum of a damaged motor exhibits spurious spectral components around the fundamental frequency [10], [11]. Let us denote f s as the fundamental frequency and those spurious frequencies as f bL and f bR . Location and magnitude of these spurious frequencies are related to the severity of the fault and the motor load: the larger the severity of the fault, the larger the magnitude of f bL and f bR . Similarly, the more loaded the engine is, the farther away are the spurious spectral components from f s [7]. Therefore, separation techniques of the referred spectral components are crucial as the first step to come up with analysis systems aiming to do fault diagnosis. In this paper, a novel approach to motor current signature analysis (MCSA) for IM fault detection, using independent component analysis (ICA) applied to stator current data in the spectral domain, is proposed. In recent years, ICA has been used as a relevant technique in support of fault diagnosis through several approaches. In [12], a vibration-based bearing defect analysis is presented. In their approach, ICA is used to separate multichannel signals in a bearing fault detection 0018-9456 © 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.