Stator Winding Faults Diagnosis in Induction Motor Based on ANN and ANFIS

Stator Winding Faults Diagnosis in Induction Motor Based on ANN and ANFIS | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Stator Winding Faults Diagnosis in Induction Motor Based on ANN and ANFIS Menshawy Mohamed, Mohamed Moustafa This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7242428/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract This paper proposes a new method using Artificial Neural Network (ANN) and Adaptive Neural Fuzzy Inference System (ANFIS) for diagnosis Inter Turn Short Circuit (ITSC) faults in induction motor. The proposed diagnosis procedure is based on the analysis of stator current. The study includes a comparative analysis of a various diagnostic methods such as decision tree, k-nearest neighbours, naive bayes, random forest and support vector machine. The time domain features extraction pre-processes the input data before entering to classifier model. The test accuracy and cross-validation analysis evaluate the model efficiency. The most important time domain features are selected to improve the performance of the classifier. ANN based the most important time domain features gives better performance as compared to all others classifier. This paper presents a comparison between ANN and ANFIS based on the most important time domain features, auto-regressive model and discrete wavelet transform. ANFIS based discrete wavelet transform achieves higher performance than ANN. The laboratory experiments on 1.5 HP squirrel cage induction motor under different loading conditions verify the proposed technique efficiency to diagnose fully various ITSC faults Inter Turn Short Circuit Faults Artificial Intelligence Machine Learning Time Domain Features Auto-Regressive Model Discrete Wavelet Transform Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 1. INTRODUCTION Electrical motors are widely utilized in manufacturing cycle, since it represents an important part in industrial processes because of its simple structure, reliability, ruggedness, cost effective design and ease of control [ 1 ]. The most common Induction Motor (IM) faults are short circuits winding or opening in the stator phase, broken rotor bar and bearing failures [ 2 – 4 ]. ITSC fault represents between 21% and 40% of the machine faults [ 5 ]. Due to high current flow in the short circuited coils and the decreased insulation the stator inter turn defect is occurred [ 6 ]. The breakdown of stator winding insulation leads to a short circuit fault. It takes very short time for short circuitry in the stator winding to damage the motor. This will break production in the industry. IM fault detection improves reliability and availability of an existing system at an early stage [ 7 ]. Recently many researchers have been concentrated on IM faults. Many fault detection techniques were illustrated. The stator winding inductance analysis and the ITSC defect diagnostic method in a several operating circumstances are proposed in [ 8 ]. Wavelet transformation algorithms and advanced digital signal processing technique have detected the fault in ITSC in induction motor [ 9 ]. Fuzzy Logic, genetic algorithms, ANN, and ANFIS are high potential data processing tool that creates fault diagnosis techniques [ 10 , 11 ]. ANN and ANFIS techniques are used to increase the accuracy for diagnosing of IM faults and overcome the drawbacks of the traditional techniques. ANN is a powerful tool that has been suggested for IM fault diagnosis [ 12 , 13 ]. ANN is used to classify IM faults based on vibration signal analysis using statistical data feature extraction [ 14 ]. Convolutional neural networks plays an important role in artificial intelligence applications, such as speech recognition [ 15 ] and action recognition [ 16 ]. Bearing faults and broken rotor bar fault detection in squirrel cage IM using the dilated convolutional neural network is presented in [ 17 ]. ANFIS is applied for the detection and classification of IM faults [ 18 ]. Detection and classification of combined inter turn short circuit and broken rotor bars faults are verified using ANFIS [ 19 ]. IM faults diagnosis depend on Machine Learning (ML) algorithms are mostly investigated [ 20 , 21 ]. A review of the ML algorithms in induction motors fault detection is presented in [ 22 ]. The stator current analysis has been considered as one of the most popular fault diagnostic techniques to detect the common faults in electrical rotating machines [ 23 ]. Broken rotor bars fault in IM are detected using spectra analysis of the stator current [ 24 ]. The most public technique of monitoring ITSC fault is stator current signals analysis. However, most of this technique gives reasonable results without data pre-processing [ 25 ]. The time domain features, auto-regressive model and discrete wavelet transform are pre-processing techniques which are used in this paper [ 21 , 26 – 28 ]. The pre-processing technique has an important role in reducing the large amount of information contained in the signal to some features that reflect the overall characteristics of the signal [ 21 , 26 ]. Random forest based on the extracted time domain features from the startup transient current signal are used to determine broken rotor bar fault [ 21 ]. Fuzzy-based time domain feature extraction from the air gap disturbances are used to diagnose broken rotor bars fault in large induction motors [ 26 ]. The Discrete Wavelet Transform (DWT) is used for the Park’s vector modulus of current signals [ 28 ]. The DWT has proven to be is a very effective and reliable technique for diagnosing broken rotor bar faults [ 29 ]. Continuous wavelet transform, discrete wavelet transform, wavelet packet transform and second generation wavelet transform for rotary machines fault diagnosis are studied in [ 30 ]. The wavelet transform is used for feature extraction, while the ANN is used for decision making and classification of winding faults in windmill generators [ 31 ]. The discrete wavelet energy ratio and neural networks are accurate and robust techniques to diagnose the ITSC fault [ 32 ]. Broken rotor bars fault detection in a three-phase squirrel cage IM is proposed based on harmonic analysis of fault components using adaptive notch filter and discrete wavelet transform [ 33 ]. Artificial neural network and support vector machine are used for planetary gearbox fault diagnosis based on DWT feature extraction [ 34 ]. The stationary wavelet packet transform and multiclass wavelet support vector machines are used to diagnose broken bar fault [ 35 ]. A multiscale feature extraction based on spectral graph wavelet transform combined with improved random forest are used to diagnose hob fault [ 36 ]. The Auto-Regressive (AR) model is effective pre-processing technique to diagnose bearing failure [ 27 ]. The integration of the DWT, AR model and principal component analysis are developed for gear multi-fault diagnosis [ 37 ]. In this work, Several experimental tests are discussed which are implemented on a three-phase IM with different fault conditions: 2%, 5%, 7% and 10% ITSC faults at different loading conditions: 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90% and 100% of load. The ITSC faults diagnostic methods are based on the stator current signature analysis. The three-phase stator currents are converted to a stationary axis using the Clark’s Transformation method to improve the diagnostic possibilities of induction motor ITSC faults. These signals are pre-processed by time domain features, auto-regressive model and discrete wavelet transform. ANN, Decision Tree (DT), K-Nearest Neighbours (KNN), Naive Bayes (NB), Random Forest (RF) and Support Vector Machine (SVM) techniques are proposed to diagnose the ITSC faults based on data pre-processed using 13 time domain features extraction. The most important time domain features were selected to diagnose the ITSC faults using ANN, DT, KNN, NB, RF and SVM techniques. The performance of ANN was compared with ANFIS based on time domain features, AR model and DWT. The proposed method achieves higher accuracy and gives better resolution. It is able to diagnose different states of faults with satisfying accuracy. The rest of this paper is organized as follows: Section 1 gives an introduction while Section 2 present intelligent techniques. Section 3 illustrates the research method. Section 4 discusses the experimental setup. Section 5 explains the results and discussion. Section 6 concludes the article. 2. INTELLIGENT TECHNIQUES ANNs learn to recognize certain patterns and give the correct output response to these patterns. Neural network learning methods can be divided into two types supervised and unsupervised learning. The back propagation is most commonly used to train ANN. Multi-layer perceptron can function efficiently with non-linear data while the accuracy of single layer perceptron decreases significantly. So Multi-Layer perceptron is better for diagnose of IM ITSC faults. ANFIS is a combination of neural networks and fuzzy inference system to introduce the learning ability to the fuzzy system. ANFIS uses back-propagation or a combination of least square estimation and back-propagation for membership function parameter estimation. DT is a dendritic classification model used both classification and regression problems. The classification is performed by the breakdown of data into smaller subsets and it is mainly based on the feature selection. KNN is non-parametric, versatile, and lazy learning algorithm used for both classification and regression problems. KNN performs classification of testing data based on the k-nearest training samples round the test data. NB is based on conditional probability with the independence assumption of attributes. It is suitable for continuous, discrete, and categorical features data sets. NB mainly classified into three types; Multinomial Naive, Bernoulli Naive, and Gaussian Naive. RF is a classification method with majority rule using results of plural DTs. Number of DTs is constructed, and the class function is established. The output is concerned with majority of the voting and the final class is declared. SVM was originally introduced to the classification of linearly separable classes of object. SVM can also effectively define a non-linear kernel classification and map its inputs into spaces for high dimensions [ 18 – 22 , 34 – 36 ]. 3. RESEARCH METHOD The intelligent diagnosis procedure begins with the act of data collection of the three-phase stator currents of an IM has different faults conditions: 2%, 5%, 7% and 10% ITSC faults at a different load conditions: 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90% and 100% of load conditions Fig. 1 . These currents are converted to the qd signals. Firstly the current signals are pre-processed using 13 time domain features and fed to ANN, DT, KNN, NB, RF and SVM techniques. Secondly these signal are pre- processed using the most important time domain features and applied to ANN, DT, KNN, NB, RF and SVM techniques to diagnose ITSC faults. Thirdly auto-regressive model and discrete wavelet transform are used to pre-process the current signals then fed to ANN and ANFIS to diagnose ITSC faults. The test accuracy and the Area Under Curve (AUC) of the Receiver Operating Characteristic (ROC) curve is used to evaluate the performance of the classifier. A ROC curve is the resulting true positive rate (Sensitivity) against the false positive rate (Specificity) for different thresholds. If the ROC curve is more to the upper left corner, the classifier performance is better [ 38 ]. The data pre-processing using features extraction gives better results in ITSC faults analysis. It is clear, that the classifier models were improved when these models were provided with more information about the training data. So, the data pre-processing methods should be chosen before applying the training in classification models. 