A multi-scale feature extraction and fusion method for diagnosing bearing faults

Online knowledge distillation conducts knowledge transfer among all student models to alleviate the reliance on pre-trained models. However, existing online methods rely heavily on the prediction distributions and neglect the further exploration of the representational knowledge. In this paper, we propose a novel Multi-scale Feature Extraction and Fusion method (MFEF) for online knowledge distillation, which comprises three key components: Multi-scale Feature Extraction, Dual-attention and Feature Fusion, towards generating more informative feature maps for distillation. The multiscale feature extraction exploiting divide-and-concatenate in channel dimension is proposed to improve the multi-scale representation ability of feature maps. To obtain more accurate information, we design a dual-attention to strengthen the important channel and spatial regions adaptively. Moreover, we aggregate and fuse the former processed feature maps via feature fusion to assist the training of student models. Extensive experiments on CIF AR-10, CIF AR-100, and CINIC-10 show that MFEF transfers more beneficial representational knowledge for distillation and outperforms alternative methods among various network architectures

Fault diagnosis of rotating machinery is vital to identify incipient failures and avoid unexpected downtime in industrial systems. This paper proposes a new rolling bearing fault diagnosis method by integrating the fine-to-coarse multiscale permutation entropy (F2CMPE), Laplacian score (LS) and support vector machine (SVM). A novel entropy measure, named F2CMPE, was proposed by calculating permutation entropy via multiple-scale fine-grained and coarse-grained signals based on the wavelet packet decomposition. The entropy measure estimates the complexity of time series from both low- and high-frequency components. Moreover, the F2CMPE mitigates the drawback of producing time series with sharply reduced data length via the coarse-grained procedure in the conventional composite multiscale permutation entropy (CMPE). The comparative performance of the F2CMPE and CMPE is investigated by analyzing the synthetic and experimental signals for entropy-based feature extraction. In the proposed bearing fault diagnosis method, the F2CMPE is first used to extract the entropy-based features from bearing vibration signals. Then, LS and SVM are used for selection of features and fault classification, respectively. Finally, the effectiveness of the proposed method is verified for rolling bearing fault diagnosis using experimental vibration data sets, and the results have demonstrated the capability of the proposed method to recognize and identify the bearing fault patterns under different fault states and severity levels.

The surface electromyography (sEMG) technique is proposed for muscle activation detection and intuitive control of prostheses or robot arms. Motion recognition is widely used to map sEMG signals to the target motions. One of the main factors preventing the implementation of this kind of method for real-time applications is the unsatisfactory motion recognition rate and time consumption. The purpose of this paper is to compare eight combinations of four feature extraction methods (Root Mean Square (RMS), Detrended Fluctuation Analysis (DFA), Weight Peaks (WP), and Muscular Model (MM)) and two classifiers (Neural Networks (NN) and Support Vector Machine (SVM)), for the task of mapping sEMG signals to eight upper-limb motions, to find out the relation between these methods and propose a proper combination to solve this issue. Seven subjects participated in the experiment and six muscles of the upper-limb were selected to record sEMG signals. The experimental results showed that NN classifier obtained the highest recognition accuracy rate (88.7%) during the training process while SVM performed better in real-time experiments (85.9%). For time consumption, SVM took less time than NN during the training process but needed more time for real-time computation. Among the four feature extraction methods, WP had the highest recognition rate for the training process (97.7%) while MM performed the best during real-time tests (94.3%). The combination of MM and NN is recommended for strict real-time applications while a combination of MM and SVM will be more suitable when time consumption is not a key requirement.

