News Updates Monday 24th Oct 2016 :
  • Welcome to International Press Corporation, world's leading publishers, We have served more than 10000+ authors
  • Articles are invited in engineering, science, technology, management, industrial engg, biotechnology etc.
  • Paper submission last date of Sept/Oct 2016 issue is 25 Oct 2016, Submit online or at
  • Our journals are indexed in University of Regensburg Germany, Google Scholar, Cross Reference data bases
  • Applications for reviewers are invited and can be sent directly to concerned editor's mail

Noninvasive External Faults Detection of Induction Motor using Feedforward Neural Network

Author : Kalpesh J. Chudasama and VipulShah

Pages : 307-315
Download PDF


Induction motors are come across various abnormalities and faults categorized as external and internal faults. This paper discusses the performance of the Feedforward Neural Network (FFNN) as non invasive fault detection technique for the external faults/abnormalities in induction motor. Alongwith normal condition various external faults like Overload (O/L), Overvoltage (O/V), Undervoltage ( U/V), Stalling, opening of any stator phase, and voltage unbalances are simulated on Squirrel cage induction motor and FFNN train and tested using Levenberg-Maquardt (LM) Backpropagation algorithm using Matlab/Simulink. The various external faults are produced with wide range of operating three phase voltages which may exist in field for the find suitability of Artificial Neural Network (ANN) as fault detector. The main objective is to detect such external fault conditions correctly and also to find best ANN configuration. Number of ANN configurations are trained, parallel validated and tested for test (unseen) data with different combination like Early Stopping (ES) or Bayesian Regularization (BR) as generalization technique, Tansig or Purelin as activation performance function in the output layer and different processing technique of data. Good result of each configuration is shown and best among them found. 100% stastical parameters such as total classification accuracy, sensitivity and specificity for both test (unseen) and input data obtained for best Neural Network (NN) configuration. The best NN configuration found with BR as generalization, Purelin as activation function in output layer and data preporcessed with principle component analysis (PCA) after mapping each row of data mapped to have zero mean and unity Standard Deviation (SD). All kind of voltage unbalance like 1phase (ø) O/V, 2ø O/V, 1ø U/V, 2ø U/V, 3ø O/V, 3ø U/V and phase displacement is considered in voltage unbalance case and opening of any phase condition considered in stator open phase. Stalling at start detected within first few cycles which can eliminate the need of waiting for safe stalling time like in conventional protection.

Keywords: Induction Motor, Neural Network (NN), Early stopping, Feedforward Neural Network, Backpropagation, Bayesian Regularization, Standard deviation, Principle component analysis , root means square(RMS)


Article published in International Journal of Current  Engineering  and Technology, Vol.3,No.2 (June- 2013)



Call for Papers
  1. IJCET- Sept/Oct-2016 Issue

    Submission Last Date
    25 Oct 2016
  2. IJTT-Sept-2016
  3. IJAIE-Sept-2016
  4. IJCSB-Sept-2016
  • Inpressco Google Scholar
  • Inpressco Science Central
  • Inpressco Global impact factor
  • Inpressco aap

International Press corporation is licensed under a Creative Commons Attribution-Non Commercial NoDerivs 3.0 Unported License
©2010-2016 INPRESSCO® All Rights Reserved