4. EXPERIMENTAL SETUP The experimental tests is implemented using a 1.5 Hp/380 V three-phase squirrel cage induction motor. The electrical specification of the IM are represented in Table 1 . Figure 2 shows the three-phase Squirrel Cage Induction Motor (SC-IM) is coupled to a DC generator. The generator is supplied by a DC voltage source. The variation of the motor load is achieved by the variation of the resistance connected to the generator by a selector switch that is designed in a printed circuit board. The motor is supplied directly by a balanced three-phase sinusoidal voltage source. The stator windings are 348 turns per phase. The motor is equipped with specific access points to diverse turns of stator winding to achieve different cases of faults. They are arranged as 7, 17, 24, and 35 turns in phase "a" that represents 2%, 5%, 7%, and 10% of turns per stator phase, respectively. Figure 3 gives the ITSC faults schematic diagram. The choice of this number of shorted turns is imposed using switches on the printed circuit board as provided in Fig. 2 . Table 1 Tested squirrel cage IM specification IM specifications Unit Value Power HP 1.5 Voltage Volt 380 Rated current Amp 2.8 Rated speed RPM 1400 Frequency Hz 50 Number of turns per phase -- 348 In order to carry out the different experimental tests, the Current/Voltage isolator and an Oscilloscope are connected to measure the three-phase stator currents. Several measurements were performed in which the three-phase stator currents waveform were acquired for a healthy stator winding and for the IM with the different number of shortened coils under different cases of load conditions. The three-phase stator currents are converted to qd signal currents. 5. RESULTS AND DISCUSSIONS 5.1. ITSC Fault Diagnosis based on Time Domain Feature The stator current signals are obtained in different states of fault; 2%, 5%, 7% and 10% ITSC at no load. The three-phase stator current are converted to the qd signals using Clark’s Transformation. There are 125000 samples for each signal. Each signal was divided into 40 segments of 3125 samples as described in Table 2 . These signals are pre-processed using 13 feature extraction parameters. These parameters are Mean (X Mean ), Root Sum of Squares (X RSS ), Root Mean Square value (X RMS ), Peak-to-Peak value (X PP ), Crest Factor (X CF ), Impulse Factor (X IF ), Shape Factor (X SF ), Margin Factor (X MF ), Peak to Average Power ratio (X PAP ), Energy (X E ), Variance (X V ), Skewness Value (X SV ) and Kurtosis Value (X KV ) [ 21 , 26 , 39 ]. The extracted features are the data set of ANN model. Table 2 Data set description Condition Class label Total samples Samples segments Feature extraction samples Data set Training data set Testing data set Healthy 0 125000 40 x 3125 40 200 140 60 2% ITSC Fault 1 125000 40 x 3125 40 5% ITSC Fault 2 125000 40 x 3125 40 7% ITSC Fault 3 125000 40 x 3125 40 10% ITSC Fault 4 125000 40 x 3125 40 The data set is split randomly into training and testing. The training data set is 70%, while the testing data set is 30%. The ANN model is given as Fig. 4 . ANN algorithm has been designed with python programing language. MLPClassifier function is used from the sklearn.neural_network. MLPClassifier, which means the Multi-layer Perceptron classifier. The hidden layers are three, the activation is 'tanh' and the weight optimization solver is 'adam'. The number of epochs that used for training is 1500. The input data is normalized in range from 0 to 1 as defined as: \(\:\text{n}\text{o}\text{r}\text{m}\_\text{d}\text{a}\text{t}\text{a}=(\text{x}-\text{m}\text{i}\text{n}(\text{x}\left)\right)/(\text{max}\left(\text{x}\right)-\text{min}\left(x\right))\) (1)(1) where x(t) is the original input data and norm_data is the data after normalization. The testing accuracy and Cross Validation (CV) metric are employed to evaluate the performance of the classifier. The ROC curve including its AUC metric are used to assessed the classifier. The CV accuracy is found to be 95%, testing accuracy is 96.67%, CV AUC is 0.97, and testing AUC is 0.98. This experimental is done with DT, KNN, NB, RF, and SVM techniques. In DT, the CV accuracy is 90%, testing accuracy is 93%, CV AUC is 0.94, and testing AUC is 0.95. While in KNN, the CV accuracy is 93.33%, testing accuracy is 91.67%, CV AUC is 0.96, and testing AUC is 0.95. The CV accuracy of NB is indicated to be 45%, its testing accuracy is 40%, CV AUC is 0.67, and testing AUC is 0.62. The CV accuracy is found to be 92%, testing accuracy is 95%, CV AUC is 0.95, and testing AUC is 0.97 in RF. SVM realized CV accuracy of 81.67%, testing accuracy is 84%, CV AUC is 0.89, and testing AUC is 0.90. Table 3 and Fig. 5 display the diagnosis results using the ML techniques. The ANN and RF achieve better performance. The ANN obtains the highest scores of CV accuracy, testing accuracy, CV AUC, and testing AUC. NB performs lower performance which obtains the lowest scores of CV accuracy, testing accuracy, CV AUC, and testing AUC. Figure 6 shows a testing ROC curve for the evaluation of ANN, DT, KNN, NB, RF, and SVM techniques. Table 3 Evaluation of ML classifiers based on time domain feature Classifier CV accuracy Testing Accuracy CV AUC Testing AUC ANN 95 96.67 0.97 0.98 DT 90 93 0.94 0.95 KNN 93.33 91.67 0.96 0.95 NB 45 40 0.67 0.62 RF 92 95 0.95 0.97 SVM 81.67 84 0.89 0.90 5.2. ITSC Fault Diagnosis based on the Most Important Features The most important features parameter are Root Mean Square value (X RMS ), Crest Factor (X CF ), Peak-to-Peak Value (X PP ), Impulse Factor (X IF ), Energy (X E ) and Kurtosis Value (X KV ) [ 19 ]. These parameters were taken for pre-processing the qd signals. The ANN model is designed for ITSC faults diagnosis as shown in Fig. 7 . The experimental is implemented on IM different states of fault 2%, 5%, 7% and 10% ITSC at No load. The ANN is provided with 12 inputs during the ITSC faults diagnosis process. The inputs are 6 features obtained from Iq signal and 6 features obtained from Id signal. The dimension of the data set is 12 × 200. The data set is split randomly into subsets of training and testing. The ratio of each subset is defined as 70% and 30%. This experimental is performed with DT, KNN, NB, RF and SVM algorithms. The cross validation accuracy, testing accuracy and AUC metric are determined to evaluate ANN, DT, KNN, NB, RF, and SVM techniques. The comparison between ML algorithms is indicated in Table 4 and Fig. 8. It was observed that the ANN and RF provide the highest performance. The ANN achieves CV accuracy of 99%, testing accuracy of 100%, CV AUC of 1 and testing AUC of 1 while RF provides CV accuracy of 99%, testing accuracy of 100%, CV AUC of 0.99, and testing AUC of 1. NB obtains the worst results that gets the lowest CV accuracy of 47%, testing accuracy of 45%, CV AUC of 0.67, and testing AUC of 0.67. Other classifiers provide CV accuracy higher than 84%, testing accuracy higher than 85%, CV AUC scores higher than 0.90, and testing AUC scores higher than 0.90. Moreover, it was proven that classifiers achieve high performance when the important features are used as compared with the classifiers using all time domain features as displayed in Table 3 and Table 4 . Therefore, the important features will be used in the following experimental due to its high accuracy. The ROC curve is presented in Fig. 9. It was observed that the ROC curve of ANN and RF are more to the upper left corner. Table 4 Evaluation of ML classifiers based on the important features Classifier CV accuracy Testing accuracy CV AUC Testing AUC ANN 99 100 1 1 DT 92 98.33 0.95 0.99 KNN 98 96.67 0.99 0.98 NB 47 45 0.67 0.67 RF 99 100 0.99 1 SVM 84 85 0.90 0.90 5.3. ITSC Fault Diagnosis using ANN and RF based on the Most Important Features Several experimental test are implemented on the IM at a loading condition of: 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90% and 100%. Different conditions of stator winding were tested: healthy, 2%, 5%, 7%, and 10% ITSC. The important features obtained from the qd signals are used to generate the data set. These data sets are used to train and test ANN and RF models. The performance of the classifier is evaluated by the CV accuracy, testing accuracy, and scores of the AUC metric. Table 5 and Fig. 10 illustrate the diagnosis results using the ANN compared with RF. ANN achieves the highest overall CV accuracy of 96.30%, overall testing accuracy of 96.50%, overall CV AUC of 0.97, and overall testing AUC of 0.98. The RF obtains the lowest overall CV accuracy of 93.30%, overall testing accuracy of 94.17%, overall CV AUC of 0.95, and overall testing AUC of 0.96. Hence, it can be concluded that the ANN based statistical feature extraction gives better performance as compared to all others approaches for ITSC fault diagnosis. Table 5 Evaluation of ANN and RF classifiers based on the important features Loading Condition ANN RF CV accuracy Testing accuracy CV AUC Testing AUC CV accuracy Testing accuracy CV AUC Testing AUC 10% 93 90 0.94 0.94 89 86.67 0.91 0.91 20% 96 91.67 0.97 0.95 90 91.67 0.93 0.95 30% 98 100 0.97 1 97 100 0.97 1 40% 97 100 0.97 1 95 95 0.95 0.97 50% 97 100 0.98 1 93 93.33 0.95 0.96 60% 97 98.33 0.97 0.99 97 98.33 0.97 0.99 70% 93 95 0.95 0.97 88 90 0.94 0.94 80% 98 96.67 0.97 0.98 93 95 0.96 0.97 90% 97 96.67 0.98 0.98 97 100 0.99 1 100% 97 96.67 0.98 0.98 94 91.67 0.96 0.95 Overall 96.30 96.50 0.97 0.98 93.30 94.17 0.95 0.96 5.4. Comparison between ANN and ANFIS The tests are performed on the IM with different fault conditions: 2%, 5%, 7%, and 10% ITSC faults at a full load. The algorithms have been designed in Matlab software. In this work, the input data is normalized in range from 0 to 1. The normalized data is pre-processed using two methods, the time domain features and the auto-regressive model. 5.4.1. Comparison between ANN and ANFIS based on Time Domain Feature The ANN and ANFIS models for classification relying on the most important features is performed. The important features are obtained from the qd signals. These features are used to classify ITSC faults conditions. The performance of classifiers is assisted by testing RMSE as shown in Table 6 . It is observed that the ANN achieves better performance than the ANFIS. Table 7 Evaluation of ANN and ANFIS based on time domain feature Classifier RMSE ANN 0.2326 ANFIS 0.3442 5.4.2. Comparison between ANN and ANFIS based on AR Model The auto-regressive model is used for pre-processing the qd signals to ANN and ANFIS models. The AR model depends on extracting the hidden features of the signal with attenuating the noise as possible via the auto-correlation method. The coefficients of the AR model is calculated by linear prediction coding. The number of parameters of the AR model is five, which is considered the features of the AR model. Testing Root Mean Square Error (RMSE) is developed to evaluate the performance of the classifiers as indicated in Table 7 . It was observed that the ANFIS obtains better accuracy than the ANN. Table 6 Evaluation of ANN and ANFIS based on AR Model Classifier RMSE ANN 0.2638 ANFIS 0.0244 5.4.3. Comparison between ANN and ANFIS based on DWT The normalized data of qd signals is pre-processed using DWT. The mother wavelet "haar" is selected and the maximum decomposition level was obtained. The data set generated from DWT is applied to ANFIS and ANN. The testing data set is utilized to assess the efficiency of ANFIS and ANN using RMSE score as given in Table 8 . It was observed that ANFIS achieves higher performance than ANN. Table 8 Evaluation of ANN and ANFIS based on DWT Classifier RMSE ANN 4.9393 × 10 − 4 ANFIS 6.4608 × 10 − 9 From the previous discussion, thee different techniques were used for pre-processing the raw qd data before entering the classifier. The pre-processed data was employed as input to help the ANN and ANFIS training. Thus, the different pre-processing data aim to build the network model in the best possible structure. It is clear, that the DWT as pre-processing for input data to ANFIS gives beater results compared to ANN. On the other hand, the time domain feature as pre-processing for input data to ANN achieves good results compared to ANFIS. It is clear, the pre-processing techniques play an important part in ITSC faults diagnosis. ANN based on the most important time domain feature provides better performance than all other techniques. The worst output is obtained by NB based on time domain feature. ANN based on auto-regressive model gives weak performance. ANFIS performing well based on DWT. Therefore, the pre-processing of input data had an important role in classification issues where it was revealed in the performance of the selected model for diagnosing the ITSC faults in IM. So, before applying the training in classification models, the appropriate pre-processing methods for the input data to classifier models should be chosen. On the other hand, data pre-processing using features extraction gives distinct results in induction motor faults analysis. 