Diabetic retinopathy is a serious complication of diabetes that can lead to blindness if not detected and treated early. Automated detection of diabetic retinopathy requires effective feature extraction techniques to enhance diagnostic accuracy. This study aims to develop a method for detecting diabetic retinopathy by utilizing Local Binary Pattern (LBP) combined with wavelet transform, and then classifying the extracted features using Support Vector Machine (SVM). The approach includes feature extraction from retinal images using LBP and wavelet transform. The extracted features are subsequently classified with SVM to evaluate performance in detecting diabetic retinopathy. Analysis results show that the dominant feature is found in the fifth row with a value of 0.57006, indicating the effectiveness of the LBP method in feature extraction. The developed model demonstrates high performance with an accuracy of 95.59%, precision of 96%, recall of 97.96%, and F1-score of 96.97%. The combination of feature extraction methods with SVM proves to be effective and reliable in detecting diabetic retinopathy, offering low error rates and high accuracy, thus potentially serving as a valuable tool in clinical diagnosis

Facial Expression Recognition (FER), also known as Facial Emotion Recognition, is an active topic in computer vision and machine learning fields. This paper analyzes different feature extraction and classification methods to propose an efficient facial expression recognition system. We have studied several feature extraction methods, including Histogram of Oriented Gradients (HOG), face-encoding, and the features extracted by a VGG16 Network. For classification, different classical classifiers, including Support Vector Machines (SVM), Adaptive Boosting (AdaBoost), and Logistic Regression, are evaluated with these features. Besides, we have trained a ResNet50 model from scratch and also tuned a ResNet50 which is pre-trained on VGGFace2 dataset. Finally, a part-based ensemble classifier is also proposed by focusing on different parts of face images. The experimental results provided on FER-2013 Dataset show that the tuned model of ResNet50 with a complete image of face, achieves higher performance than the other methods.

Feature extraction is one of the most important, pivotal, and difficult problems in mechanical fault diagnosis, which directly relates to the accuracy of fault diagnosis and the reliability of early fault prediction. Therefore, a new fault feature extraction method, called the EDOMFE method based on integrating ensemble empirical mode decomposition (EEMD), mode selection, and multi-scale fuzzy entropy is proposed to accurately diagnose fault in this paper. The EEMD method is used to decompose the vibration signal into a series of intrinsic mode functions (IMFs) with a different physical significance. The correlation coefficient analysis method is used to calculate and determine three improved IMFs, which are close to the original signal. The multi-scale fuzzy entropy with the ability of effective distinguishing the complexity of different signals is used to calculate the entropy values of the selected three IMFs in order to form a feature vector with the complexity measure, which is regarded as the inputs of the support vector machine (SVM) model for training and constructing a SVM classifier (EOMSMFD based on EDOMFE and SVM) for fulfilling fault pattern recognition. Finally, the effectiveness of the proposed method is validated by real bearing vibration signals of the motor with different loads and fault severities. The experiment results show that the proposed EDOMFE method can effectively extract fault features from the vibration signal and that the proposed EOMSMFD method can accurately diagnose the fault types and fault severities for the inner race fault, the outer race fault, and rolling element fault of the motor bearing. Therefore, the proposed method provides a new fault diagnosis technology for rotating machinery.

According to the rolling bearing local fault vibration mechanism, a monopulse feature extraction and fault diagnosis method of rolling bearing under low‐speed and heavy‐load conditions based on phase scan and CNN is proposed. The synchronous collected speed signal is used to calculate bearing phase function and divide fault monopulse periods. The monopulse waveforms of multiple fault periods are scanned and ensemble averaged to suppress noise interference and detail feature loss at the same time of feature extraction. By iteratively calibrating phase function, the feature matrix containing bearing fault information can be obtained. Finally, CNN is used to recognize and classify different bearing states. The experimental and analysis results show that bearing fault diagnosis can be achieved. The total recognition rates of constant and variable speed samples are 99.67% and 99.89%, respectively. The trained network has fast convergence speed and good generalization ability for different fault sizes and working conditions. Further experiments show that the method can also accurately identify different bearing degradation states. The total recognition rates of constant and variable speed samples are 96.67% and 95.56%, respectively. The limited errors are concentrated between the degradation states with the same type weak fault. The experimental results using Case Western Reserve University bearing data show that feature extraction and network training are better, and the recognition rates of 5 bearing states are all 100%. Therefore, the proposed method is an effective rolling bearing feature extraction and fault diagnosis technology.

