Which type of input data are used most frequently in fault diagnosis of electric motors?
Best insight from top research papers
In fault diagnosis of electric motors, various types of input data are utilized. The predominant input data types include audio signals for potential engine failure detection , vibration signatures for stator fault detection , and acoustic signatures for early fault diagnosis in Electric Vehicle (EV) motors . These data types are crucial for timely detection of faults, preventive maintenance, and ensuring operational safety. Audio-based fault classification using AutoEncoder Neural Networks is common for engine failure detection, while vibration data is employed for stator fault identification with high accuracy. Additionally, acoustic signature-based transfer learning is utilized for early fault identification in EV motors, enhancing operational reliability and safety. The integration of these diverse data sources showcases the importance of multi-modal approaches in fault diagnosis for electric motors.
Answers from top 5 papers
More filters
| Papers (5) | Insight |
|---|---|
14 Dec 2022 | Acoustic signature-based data, specifically Constant Q-Transform (CQT) based on time-frequency scalograms, is used most frequently for fault diagnosis of electric motors in the research. |
22 May 2023 | Vibration signals are predominantly utilized for fault diagnosis in electric motors, offering non-invasive and efficient stator fault detection compared to traditional current-based methods. |
2 Citations | The paper discusses using real-world data from a permanent magnet synchronous machine for fault diagnosis, highlighting the benefits of a combined approach using both data-driven and model-based techniques. |
16 Nov 2022 | Spectrogram images derived from audio time series data are frequently used as input data for fault diagnosis of electric motors, enabling accurate classification using deep learning techniques. |
| Spectrogram images derived from audio time series data are frequently used as input for fault diagnosis of electric motors, enhancing classification accuracy above 97% and AUC above 99%. |
Related Questions
What other methods can be used for detection and classification of faults in pmsm?5 answersVarious methods can be employed for the detection and classification of faults in Permanent Magnet Synchronous Motors (PMSMs). These include the application of metric learning (ML) for fault diagnosis, utilizing structural characteristics and amplitude-trend distance (ATD) to differentiate between healthy and faulty sectors. Additionally, techniques like input-output linearization without a speed sensor, Extended Kalman Filter (EKF) estimation, Fast Fourier Transform (FFT), and Discrete Wavelet Transform (DWT) can be utilized for detecting supply faults and short circuit faults in PMSMs. Moreover, the use of bispectrum analysis of stator phase current and Convolutional Neural Network (CNN) for intelligent diagnosis, along with Continuous Wavelet Transform (CWT) and artificial intelligence (AI) techniques for stator winding fault detection and classification, have shown promising results in fault detection in PMSMs.
Using graph network s for understanding complex relationships between motor components in motor fault diagnosis?5 answersGraph networks, such as Graph Attention Networks (GATs) and Graph Neural Networks (GNNs), are instrumental in understanding complex relationships between motor components for fault diagnosis. These networks excel in mining graph structures, integrating multiple node relationships, and effectively fusing multi-sensor information to enhance fault diagnosis accuracy. Additionally, the use of knowledge graphs in conjunction with Convolutional Neural Networks (CNNs) further enriches fault diagnosis by leveraging semantic knowledge and structured data. Moreover, the introduction of multireceptive field graph convolutional networks (MRF-GCNs) addresses limitations in existing GCNs by considering weighted graphs and enhancing feature representation, proving superior performance even with imbalanced datasets. Overall, these advanced graph-based models offer a robust framework for comprehensively analyzing and diagnosing motor faults by capturing intricate interdependencies among motor components.
How can the hidden Markov model be used for fault diagnosis in rotary machinery?5 answersHidden Markov models (HMMs) can be used for fault diagnosis in rotary machinery. HMMs have been applied to fault detection and prediction in continuous industrial processes. However, the EM algorithm in HMMs can have local optimality problems. To address this, a hidden Markov model-Bayesian network (HMM-BN) hybrid model has been proposed. The model combines expert empirical knowledge with the EM algorithm to accurately diagnose the fault root cause variable. The improved EM algorithm uses sequential and parallel learning to alleviate initial sensitivity and local optimum problems. The model provides fault detection, prediction, and root cause variable detection results based on log-likelihood estimates calculated by the improved HMM. Experimental results using the Tennessee Eastman (TE) process and the continuously stirred tank reactor (CSTR) process demonstrate the feasibility and effectiveness of the model in finding faults and identifying their causes.
What are the percentage accuracy in data driven methods used to detect eccentricity faults in electrical machines?4 answersData-driven methods used to detect eccentricity faults in electrical machines have shown high accuracy rates. In one study, a new diagnosis method for mixed eccentricity in external rotor permanent magnet motors (ERPMMs) achieved a maximum absolute diagnostic error of less than 4.0%. Another study compared the performance of four machine learning techniques in detecting eccentricity faults of induction motors and found that the highest accuracy rates were achieved using rms and p2p features of vibration signals. A machine learning pipeline using support vector machines (SVMs) with time domain statistical features achieved overall accuracy rates of 91% to 99% and F-scores of 0.81 to 0.99 for multi-level fault diagnosis and classification of rolling-element bearings. Additionally, a linear regression model incorporating multiple fault signature features showed satisfactory accuracy in estimating eccentricity severity levels of induction machines under different operating conditions.
How can artificial intelligence be used to diagnose electric vehicles motors?5 answersArtificial intelligence can be used to diagnose electric vehicle motors by utilizing various techniques. One approach is the use of artificial neural networks (ANNs). ANNs can be trained on real-time data collected from sensors installed in the electric motors. These sensors provide information about the motor's performance and health condition. The ANN can analyze this data and identify any potential faults or abnormalities in the motor. Another technique is the use of acoustic signature-based transfer learning, which involves analyzing the acoustic signals produced by the motor using time-frequency scalograms. This approach can accurately identify motor faults at an early stage, improving operational safety and avoiding unplanned breakdowns. Additionally, machine learning techniques such as decision forest and decision tree algorithms can be used to classify the health condition of the motor based on vibration signals collected from accelerometers. These techniques enable the diagnosis and prediction of motor health in electric vehicle powertrain applications.
What are the major data driven techniques used for fault detection using SCADA?4 answersData-driven techniques used for fault detection using SCADA include structural analysis (SA) combined with machine learning algorithms. These techniques involve determining the structure of analytical redundancy relations (ARRs) between inputs and outputs using SA, and then using machine learning approaches to obtain analytical relations between inputs and outputs. Another approach is remote condition monitoring and data science techniques, which enable round-the-clock diagnoses of wind turbine subsystems based on data from the supervisory control and data acquisition system. Additionally, a clustering-based multi-turbine fault detection approach has been proposed, which involves WT clustering, single-turbine modeling, and fault indicator calculation. Finally, a method combining Synthetic Minority Over-Sampling Technique (SMOTE) with Generative Adversarial Networks (GAN) has been used to preprocess training data for fault detection of mechanical rotating parts.