6. CONCLUSION This paper presents a new application of ANN, ANFIS, DT, KNN, NB, RF, and SVM techniques to diagnose the induction motor ITSC faults. The motor is exposed to different ITSC faults under different loading conditions. The input data was pre-processed before entering to ANN, DT, KNN, NB, RF, and SVM models by time domain features extraction. The experimental was performed to diagnose ITSC faults using ANN, DT, KNN, NB, RF, and SVM techniques based on time domain features. The analysis of the results establishes that ANN archives the highest accuracy while the NB is the worst classifier. The experimental was implemented using ANN and ANFIS techniques based on time domain features and AR models and DWT. The evaluation analysis of the results proves that the ANN obtains the best results based on the time domain features as pre-processing for input data compared to ANFIS. On the other hand, the ANFIS obtains the beater results based on the DWT model as pre-processing for input data compared to ANN. The effectiveness of using the ANN and ANFIS based on data pre-processing to diagnose the ITSC faults in three-phase IM is investigated using experimental tests. In the future researches, some works could be considered as follows: Other types of faults such as bearing failure, phase-to-phase, open-phase and phase-to-ground faults may be considered based on data pre-processing. Declarations Author Contribution M.A.M. and M.A.M.H. designed the problem under study; M.A.M. performed the simulations and obtained the results; M.A.M.H. analyzed the obtained results; M.A.M. wrote the paper, which was further reviewed by M.A.M.H.All authors have read and agreed to the published version of the manuscript. References Konar, P., & Chattopadhyay, P. (2015). Multi-class fault diagnosis of induction motor using Hilbert and Wavelet Transform. Applied Soft Computing, 30, 341-352.‏ Menshawy A. M., E. Mohamed, A.-A. Mohamed, M. Abdel-Nasser, and M. A. Moustafa Hassan, “Detection of inter turn short circuit faults in induction motor using artificial neural network,” in 2020 26th Conference of Open Innovations Association (FRUCT), 2020, pp. 297-304. Menshawy A. M., E. Mohamed, A.-A. Mohamed, M. Abdel-Nasser, and M. A. Moustafa Hassan, “Induction motor broken rotor bar faults diagnosis using ANFIS-based DWT,” Int. J. Model. Simul. , pp. 1–14, 2019. S. Bensaoucha, S. A. Bessedik, A. Ameur, and A. Teta, “Induction motors broken rotor bars detection using RPVM and neural network,” COMPEL , vol. 38, no. 2, pp. 596–615, 2019. Almounajjed, A., Sahoo, A. K., & Kumar, M. K. (2021). Diagnosis of stator fault severity in induction motor based on discrete wavelet analysis. Measurement, 182, 109780.‏ M. Eftekhari, M. Moallem, S. Sadri, and M.-F. Hsieh, “A novel indicator of stator winding inter-turn fault in induction motor using infrared thermal imaging,” Infrared Phys. Technol. , vol. 61, pp. 330–336, 2013. Discrete wavelet transform and energy eigen value for rotor bars fault detection in variable speed field-oriented control of induction motor drive Y. Chen, X. Chen, and Y. Shen, “On-line detection of coil inter-turn short circuit faults in dual-redundancy permanent magnet synchronous motors,” Energies , vol. 11, no. 3, p. 662, 2018. M. S. Khadim, S. Kuldeep and V. K. Giri, “Diagnosis of Stator inter-turn fault in pwm inverter fed induction motor by advanced dsp technique,” i-manag. J. Electr. Eng. , vol. 12, no. 3, p. 9, 2019. W. Zhang, M.-P. Jia, L. Zhu, and X.-A. Yan, “Comprehensive overview on computational intelligence techniques for machinery condition monitoring and fault diagnosis,” Chin. J. Mech. Eng. , vol. 30, no. 4, pp. 782–795, 2017. N. Ngote, and M. Ouassaid, “A Hybrid TSA-Fuzzy Logic Approach to Detect Induction Motor Rotor Faults,” International Journal of Information Science and Technology , vol. 3, no. 1, pp. 26-35, 2019. L. S. Maraaba, Z. M. Al-Hamouz, A. S. Milhem, and M. A. Abido, “Neural network-based diagnostic tool for detecting Stator inter-turn faults in line start permanent magnet synchronous motors,” IEEE Access , vol. 7, pp. 89014–89025, 2019. M. Skowron, M. Wolkiewicz, T. Orlowska-Kowalska, and C. T. Kowalski, “Effectiveness of selected neural network structures based on axial flux analysis in Stator and rotor winding incipient fault detection of inverter-fed induction motors,” Energies , vol. 12, no. 12, p. 2392, 2019. R. F. Ribeiro Junior, F. A. de Almeida, and G. F. Gomes, “Fault classification in three-phase motors based on vibration signal analysis and artificial neural networks,” Neural Comput. Appl. , vol. 32, no. 18, pp. 15171–15189, 2020. W. Jing, T. Jiang, X. Zhang, and L. Zhu, “The optimisation of speech recognition based on convolutional neural network,” Int. j. high perform. comput. netw. , vol. 13, no. 2, p. 222, 2019. (15)J. Li, X. Ban, G. Yang, Y. Li, and Y. Wang, “Real-time human action recognition using depth motion maps and convolutional neural networks,” Int. j. high perform. comput. netw. , vol. 13, no. 3, p. 312, 2019. Kumar, P., & Hati, A. S. (2022). Dilated convolutional neural network based model for bearing faults and broken rotor bar detection in squirrel cage induction motors. Expert Systems with Applications, 191, 116290.‏ Menshawy A. M., Mahmoud, A. S., Essam, H. A. and M. A. Moustafa Hassan, "Detection and Classification of Broken Rotor Bars Faults in Induction Motor Using Adaptive Neuro-Fuzzy Inference System," Proceedings of MEPCON ‘14, Cairo, Egypt , 2014. Menshawy A. Mohamed, M. A. Moustafa. Hassan, Mostafa Dardeer and et al., "Diagnostic Modelling for Induction Motor Faults via ANFIS Algorithm and DWT-Based Feature Extraction," Applied Sciences. Vol. 11, No. 19, pp. 1-18, 2021.‏ S. Esakimuthu Pandarakone, Y. Mizuno, and H. Nakamura, “A comparative study between machine learning algorithm and artificial intelligence neural network in detecting minor bearing fault of induction motors,” Energies , vol. 12, no. 11, p. 2105, 2019. J. C. Quiroz, N. Mariun, M. R. Mehrjou, M. Izadi, N. Misron, and M. A. Mohd Radzi, “Fault detection of broken rotor bar in LS-PMSM using random forests,” Measurement (Lond.) , vol. 116, pp. 273–280, 2018. P. Kumar and A. S. Hati, “Review on machine learning algorithm based fault detection in induction motors,” Arch. Comput. Methods Eng. , 2020. I. O. Zaparoli, L. M. Rabelo Baccarini, P. C. M. Lamim Filho, and F. B. Batista, “Transient envelope current analysis for inter-turn short-circuit detection in induction motor stator,” J. Braz. Soc. Mech. Sci. Eng. , vol. 42, no. 2, 2020. Ebrahimi, B. M., Faiz, J., Lotfi-Fard, S., & Pillay, P. (2012). Novel indices for broken rotor bars fault diagnosis in induction motors using wavelet transform. Mechanical Systems and Signal Processing, 30, 131-145.‏ Mostafa M., Galal H. and Mohamed A. Moustafa Hassan, “Diagnosis of Rotating Machines Faults Using Artificial Intelligence Based on Preprocessing for Input Data,” in 2020 26th Conference of Open Innovations Association (FRUCT), 2020, pp. 572-582. C. G. Dias, L. C. da da Silva, and I. E. Chabu, “Fuzzy-based statistical feature extraction for detecting broken rotor bars in line-fed and inverter-fed induction motors,” Energies , vol. 12, no. 12, p. 2381, 2019. Mohamed A. Moustafa Hassan, “Effect of Pre-processing on Using ANN and ANFIS,” in 2020 26th Conference of Open Innovations Association (FRUCT), 2020, pp. 589-603. D. M. Sonje, P. Kundu, and A. Chowdhury, “A novel approach for sensitive inter-turn fault detection in induction motor under various operating conditions,” Arab. J. Sci. Eng. , vol. 44, no. 8, pp. 6887–6900, 2019. Talhaoui, H., Menacer, A., Kessal, A., & Kechida, R. (2014). Fast Fourier and discrete wavelet transforms applied to sensorless vector control induction motor for rotor bar faults diagnosis. ISA transactions, 53(5), 1639-1649.‏ Yan, R., Gao, R. X., & Chen, X. (2014). Wavelets for fault diagnosis of rotary machines: A review with applications. Signal processing, 96, 1-15.‏ Gketsis, Z. E., Zervakis, M. E., & Stavrakakis, G. (2009). Detection and classification of winding faults in windmill generators using Wavelet Transform and ANN. Electric Power Systems Research, 79(11), 1483-1494.‏ Cherif, H., Benakcha, A., Laib, I., Chehaidia, S. E., Menacer, A., Soudan, B., & Olabi, A. G. (2020). Early detection and localization of stator inter-turn faults based on discrete wavelet energy ratio and neural networks in induction motor. Energy, 212, 118684.‏ Taher, S. A., Malekpour, M., & Farshadnia, M. (2014). Diagnosis of broken rotor bars in induction motors based on harmonic analysis of fault components using modified adaptive notch filter and discrete wavelet transform. Simulation Modelling Practice and Theory, 44, 26-41. ‏ Syed, S. H., & Muralidharan, V. (2022). Feature extraction using Discrete Wavelet Transform for fault classification of planetary gearbox–A comparative study. Applied Acoustics, 188, 108572. ‏ Keskes, H., Braham, A., & Lachiri, Z. (2013). Broken rotor bar diagnosis in induction machines through stationary wavelet packet transform and multiclass wavelet SVM. Electric Power Systems Research, 97, 151-157. ‏ Dong, X., Li, G., Jia, Y., & Xu, K. (2021). Multiscale feature extraction from the perspective of graph for hob fault diagnosis using spectral graph wavelet transform combined with improved random forest. Measurement, 176, 109178. ‏ Li, Z., Yan, X., Yuan, C., Peng, Z., & Li, L. (2011). Virtual prototype and experimental research on gear multi-fault diagnosis using wavelet-autoregressive model and principal component analysis method. Mechanical Systems and Signal Processing, 25(7), 2589-2607. ‏ I. Martin-Diaz, D. Morinigo-Sotelo, O. Duque-Perez, and R. J. Romero-Troncoso, “An experimental comparative evaluation of machine learning techniques for motor fault diagnosis under various operating conditions,” IEEE Trans. Ind. Appl. , vol. 54, no. 3, pp. 2215–2224, 2018. J. Patel, V. Patel,and A. Patel, "Fault diagnostics of rolling bearing based on improve time and frequency domain features using artificial neural networks," International Journal for Scientific Research & Development , vol. 1, no. 4, pp. 781-788, 2013. