A Novel Method for Induction Motor Fault Identification Based on MSST and LightGBM

Induction motors are widely used in industries and daily lives. The sudden occurrence of motor faults can cause huge economic losses and even inflict human injuries. To identify the motor faults promptly and improve the operation reliability. This paper presents a novel method for induction motor fault identification. The method consists of five main steps: multi-synchrosqueezing transform (MSST) analysis, signal texture feature extraction, multi-domain feature vector construction, feature dimension reduction with Linear Discriminant Analysis (LDA), and fault identification with light gradient boosting machine (LightGBM) algorithm. In the process, MSST is first utilized to get an energy-concentrated time-frequency (TF) representation of the signal, the texture features of TF images are calculated with gray level co-occurrence matrix (GLCM), then combined with time-domain and frequency domain-features, a multi-domain feature matrix is constructed and a low-dimension feature matrix is obtained by LDA to improve the feature quality. After that, a LightGBM model is trained with the new low dimension feature matrix to identify the motor faults. The proposed method is validated by the experiment dataset of the drivetrain dynamics simulator, and the results proved that the approach can accurately classify different induction motor faults.

[1]  Pratyay Konar,et al.  Multi-class fault diagnosis of induction motor using Hilbert and Wavelet Transform , 2015, Appl. Soft Comput..

[2]  Kosin Chamnongthai,et al.  Fusion of color histogram and LBP-based features for texture image retrieval and classification , 2017, Inf. Sci..

[3]  Tie-Yan Liu,et al.  LightGBM: A Highly Efficient Gradient Boosting Decision Tree , 2017, NIPS.

[4]  David A. Clausi,et al.  Preserving boundaries for image texture segmentation using grey level co-occurring probabilities , 2006, Pattern Recognit..

[5]  J. Friedman.  Greedy function approximation: A gradient boosting machine. , 2001 .

[6]  Zhipeng Feng,et al.  Application of multi-domain sparse features for fault identification of planetary gearbox , 2017 .

[7]  Fei Wang,et al.  Bispectrum Texture Feature Manifold for Feature Extraction in Rolling Bear Fault Diagnosis , 2019, Mathematical Problems in Engineering.

[8]  Xiang Wang,et al.  Bearing Fault Diagnosis Based on Statistical Locally Linear Embedding , 2015, Sensors.

[9]  Vilas N. Ghate,et al.  Optimal MLP neural network classifier for fault detection of three phase induction motor , 2010, Expert Syst. Appl..

[10]  Mehmet Karaköse,et al.  Combined intelligent methods based on wireless sensor networks for condition monitoring and fault diagnosis , 2015, J. Intell. Manuf..

[11]  Iqbal Gondal,et al.  A data mining approach for machine fault diagnosis based on associated frequency patterns , 2016, Applied Intelligence.

[12]  Robert M. Haralick,et al.  Textural Features for Image Classification , 1973, IEEE Trans. Syst. Man Cybern..

[13]  Chee Peng Lim,et al.  Motor fault detection and diagnosis using a hybrid FMM-CART model with online learning , 2016, J. Intell. Manuf..

[14]  Tian Han,et al.  Fault Diagnosis System of Induction Motors Based on Neural Network and Genetic Algorithm Using Stator Current Signals , 2006 .

[15]  Shamik Sural,et al.  An Integrated Color and Intensity Co-occurrence Matrix , 2007, Pattern Recognit. Lett..

[16]  Bhim Singh,et al.  Investigation of Vibration Signatures for Multiple Fault Diagnosis in Variable Frequency Drives Using Complex Wavelets , 2014, IEEE Transactions on Power Electronics.