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7242428","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":492983400,"identity":"528af254-2757-4ec5-9621-0070e8cd4462","order_by":0,"name":"Menshawy Mohamed","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA2UlEQVRIiWNgGAWjYJACZgjFw/iAgeEAaVqYDUjWwiZBlBbd9t7DnwtqDieube89Vs1Tc0eOn4H54aMbeLSYnTmXJj3j2OHEbUDGbZ5jz4wlG9iMjXPwabmRY8bMw3Y7cRuQcZuH7XDihgM8bNJ4tdx/Y/yZ5x9ESzHPP2K03OAxkOZtg2hh5m0jRsuZHDNp3r7/xtvOnDGWnNt32FiymZBfjp8BOuxbmuy24z2GH958OyzHz9788DE+LTDg2AAkmHhATGYilIOAPYhg/EGk6lEwCkbBKBhZAAAtMFNAg/609gAAAABJRU5ErkJggg==","orcid":"","institution":"University of Calabria","correspondingAuthor":true,"prefix":"","firstName":"Menshawy","middleName":"","lastName":"Mohamed","suffix":""},{"id":492983402,"identity":"8113bff3-438c-4662-be41-aaac80bd7a45","order_by":1,"name":"Mohamed Moustafa","email":"","orcid":"","institution":"Cairo University","correspondingAuthor":false,"prefix":"","firstName":"Mohamed","middleName":"","lastName":"Moustafa","suffix":""}],"badges":[],"createdAt":"2025-07-29 10:53:27","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7242428/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7242428/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":88390617,"identity":"d4fa00a2-0693-4f2d-a330-cd19b4c994a2","added_by":"auto","created_at":"2025-08-06 04:22:52","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":100089,"visible":true,"origin":"","legend":"\u003cp\u003eFlow graph of ITSC fault diagnosis procedure\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-7242428/v1/38e57a6bd71d1eebb469617d.png"},{"id":88391273,"identity":"14bb3e42-cda9-4b32-ba66-8ffb9dd62b49","added_by":"auto","created_at":"2025-08-06 04:30:52","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":487534,"visible":true,"origin":"","legend":"\u003cp\u003eExperimental setup\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-7242428/v1/06ce668e8024a6ee2d8019aa.png"},{"id":88390632,"identity":"280b943a-bc78-49ed-a5a1-1f846103cfe2","added_by":"auto","created_at":"2025-08-06 04:22:53","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":52922,"visible":true,"origin":"","legend":"\u003cp\u003eStator ITSC faults\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-7242428/v1/6bb21876666c61f50ee97ee2.png"},{"id":88390618,"identity":"806b51b3-5947-42b7-b150-4be4126c87d3","added_by":"auto","created_at":"2025-08-06 04:22:52","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":81176,"visible":true,"origin":"","legend":"\u003cp\u003eANN model classifier\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-7242428/v1/b7dc38f38a064d4614972ad4.png"},{"id":88390623,"identity":"602c1c59-4483-40c4-b03f-7277da6ea97d","added_by":"auto","created_at":"2025-08-06 04:22:53","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":55824,"visible":true,"origin":"","legend":"\u003cp\u003eEvaluation Diagram of ML Classifiers (a) Accuracy (b) AUC\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-7242428/v1/a40f54694e875e75509586b3.png"},{"id":88390627,"identity":"b445032e-7e8d-4181-bb0f-377ba5ca9ca7","added_by":"auto","created_at":"2025-08-06 04:22:53","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":298404,"visible":true,"origin":"","legend":"\u003cp\u003eTesting ROC curve for different ITSC faults (a) ANN (b) DT (c) KNN (d) NB (e) RF (f) SVM\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-7242428/v1/5cdc71a5893d260483cbcdab.png"},{"id":88390640,"identity":"061140e7-b28e-4515-b0bf-cf5e1d858cca","added_by":"auto","created_at":"2025-08-06 04:22:53","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":75034,"visible":true,"origin":"","legend":"\u003cp\u003eANN model based on the important features\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-7242428/v1/8cd4e23d1a0df875019be623.png"},{"id":88390633,"identity":"ce5b7d50-1eec-414a-8a59-05630a291c1a","added_by":"auto","created_at":"2025-08-06 04:22:53","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":50455,"visible":true,"origin":"","legend":"\u003cp\u003eEvaluation diagram of ML classifiers based on the important features (a) Accuracy \u0026nbsp;\u0026nbsp;(b) AUC\u003c/p\u003e","description":"","filename":"8.png","url":"https://assets-eu.researchsquare.com/files/rs-7242428/v1/fc829219a1c8fba7819ee24f.png"},{"id":88391792,"identity":"7a2ef8a4-a16c-49be-97e3-ffd1ffe41656","added_by":"auto","created_at":"2025-08-06 04:38:53","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":240890,"visible":true,"origin":"","legend":"\u003cp\u003eTesting ROC curve for different ITSC faults based on the important features (a) ANN (b) DT (c) KNN (d) NB (e) RF (f) SVM\u003c/p\u003e","description":"","filename":"9.png","url":"https://assets-eu.researchsquare.com/files/rs-7242428/v1/8588fa68a3e6b0e5ba34e7df.png"},{"id":88390641,"identity":"b2587ace-5a1f-4d08-a1ed-205d3ffc0f54","added_by":"auto","created_at":"2025-08-06 04:22:53","extension":"png","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":94234,"visible":true,"origin":"","legend":"\u003cp\u003eEvaluation diagram of ANN and RF classifiers based on the important features (a) Accuracy (b) AUC\u003c/p\u003e","description":"","filename":"10.png","url":"https://assets-eu.researchsquare.com/files/rs-7242428/v1/cb43f99ba966673f161c9306.png"},{"id":89217053,"identity":"995a2efa-bd2f-46cb-82ac-84f97cfc042b","added_by":"auto","created_at":"2025-08-17 07:01:31","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2436839,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7242428/v1/229e29e4-5f92-4884-be30-2713bb66f519.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Stator Winding Faults Diagnosis in Induction Motor Based on ANN and ANFIS","fulltext":[{"header":"1. INTRODUCTION","content":"\u003cp\u003eElectrical motors are widely utilized in manufacturing cycle, since it represents an important part in industrial processes because of its simple structure, reliability, ruggedness, cost effective design and ease of control [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. The most common Induction Motor (IM) faults are short circuits winding or opening in the stator phase, broken rotor bar and bearing failures [\u003cspan additionalcitationids=\"CR3\" citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. ITSC fault represents between 21% and 40% of the machine faults [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. Due to high current flow in the short circuited coils and the decreased insulation the stator inter turn defect is occurred [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. The breakdown of stator winding insulation leads to a short circuit fault. It takes very short time for short circuitry in the stator winding to damage the motor. This will break production in the industry. IM fault detection improves reliability and availability of an existing system at an early stage [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eRecently many researchers have been concentrated on IM faults. Many fault detection techniques were illustrated. The stator winding inductance analysis and the ITSC defect diagnostic method in a several operating circumstances are proposed in [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. Wavelet transformation algorithms and advanced digital signal processing technique have detected the fault in ITSC in induction motor [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Fuzzy Logic, genetic algorithms, ANN, and ANFIS are high potential data processing tool that creates fault diagnosis techniques [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. ANN and ANFIS techniques are used to increase the accuracy for diagnosing of IM faults and overcome the drawbacks of the traditional techniques. ANN is a powerful tool that has been suggested for IM fault diagnosis [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. ANN is used to classify IM faults based on vibration signal analysis using statistical data feature extraction [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. Convolutional neural networks plays an important role in artificial intelligence applications, such as speech recognition [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e] and action recognition [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. Bearing faults and broken rotor bar fault detection in squirrel cage IM using the dilated convolutional neural network is presented in [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. ANFIS is applied for the detection and classification of IM faults [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. Detection and classification of combined inter turn short circuit and broken rotor bars faults are verified using ANFIS [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. IM faults diagnosis depend on Machine Learning (ML) algorithms are mostly investigated [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. A review of the ML algorithms in induction motors fault detection is presented in [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. The stator current analysis has been considered as one of the most popular fault diagnostic techniques to detect the common faults in electrical rotating machines [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. Broken rotor bars fault in IM are detected using spectra analysis of the stator current [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. The most public technique of monitoring ITSC fault is stator current signals analysis. However, most of this technique gives reasonable results without data pre-processing [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. The time domain features, auto-regressive model and discrete wavelet transform are pre-processing techniques which are used in this paper [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e, \u003cspan additionalcitationids=\"CR27\" citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. The pre-processing technique has an important role in reducing the large amount of information contained in the signal to some features that reflect the overall characteristics of the signal [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. Random forest based on the extracted time domain features from the startup transient current signal are used to determine broken rotor bar fault [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. Fuzzy-based time domain feature extraction from the air gap disturbances are used to diagnose broken rotor bars fault in large induction motors [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. The Discrete Wavelet Transform (DWT) is used for the Park\u0026rsquo;s vector modulus of current signals [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. The DWT has proven to be is a very effective and reliable technique for diagnosing broken rotor bar faults [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]. Continuous wavelet transform, discrete wavelet transform, wavelet packet transform and second generation wavelet transform for rotary machines fault diagnosis are studied in [\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]. The wavelet transform is used for feature extraction, while the ANN is used for decision making and classification of winding faults in windmill generators [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]. The discrete wavelet energy ratio and neural networks are accurate and robust techniques to diagnose the ITSC fault [\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]. Broken rotor bars fault detection in a three-phase squirrel cage IM is proposed based on harmonic analysis of fault components using adaptive notch filter and discrete wavelet transform [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e]. Artificial neural network and support vector machine are used for planetary gearbox fault diagnosis based on DWT feature extraction [\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e]. The stationary wavelet packet transform and multiclass wavelet support vector machines are used to diagnose broken bar fault [\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]. A multiscale feature extraction based on spectral graph wavelet transform combined with improved random forest are used to diagnose hob fault [\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e]. The Auto-Regressive (AR) model is effective pre-processing technique to diagnose bearing failure [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. The integration of the DWT, AR model and principal component analysis are developed for gear multi-fault diagnosis [\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eIn this work, Several experimental tests are discussed which are implemented on a three-phase IM with different fault conditions: 2%, 5%, 7% and 10% ITSC faults at different loading conditions: 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90% and 100% of load. The ITSC faults diagnostic methods are based on the stator current signature analysis. The three-phase stator currents are converted to a stationary axis using the Clark\u0026rsquo;s Transformation method to improve the diagnostic possibilities of induction motor ITSC faults. These signals are pre-processed by time domain features, auto-regressive model and discrete wavelet transform. ANN, Decision Tree (DT), K-Nearest Neighbours (KNN), Naive Bayes (NB), Random Forest (RF) and Support Vector Machine (SVM) techniques are proposed to diagnose the ITSC faults based on data pre-processed using 13 time domain features extraction. The most important time domain features were selected to diagnose the ITSC faults using ANN, DT, KNN, NB, RF and SVM techniques. The performance of ANN was compared with ANFIS based on time domain features, AR model and DWT. The proposed method achieves higher accuracy and gives better resolution. It is able to diagnose different states of faults with satisfying accuracy.\u003c/p\u003e\u003cp\u003eThe rest of this paper is organized as follows: Section 1 gives an introduction while Section 2 present intelligent techniques. Section 3 illustrates the research method. Section 4 discusses the experimental setup. Section 5 explains the results and discussion. Section 6 concludes the article.\u003c/p\u003e"},{"header":"2. INTELLIGENT TECHNIQUES","content":"\u003cp\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eANNs learn to recognize certain patterns and give the correct output response to these patterns. Neural network learning methods can be divided into two types supervised and unsupervised learning. The back propagation is most commonly used to train ANN. Multi-layer perceptron can function efficiently with non-linear data while the accuracy of single layer perceptron decreases significantly. So Multi-Layer perceptron is better for diagnose of IM ITSC faults. ANFIS is a combination of neural networks and fuzzy inference system to introduce the learning ability to the fuzzy system. ANFIS uses back-propagation or a combination of least square estimation and back-propagation for membership function parameter estimation. DT is a dendritic classification model used both classification and regression problems. The classification is performed by the breakdown of data into smaller subsets and it is mainly based on the feature selection. KNN is non-parametric, versatile, and lazy learning algorithm used for both classification and regression problems. KNN performs classification of testing data based on the k-nearest training samples round the test data. NB is based on conditional probability with the independence assumption of attributes. It is suitable for continuous, discrete, and categorical features data sets. NB mainly classified into three types; Multinomial Naive, Bernoulli Naive, and Gaussian Naive. RF is a classification method with majority rule using results of plural DTs. Number of DTs is constructed, and the class function is established. The output is concerned with majority of the voting and the final class is declared. SVM was originally introduced to the classification of linearly separable classes of object. SVM can also effectively define a non-linear kernel classification and map its inputs into spaces for high dimensions [\u003cspan additionalcitationids=\"CR19 CR20 CR21\" citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e, \u003cspan additionalcitationids=\"CR35\" citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e].\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e"},{"header":"3. RESEARCH METHOD","content":"\u003cp\u003eThe intelligent diagnosis procedure begins with the act of data collection of the three-phase stator currents of an IM has different faults conditions: 2%, 5%, 7% and 10% ITSC faults at a different load conditions: 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90% and 100% of load conditions Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. These currents are converted to the qd signals. Firstly the current signals are pre-processed using 13 time domain features and fed to ANN, DT, KNN, NB, RF and SVM techniques. Secondly these signal are pre- processed using the most important time domain features and applied to ANN, DT, KNN, NB, RF and SVM techniques to diagnose ITSC faults. Thirdly auto-regressive model and discrete wavelet transform are used to pre-process the current signals then fed to ANN and ANFIS to diagnose ITSC faults.\u003c/p\u003e\u003cp\u003eThe test accuracy and the Area Under Curve (AUC) of the Receiver Operating Characteristic (ROC) curve is used to evaluate the performance of the classifier. A ROC curve is the resulting true positive rate (Sensitivity) against the false positive rate (Specificity) for different thresholds. If the ROC curve is more to the upper left corner, the classifier performance is better [\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e]. The data pre-processing using features extraction gives better results in ITSC faults analysis. It is clear, that the classifier models were improved when these models were provided with more information about the training data. So, the data pre-processing methods should be chosen before applying the training in classification models.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e"},{"header":"4. EXPERIMENTAL SETUP","content":"\u003cp\u003eThe experimental tests is implemented using a 1.5 Hp/380 V three-phase squirrel cage induction motor. The electrical specification of the IM are represented in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. Figure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e shows the three-phase Squirrel Cage Induction Motor (SC-IM) is coupled to a DC generator. The generator is supplied by a DC voltage source. The variation of the motor load is achieved by the variation of the resistance connected to the generator by a selector switch that is designed in a printed circuit board. The motor is supplied directly by a balanced three-phase sinusoidal voltage source. The stator windings are 348 turns per phase. The motor is equipped with specific access points to diverse turns of stator winding to achieve different cases of faults. They are arranged as 7, 17, 24, and 35 turns in phase \"a\" that represents 2%, 5%, 7%, and 10% of turns per stator phase, respectively. Figure\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e gives the ITSC faults schematic diagram. The choice of this number of shorted turns is imposed using switches on the printed circuit board as provided in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eTested squirrel cage IM specification\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"3\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eIM specifications\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eUnit\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eValue\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePower\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eHP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.5\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eVoltage\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eVolt\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e380\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRated current\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAmp\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2.8\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRated speed\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eRPM\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1400\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFrequency\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eHz\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e50\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNumber of turns per phase\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e--\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e348\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eIn order to carry out the different experimental tests, the Current/Voltage isolator and an Oscilloscope are connected to measure the three-phase stator currents. Several measurements were performed in which the three-phase stator currents waveform were acquired for a healthy stator winding and for the IM with the different number of shortened coils under different cases of load conditions. The three-phase stator currents are converted to qd signal currents.\u003c/p\u003e"},{"header":"5. RESULTS AND DISCUSSIONS","content":"\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\n \u003ch2\u003e5.1. ITSC Fault Diagnosis based on Time Domain Feature\u003c/h2\u003e\n \u003cp\u003eThe stator current signals are obtained in different states of fault; 2%, 5%, 7% and 10% ITSC at no load. The three-phase stator current are converted to the qd signals using Clark\u0026rsquo;s Transformation. There are 125000 samples for each signal. Each signal was divided into 40 segments of 3125 samples as described in Table \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e. These signals are pre-processed using 13 feature extraction parameters. These parameters are Mean (X\u003csub\u003eMean\u003c/sub\u003e), Root Sum of Squares (X\u003csub\u003eRSS\u003c/sub\u003e), Root Mean Square value (X\u003csub\u003eRMS\u003c/sub\u003e), Peak-to-Peak value (X\u003csub\u003ePP\u003c/sub\u003e), Crest Factor (X\u003csub\u003eCF\u003c/sub\u003e), Impulse Factor (X\u003csub\u003eIF\u003c/sub\u003e), Shape Factor (X\u003csub\u003eSF\u003c/sub\u003e), Margin Factor (X\u003csub\u003eMF\u003c/sub\u003e), Peak to Average Power ratio (X\u003csub\u003ePAP\u003c/sub\u003e), Energy (X\u003csub\u003eE\u003c/sub\u003e), Variance (X\u003csub\u003eV\u003c/sub\u003e), Skewness Value (X\u003csub\u003eSV\u003c/sub\u003e) and Kurtosis Value (X\u003csub\u003eKV\u003c/sub\u003e) [\u003cspan class=\"CitationRef\"\u003e21\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e26\u003c/span\u003e,\u0026nbsp;\u003cspan class=\"CitationRef\"\u003e39\u003c/span\u003e]. The extracted features are the data set of ANN model.\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003ctable id=\"Tab2\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eData set description\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eCondition\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eClass label\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTotal samples\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSamples segments\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eFeature extraction samples\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eData set\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTraining\u003c/p\u003e\n \u003cp\u003edata set\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTesting\u003c/p\u003e\n \u003cp\u003edata set\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHealthy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e125000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e40 x 3125\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e40\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" rowspan=\"5\"\u003e\n \u003cp\u003e200\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" rowspan=\"5\"\u003e\n \u003cp\u003e140\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" rowspan=\"5\"\u003e\n \u003cp\u003e60\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2% ITSC Fault\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e125000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e40 x 3125\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e40\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5% ITSC Fault\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e125000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e40 x 3125\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e40\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e7% ITSC Fault\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e125000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e40 x 3125\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e40\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e10% ITSC Fault\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e125000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e40 x 3125\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e40\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cp\u003eThe data set is split randomly into training and testing. The training data set is 70%, while the testing data set is 30%. The ANN model is given as Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e. ANN algorithm has been designed with python programing language. MLPClassifier function is used from the sklearn.neural_network. MLPClassifier, which means the Multi-layer Perceptron classifier. The hidden layers are three, the activation is \u0026apos;tanh\u0026apos; and the weight optimization solver is \u0026apos;adam\u0026apos;. The number of epochs that used for training is 1500. The input data is normalized in range from 0 to 1 as defined as:\u003c/p\u003e\n \u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\text{n}\\text{o}\\text{r}\\text{m}\\_\\text{d}\\text{a}\\text{t}\\text{a}=(\\text{x}-\\text{m}\\text{i}\\text{n}(\\text{x}\\left)\\right)/(\\text{max}\\left(\\text{x}\\right)-\\text{min}\\left(x\\right))\\)\u003c/span\u003e\u003c/span\u003e (1)(1)\u003c/p\u003e\n \u003cp\u003ewhere x(t) is the original input data and norm_data is the data after normalization.\u003c/p\u003e\n \u003cp\u003eThe testing accuracy and Cross Validation (CV) metric are employed to evaluate the performance of the classifier. The ROC curve including its AUC metric are used to assessed the classifier. The CV accuracy is found to be 95%, testing accuracy is 96.67%, CV AUC is 0.97, and testing AUC is 0.98.\u003c/p\u003e\n \u003cp\u003eThis experimental is done with DT, KNN, NB, RF, and SVM techniques. In DT, the CV accuracy is 90%, testing accuracy is 93%, CV AUC is 0.94, and testing AUC is 0.95. While in KNN, the CV accuracy is 93.33%, testing accuracy is 91.67%, CV AUC is 0.96, and testing AUC is 0.95. The CV accuracy of NB is indicated to be 45%, its testing accuracy is 40%, CV AUC is 0.67, and testing AUC is 0.62. The CV accuracy is found to be 92%, testing accuracy is 95%, CV AUC is 0.95, and testing AUC is 0.97 in RF. SVM realized CV accuracy of 81.67%, testing accuracy is 84%, CV AUC is 0.89, and testing AUC is 0.90.\u003c/p\u003e\n \u003cp\u003eTable\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e and Fig. 5 display the diagnosis results using the ML techniques. The ANN and RF achieve better performance. The ANN obtains the highest scores of CV accuracy, testing accuracy, CV AUC, and testing AUC. NB performs lower performance which obtains the lowest scores of CV accuracy, testing accuracy, CV AUC, and testing AUC. Figure 6 shows a testing ROC curve for the evaluation of ANN, DT, KNN, NB, RF, and SVM techniques.\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003ctable id=\"Tab3\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eEvaluation of ML classifiers based on time domain feature\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eClassifier\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eCV accuracy\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTesting Accuracy\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eCV AUC\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTesting AUC\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eANN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e95\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e96.67\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.98\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDT\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e90\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e93\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.94\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.95\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eKNN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e93.33\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e91.67\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.96\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.95\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e45\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e40\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.67\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.62\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRF\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e92\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e95\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.95\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.97\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSVM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e81.67\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e84\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.89\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.90\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003cdiv align=\"left\" class=\"colspec\"\u003e\u003cbr\u003e\u003c/div\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\n \u003ch2\u003e5.2. ITSC Fault Diagnosis based on the Most Important Features\u003c/h2\u003e\n \u003cp\u003eThe most important features parameter are Root Mean Square value (X\u003csub\u003eRMS\u003c/sub\u003e), Crest Factor (X\u003csub\u003eCF\u003c/sub\u003e), Peak-to-Peak Value (X\u003csub\u003ePP\u003c/sub\u003e), Impulse Factor (X\u003csub\u003eIF\u003c/sub\u003e), Energy (X\u003csub\u003eE\u003c/sub\u003e) and Kurtosis Value (X\u003csub\u003eKV\u003c/sub\u003e) [\u003cspan class=\"CitationRef\"\u003e19\u003c/span\u003e]. These parameters were taken for pre-processing the qd signals. The ANN model is designed for ITSC faults diagnosis as shown in Fig. \u003cspan class=\"InternalRef\"\u003e7\u003c/span\u003e.\u003c/p\u003e\n \u003cp\u003eThe experimental is implemented on IM different states of fault 2%, 5%, 7% and 10% ITSC at No load. The ANN is provided with 12 inputs during the ITSC faults diagnosis process. The inputs are 6 features obtained from Iq signal and 6 features obtained from Id signal. The dimension of the data set is 12 \u0026times; 200. The data set is split randomly into subsets of training and testing. The ratio of each subset is defined as 70% and 30%.\u003c/p\u003e\n \u003cp\u003eThis experimental is performed with DT, KNN, NB, RF and SVM algorithms. The cross validation accuracy, testing accuracy and AUC metric are determined to evaluate ANN, DT, KNN, NB, RF, and SVM techniques. The comparison between ML algorithms is indicated in Table \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e and Fig. 8. It was observed that the ANN and RF provide the highest performance. The ANN achieves CV accuracy of 99%, testing accuracy of 100%, CV AUC of 1 and testing AUC of 1 while RF provides CV accuracy of 99%, testing accuracy of 100%, CV AUC of 0.99, and testing AUC of 1. NB obtains the worst results that gets the lowest CV accuracy of 47%, testing accuracy of 45%, CV AUC of 0.67, and testing AUC of 0.67. Other classifiers provide CV accuracy higher than 84%, testing accuracy higher than 85%, CV AUC scores higher than 0.90, and testing AUC scores higher than 0.90. Moreover, it was proven that classifiers achieve high performance when the important features are used as compared with the classifiers using all time domain features as displayed in Table \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e and Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e. Therefore, the important features will be used in the following experimental due to its high accuracy. The ROC curve is presented in Fig. 9. It was observed that the ROC curve of ANN and RF are more to the upper left corner.\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003ctable id=\"Tab4\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eEvaluation of ML classifiers based on the important features\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eClassifier\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eCV accuracy\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTesting accuracy\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eCV AUC\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTesting AUC\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eANN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDT\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e92\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e98.33\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.95\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.99\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eKNN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e96.67\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.98\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e47\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e45\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.67\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.67\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRF\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSVM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e84\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e85\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.90\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.90\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003cdiv align=\"left\" class=\"colspec\"\u003e\u003cbr\u003e\u003c/div\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\n \u003ch2\u003e5.3. ITSC Fault Diagnosis using ANN and RF based on the Most Important Features\u003c/h2\u003e\n \u003cp\u003eSeveral experimental test are implemented on the IM at a loading condition of: 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90% and 100%. Different conditions of stator winding were tested: healthy, 2%, 5%, 7%, and 10% ITSC. The important features obtained from the qd signals are used to generate the data set. These data sets are used to train and test ANN and RF models. The performance of the classifier is evaluated by the CV accuracy, testing accuracy, and scores of the AUC metric. Table \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003e and Fig. \u003cspan class=\"InternalRef\"\u003e10\u003c/span\u003e illustrate the diagnosis results using the ANN compared with RF. ANN achieves the highest overall CV accuracy of 96.30%, overall testing accuracy of 96.50%, overall CV AUC of 0.97, and overall testing AUC of 0.98. The RF obtains the lowest overall CV accuracy of 93.30%, overall testing accuracy of 94.17%, overall CV AUC of 0.95, and overall testing AUC of 0.96. Hence, it can be concluded that the ANN based statistical feature extraction gives better performance as compared to all others approaches for ITSC fault diagnosis.\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003ctable id=\"Tab5\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 5\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eEvaluation of ANN and RF classifiers based on the important features\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003eLoading Condition\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colspan=\"4\"\u003e\n \u003cp\u003eANN\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colspan=\"4\"\u003e\n \u003cp\u003eRF\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eCV accuracy\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTesting accuracy\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eCV AUC\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTesting AUC\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eCV accuracy\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTesting accuracy\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eCV AUC\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTesting AUC\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e10%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e93\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e90\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.94\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.94\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e89\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e86.67\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.91\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.91\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e20%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e96\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e91.67\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.95\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e90\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e91.67\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.93\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.95\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e30%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e40%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e95\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e95\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.95\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.97\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e50%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e93\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e93.33\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.95\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.96\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e60%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e98.33\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e98.33\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.99\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e70%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e93\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e95\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.95\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e88\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e90\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.94\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.94\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e80%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e96.67\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e93\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e95\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.96\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.97\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e90%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e96.67\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e100%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e96.67\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e94\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e91.67\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.96\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.95\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOverall\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e96.30\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e96.50\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e93.30\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e94.17\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.95\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.96\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\n \u003ch2\u003e5.4. Comparison between ANN and ANFIS\u003c/h2\u003e\n \u003cp\u003eThe tests are performed on the IM with different fault conditions: 2%, 5%, 7%, and 10% ITSC faults at a full load. The algorithms have been designed in Matlab software. In this work, the input data is normalized in range from 0 to 1. The normalized data is pre-processed using two methods, the time domain features and the auto-regressive model.\u003c/p\u003e\n \u003cdiv id=\"Sec10\" class=\"Section3\"\u003e\n \u003ch2\u003e5.4.1. Comparison between ANN and ANFIS based on Time Domain Feature\u003c/h2\u003e\n \u003cp\u003eThe ANN and ANFIS models for classification relying on the most important features is performed. The important features are obtained from the qd signals. These features are used to classify ITSC faults conditions. The performance of classifiers is assisted by testing RMSE as shown in Table \u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003e. It is observed that the ANN achieves better performance than the ANFIS.\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003ctable id=\"Tab6\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 7\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eEvaluation of ANN and ANFIS based on time domain feature\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eClassifier\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eRMSE\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eANN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.2326\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eANFIS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.3442\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003c/div\u003e\n \u003cdiv id=\"Sec11\" class=\"Section3\"\u003e\n \u003ch2\u003e5.4.2. Comparison between ANN and ANFIS based on AR Model\u003c/h2\u003e\n \u003cp\u003eThe auto-regressive model is used for pre-processing the qd signals to ANN and ANFIS models. The AR model depends on extracting the hidden features of the signal with attenuating the noise as possible via the auto-correlation method. The coefficients of the AR model is calculated by linear prediction coding. The number of parameters of the AR model is five, which is considered the features of the AR model. Testing Root Mean Square Error (RMSE) is developed to evaluate the performance of the classifiers as indicated in Table \u003cspan class=\"InternalRef\"\u003e7\u003c/span\u003e. It was observed that the ANFIS obtains better accuracy than the ANN.\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003ctable id=\"Tab7\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 6\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eEvaluation of ANN and ANFIS based on AR Model\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eClassifier\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eRMSE\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eANN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.2638\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eANFIS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.0244\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003c/div\u003e\n \u003cdiv id=\"Sec12\" class=\"Section3\"\u003e\n \u003ch2\u003e5.4.3. Comparison between ANN and ANFIS based on DWT\u003c/h2\u003e\n \u003cp\u003eThe normalized data of qd signals is pre-processed using DWT. The mother wavelet \u0026quot;haar\u0026quot; is selected and the maximum decomposition level was obtained. The data set generated from DWT is applied to ANFIS and ANN. The testing data set is utilized to assess the efficiency of ANFIS and ANN using RMSE score as given in Table \u003cspan class=\"InternalRef\"\u003e8\u003c/span\u003e. It was observed that ANFIS achieves higher performance than ANN.\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003ctable id=\"Tab8\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 8\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eEvaluation of ANN and ANFIS based on DWT\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eClassifier\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eRMSE\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eANN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4.9393 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eANFIS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e6.4608 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;9\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cp\u003eFrom the previous discussion, thee different techniques were used for pre-processing the raw qd data before entering the classifier. The pre-processed data was employed as input to help the ANN and ANFIS training. Thus, the different pre-processing data aim to build the network model in the best possible structure. It is clear, that the DWT as pre-processing for input data to ANFIS gives beater results compared to ANN. On the other hand, the time domain feature as pre-processing for input data to ANN achieves good results compared to ANFIS.\u003c/p\u003e\n \u003cp\u003eIt is clear, the pre-processing techniques play an important part in ITSC faults diagnosis. ANN based on the most important time domain feature provides better performance than all other techniques. The worst output is obtained by NB based on time domain feature. ANN based on auto-regressive model gives weak performance. ANFIS performing well based on DWT. Therefore, the pre-processing of input data had an important role in classification issues where it was revealed in the performance of the selected model for diagnosing the ITSC faults in IM. So, before applying the training in classification models, the appropriate pre-processing methods for the input data to classifier models should be chosen. On the other hand, data pre-processing using features extraction gives distinct results in induction motor faults analysis.\u003c/p\u003e\n \u003c/div\u003e\n\u003c/div\u003e"},{"header":"6. CONCLUSION","content":"\u003cp\u003eThis paper presents a new application of ANN, ANFIS, DT, KNN, NB, RF, and SVM techniques to diagnose the induction motor ITSC faults. The motor is exposed to different ITSC faults under different loading conditions. The input data was pre-processed before entering to ANN, DT, KNN, NB, RF, and SVM models by time domain features extraction. The experimental was performed to diagnose ITSC faults using ANN, DT, KNN, NB, RF, and SVM techniques based on time domain features. The analysis of the results establishes that ANN archives the highest accuracy while the NB is the worst classifier. The experimental was implemented using ANN and ANFIS techniques based on time domain features and AR models and DWT. The evaluation analysis of the results proves that the ANN obtains the best results based on the time domain features as pre-processing for input data compared to ANFIS. On the other hand, the ANFIS obtains the beater results based on the DWT model as pre-processing for input data compared to ANN. The effectiveness of using the ANN and ANFIS based on data pre-processing to diagnose the ITSC faults in three-phase IM is investigated using experimental tests. In the future researches, some works could be considered as follows: Other types of faults such as bearing failure, phase-to-phase, open-phase and phase-to-ground faults may be considered based on data pre-processing.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eM.A.M. and M.A.M.H. designed the problem under study; M.A.M. performed the simulations and obtained the results; M.A.M.H. analyzed the obtained results; M.A.M. wrote the paper, which was further reviewed by M.A.M.H.All authors have read and agreed to the published version of the manuscript.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eKonar, P., \u0026amp; Chattopadhyay, P. (2015). Multi-class fault diagnosis of induction motor using Hilbert and Wavelet Transform. Applied Soft Computing, 30, 341-352.\u0026rlm;\u003c/li\u003e\n \u003cli\u003eMenshawy A. M., E. Mohamed, A.-A. Mohamed, M. Abdel-Nasser, and M. A. Moustafa Hassan, \u0026ldquo;Detection of inter turn short circuit faults in induction motor using artificial neural network,\u0026rdquo; \u003cem\u003ein 2020 26th Conference of Open Innovations Association (FRUCT),\u003c/em\u003e 2020, pp. 297-304. \u0026nbsp;\u0026nbsp;\u003c/li\u003e\n \u003cli\u003e\u0026nbsp;Menshawy A. M., E. Mohamed, A.-A. Mohamed, M. Abdel-Nasser, and M. A. Moustafa Hassan, \u0026ldquo;Induction motor broken rotor bar faults diagnosis using ANFIS-based DWT,\u0026rdquo; \u003cem\u003eInt. J. Model. Simul.\u003c/em\u003e, pp. 1\u0026ndash;14, 2019.\u003c/li\u003e\n \u003cli\u003eS. Bensaoucha, S. A. Bessedik, A. Ameur, and A. Teta,\u0026nbsp;\u0026ldquo;Induction motors broken rotor bars detection using RPVM and neural network,\u0026rdquo; \u003cem\u003eCOMPEL\u003c/em\u003e, vol. 38, no. 2, pp. 596\u0026ndash;615, 2019.\u003c/li\u003e\n \u003cli\u003e\u0026nbsp;Almounajjed, A., Sahoo, A. K., \u0026amp; Kumar, M. K. (2021). Diagnosis of stator fault severity in induction motor based on discrete wavelet analysis. Measurement, 182, 109780.\u0026rlm;\u003c/li\u003e\n \u003cli\u003eM. Eftekhari, M. Moallem, S. Sadri, and M.-F. Hsieh,\u0026nbsp;\u0026ldquo;A novel indicator of stator winding inter-turn fault in induction motor using infrared thermal imaging,\u0026rdquo; \u003cem\u003eInfrared Phys. Technol.\u003c/em\u003e, vol. 61, pp. 330\u0026ndash;336, 2013.\u003c/li\u003e\n \u003cli\u003eDiscrete wavelet transform and energy eigen value for rotor bars fault detection in variable speed field-oriented control of induction motor drive\u003c/li\u003e\n \u003cli\u003eY. Chen, X. Chen, and Y. Shen, \u0026ldquo;On-line detection of coil inter-turn short circuit faults in dual-redundancy permanent magnet synchronous motors,\u0026rdquo; \u003cem\u003eEnergies\u003c/em\u003e, vol. 11, no. 3, p. 662, 2018.\u003c/li\u003e\n \u003cli\u003eM. S. Khadim, S. Kuldeep and V. K. Giri, \u0026ldquo;Diagnosis of Stator inter-turn fault in pwm inverter fed induction motor by advanced dsp technique,\u0026rdquo; \u003cem\u003ei-manag. J. Electr. Eng.\u003c/em\u003e, vol. 12, no. 3, p. 9, 2019.\u003c/li\u003e\n \u003cli\u003eW. Zhang, M.-P. Jia, L. Zhu, and X.-A. Yan, \u0026ldquo;Comprehensive overview on computational intelligence techniques for machinery condition monitoring and fault diagnosis,\u0026rdquo; \u003cem\u003eChin. J. Mech. Eng.\u003c/em\u003e, vol. 30, no. 4, pp. 782\u0026ndash;795, 2017.\u003c/li\u003e\n \u003cli\u003eN. Ngote, and M. Ouassaid,\u0026nbsp;\u0026ldquo;A Hybrid TSA-Fuzzy Logic Approach to Detect Induction Motor Rotor Faults,\u0026rdquo; \u003cem\u003eInternational Journal of Information Science and Technology\u003c/em\u003e, vol. 3, no. 1, pp. 26-35, 2019.\u003c/li\u003e\n \u003cli\u003eL. S. Maraaba, Z. M. Al-Hamouz, A. S. Milhem, and M. A. Abido, \u0026ldquo;Neural network-based diagnostic tool for detecting Stator inter-turn faults in line start permanent magnet synchronous motors,\u0026rdquo; \u003cem\u003eIEEE Access\u003c/em\u003e, vol. 7, pp. 89014\u0026ndash;89025, 2019.\u003c/li\u003e\n \u003cli\u003eM. Skowron, M. Wolkiewicz, T. Orlowska-Kowalska, and C. T. Kowalski, \u0026ldquo;Effectiveness of selected neural network structures based on axial flux analysis in Stator and rotor winding incipient fault detection of inverter-fed induction motors,\u0026rdquo; \u003cem\u003eEnergies\u003c/em\u003e, vol. 12, no. 12, p. 2392, 2019.\u003c/li\u003e\n \u003cli\u003eR. F. Ribeiro Junior, F. A. de Almeida, and G. F. Gomes, \u0026ldquo;Fault classification in three-phase motors based on vibration signal analysis and artificial neural networks,\u0026rdquo; \u003cem\u003eNeural Comput. Appl.\u003c/em\u003e, vol. 32, no. 18, pp. 15171\u0026ndash;15189, 2020.\u003c/li\u003e\n \u003cli\u003eW. Jing, T. Jiang, X. Zhang, and L. Zhu, \u0026ldquo;The optimisation of speech recognition based on convolutional neural network,\u0026rdquo; \u003cem\u003eInt. j. high perform. comput. netw.\u003c/em\u003e, vol. 13, no. 2, p. 222, 2019.\u003c/li\u003e\n \u003cli\u003e(15)J. Li, X. Ban, G. Yang, Y. Li, and Y. Wang, \u0026ldquo;Real-time human action recognition using depth motion maps and convolutional neural networks,\u0026rdquo; \u003cem\u003eInt. j. high perform. comput. netw.\u003c/em\u003e, vol. 13, no. 3, p. 312, 2019.\u003c/li\u003e\n \u003cli\u003eKumar, P., \u0026amp; Hati, A. S. (2022). Dilated convolutional neural network based model for bearing faults and broken rotor bar detection in squirrel cage induction motors. Expert Systems with Applications, 191, 116290.\u0026rlm;\u003c/li\u003e\n \u003cli\u003eMenshawy A. M., Mahmoud, A. S., Essam, H. A. and M. A. Moustafa Hassan,\u0026nbsp;\u0026quot;Detection and Classification of Broken Rotor Bars Faults in Induction Motor Using Adaptive Neuro-Fuzzy Inference System,\u0026quot; \u003cem\u003eProceedings of MEPCON \u0026lsquo;14, Cairo, Egypt\u003c/em\u003e, 2014.\u003c/li\u003e\n \u003cli\u003eMenshawy A. Mohamed, M. A. Moustafa. Hassan, Mostafa Dardeer and et al., \u0026quot;Diagnostic Modelling for Induction Motor Faults via ANFIS Algorithm and DWT-Based Feature Extraction,\u0026quot; Applied Sciences. Vol. 11, No. 19, pp. 1-18, 2021.\u0026rlm;\u003c/li\u003e\n \u003cli\u003eS. Esakimuthu Pandarakone, Y. Mizuno, and H. Nakamura, \u0026ldquo;A comparative study between machine learning algorithm and artificial intelligence neural network in detecting minor bearing fault of induction motors,\u0026rdquo; \u003cem\u003eEnergies\u003c/em\u003e, vol. 12, no. 11, p. 2105, 2019.\u003c/li\u003e\n \u003cli\u003eJ. C. Quiroz, N. Mariun, M. R. Mehrjou, M. Izadi, N. Misron, and M. A. Mohd Radzi,\u0026nbsp;\u0026ldquo;Fault detection of broken rotor bar in LS-PMSM using random forests,\u0026rdquo; \u003cem\u003eMeasurement (Lond.)\u003c/em\u003e, vol. 116, pp. 273\u0026ndash;280, 2018.\u003c/li\u003e\n \u003cli\u003eP. Kumar and A. S. Hati, \u0026ldquo;Review on machine learning algorithm based fault detection in induction motors,\u0026rdquo; \u003cem\u003eArch. Comput. Methods Eng.\u003c/em\u003e, 2020.\u003c/li\u003e\n \u003cli\u003eI. O. Zaparoli, L. M. Rabelo Baccarini, P. C. M. Lamim Filho, and F. B. Batista, \u0026ldquo;Transient envelope current analysis for inter-turn short-circuit detection in induction motor stator,\u0026rdquo; \u003cem\u003eJ. Braz. Soc. Mech. Sci. Eng.\u003c/em\u003e, vol. 42, no. 2, 2020.\u003c/li\u003e\n \u003cli\u003eEbrahimi, B. M., Faiz, J., Lotfi-Fard, S., \u0026amp; Pillay, P. (2012). Novel indices for broken rotor bars fault diagnosis in induction motors using wavelet transform. Mechanical Systems and Signal Processing, 30, 131-145.\u0026rlm;\u003c/li\u003e\n \u003cli\u003eMostafa M., Galal H. and Mohamed A. Moustafa Hassan,\u0026nbsp;\u0026ldquo;Diagnosis of Rotating Machines Faults Using Artificial Intelligence Based on Preprocessing for Input Data,\u0026rdquo; \u003cem\u003ein 2020 26th Conference of Open Innovations Association (FRUCT),\u003c/em\u003e 2020, pp. 572-582. \u0026nbsp;\u003c/li\u003e\n \u003cli\u003eC. G. Dias, L. C. da da Silva, and I. E. Chabu, \u0026ldquo;Fuzzy-based statistical feature extraction for detecting broken rotor bars in line-fed and inverter-fed induction motors,\u0026rdquo; \u003cem\u003eEnergies\u003c/em\u003e, vol. 12, no. 12, p. 2381, 2019.\u003c/li\u003e\n \u003cli\u003eMohamed A. Moustafa Hassan, \u0026ldquo;Effect of Pre-processing on Using ANN and ANFIS,\u0026rdquo; in 2020 26th Conference of Open Innovations Association (FRUCT), 2020, pp. 589-603. \u0026nbsp;\u003c/li\u003e\n \u003cli\u003eD. M. Sonje, P. Kundu, and A. Chowdhury, \u0026ldquo;A novel approach for sensitive inter-turn fault detection in induction motor under various operating conditions,\u0026rdquo; \u003cem\u003eArab. J. Sci. Eng.\u003c/em\u003e, vol. 44, no. 8, pp. 6887\u0026ndash;6900, 2019.\u003c/li\u003e\n \u003cli\u003eTalhaoui, H., Menacer, A., Kessal, A., \u0026amp; Kechida, R. (2014). Fast Fourier and discrete wavelet transforms applied to sensorless vector control induction motor for rotor bar faults diagnosis. ISA transactions, 53(5), 1639-1649.\u0026rlm;\u003c/li\u003e\n \u003cli\u003eYan, R., Gao, R. X., \u0026amp; Chen, X. (2014). Wavelets for fault diagnosis of rotary machines: A review with applications. Signal processing, 96, 1-15.\u0026rlm;\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eGketsis, Z. E., Zervakis, M. E., \u0026amp; Stavrakakis, G. (2009). Detection and classification of winding faults in windmill generators using Wavelet Transform and ANN. Electric Power Systems Research, 79(11), 1483-1494.\u0026rlm;\u003c/li\u003e\n \u003cli\u003eCherif, H., Benakcha, A., Laib, I., Chehaidia, S. E., Menacer, A., Soudan, B., \u0026amp; Olabi, A. G. (2020). Early detection and localization of stator inter-turn faults based on discrete wavelet energy ratio and neural networks in induction motor. Energy, 212, 118684.\u0026rlm;\u003c/li\u003e\n \u003cli\u003eTaher, S. A., Malekpour, M., \u0026amp; Farshadnia, M. (2014). Diagnosis of broken rotor bars in induction motors based on harmonic analysis of fault components using modified adaptive notch filter and discrete wavelet transform. Simulation Modelling Practice and Theory, 44, 26-41.\u003cspan dir=\"RTL\"\u003e\u0026rlm;\u003c/span\u003e \u0026nbsp;\u003c/li\u003e\n \u003cli\u003eSyed, S. H., \u0026amp; Muralidharan, V. (2022). Feature extraction using Discrete Wavelet Transform for fault classification of planetary gearbox\u0026ndash;A comparative study. Applied Acoustics, 188, 108572.\u003cspan dir=\"RTL\"\u003e\u0026rlm;\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003eKeskes, H., Braham, A., \u0026amp; Lachiri, Z. (2013). Broken rotor bar diagnosis in induction machines through stationary wavelet packet transform and multiclass wavelet SVM. Electric Power Systems Research, 97, 151-157.\u003cspan dir=\"RTL\"\u003e\u0026rlm;\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003eDong, X., Li, G., Jia, Y., \u0026amp; Xu, K. (2021). Multiscale feature extraction from the perspective of graph for hob fault diagnosis using spectral graph wavelet transform combined with improved random forest. Measurement, 176, 109178.\u003cspan dir=\"RTL\"\u003e\u0026rlm;\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003eLi, Z., Yan, X., Yuan, C., Peng, Z., \u0026amp; Li, L. (2011). Virtual prototype and experimental research on gear multi-fault diagnosis using wavelet-autoregressive model and principal component analysis method. Mechanical Systems and Signal Processing, 25(7), 2589-2607.\u003cspan dir=\"RTL\"\u003e\u0026rlm;\u0026nbsp;\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003eI. Martin-Diaz, D. Morinigo-Sotelo, O. Duque-Perez, and R. J. Romero-Troncoso, \u0026ldquo;An experimental comparative evaluation of machine learning techniques for motor fault diagnosis under various operating conditions,\u0026rdquo; \u003cem\u003eIEEE Trans. Ind. Appl.\u003c/em\u003e, vol. 54, no. 3, pp. 2215\u0026ndash;2224, 2018.\u003c/li\u003e\n \u003cli\u003eJ. Patel, \u0026nbsp;V. Patel,and A. Patel,\u003cem\u003e\u0026nbsp;\u003c/em\u003e\u0026quot;Fault diagnostics of rolling bearing based on improve time and frequency domain features using artificial neural networks,\u0026quot; \u003cem\u003eInternational Journal for Scientific Research \u0026amp; Development\u003c/em\u003e, vol. 1, no. 4, pp. 781-788, 2013.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Inter Turn Short Circuit Faults, Artificial Intelligence, Machine Learning, Time Domain Features, Auto-Regressive Model, Discrete Wavelet Transform","lastPublishedDoi":"10.21203/rs.3.rs-7242428/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7242428/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThis paper proposes a new method using Artificial Neural Network (ANN) and Adaptive Neural Fuzzy Inference System (ANFIS) for diagnosis Inter Turn Short Circuit (ITSC) faults in induction motor. The proposed diagnosis procedure is based on the analysis of stator current. The study includes a comparative analysis of a various diagnostic methods such as decision tree, k-nearest neighbours, naive bayes, random forest and support vector machine. The time domain features extraction pre-processes the input data before entering to classifier model. The test accuracy and cross-validation analysis evaluate the model efficiency. The most important time domain features are selected to improve the performance of the classifier. ANN based the most important time domain features gives better performance as compared to all others classifier. This paper presents a comparison between ANN and ANFIS based on the most important time domain features, auto-regressive model and discrete wavelet transform. ANFIS based discrete wavelet transform achieves higher performance than ANN. The laboratory experiments on 1.5 HP squirrel cage induction motor under different loading conditions verify the proposed technique efficiency to diagnose fully various ITSC faults\u003c/p\u003e","manuscriptTitle":"Stator Winding Faults Diagnosis in Induction Motor Based on ANN and ANFIS","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-08-06 04:22:48","doi":"10.21203/rs.3.rs-7242428/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"0925295c-b559-4887-8dae-3ddc7e00ff7e","owner":[],"postedDate":"August 6th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-08-17T06:53:22+00:00","versionOfRecord":[],"versionCreatedAt":"2025-08-06 04:22:48","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7242428","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7242428","identity":"rs-7242428","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}