In the field of computer vision and image processing CBIR (Content based image retrieval) system become an active research area. The CBIR technique involves feature extraction based on image features (like color, shape and texture) and defining rules for comparing image. A classifier is also used by authors for image classification. For this purpose an image database taken from James Z. Wang research group website is used by many authors. Due to enormous increase in the size of database traditional methods of image retrieval by using keywords does not provide accurate result. Thus the concept of CBIR system is emerged. In last couple of years many method of feature extraction is developed but since now researcher are not able to develop a system which has gained high accuracy of human visual perception in distinguishing pictures. Therefore earlier researchers have concentrated to increase the retrieval accuracy of CBIR system by using different combination of features. The objective of this paper is to present review on a CBIR technique which combined four kinds of features set including every kind of feature for developing feature vector. Color based features (histogram), statistical features (mean, median and standard deviation), spatial features (Gabor feature and wavelet features) and texture features (GLCM) is used for forming feature vector. A SVM classifier is used with CBIR system to increase retrieval accuracy. SVM (support vector machine) is a machine learning classifier which provides good classification accuracy. Keywords: CBIR, SVM, NN, GLCM, HSV Color space, Euclidean distance matrix.

Rolling bearings are the key components of rotating machinery. Incipient fault diagnosis of bearing plays an increasingly important role in guaranteeing normal and safe operation of rotating machinery. However, because of the high complexity of the fault feature extraction, the incipient faults of rolling bearings are difficult to diagnose. To solve this problem, this paper presents a new incipient fault intelligent identification method of rolling bearings based on variational mode decomposition (VMD), principal component analysis (PCA), and support vector machines (SVM). In the proposed method, the bearing vibration signals are decomposed by using VMD, and a series of intrinsic mode functions (IMFs) with different frequencies are obtained. Then, the energy and kurtosis values of each IMF are calculated to reveal the intrinsic characteristics of the vibration signals in different scales. Finally, all energy and kurtosis values of IMFs are processed via PCA and subsequently fed into SVM to achieve the bearing fault identification automatically. The effectiveness of this method is verified through the experimental bearing data. The verification results indicate that the proposed method can effectively extract the bearing fault features and accurately identify the bearing incipient faults, and outperform the two compared methods obviously in identification accuracy and computation time.

In recent years, machine learning technology is widely used in the field of fault diagnosis for bearings. Although these methods usually work well, the following defects still exist when they are dealing with large amount of fault data: (1) feature extraction methods need to rely on expertise or signal processing technologies. Therefore, there is a lack of a feature extraction method that is common to different diagnostic problems; (2) shallow models can't learn more complex mapping relationships well; (3) traditional intelligent diagnostic methods are usually computationally intensive and slow in convergence. Inspired by the Auto-encoder’s (AE) feature extraction capability and fast training speed of the Extreme Learning Machine (ELM), a new fault diagnosis method for bearings based on Extreme Learning Machine-Autoencoder (ELM-AE) is proposed in this paper. With its automatic feature extraction capability and very efficient learning strategy, the raw vibration signals of bearings are directly sent to the model without any manual feature extraction for fault diagnosis, which overcomes the above drawbacks. The experimental results on CWRU bearing dataset show that the proposed method takes into account both diagnostic accuracy and time efficiency. Compared with existing literatures, our proposed method obtains superior accuracy.

Tail rotor bearing is the key components of Unmanned Aerial Vehicle (UAV) tail rotor to help the tail rotor blade to keep the balance and control the course. The previous researches about tail rotor bearing diagnosis are based on vibration, acoustic, temperature and others. However, the existing monitoring signals are easily contaminated by environment or measurement noises and fail to detect early fault with high accuracy. To overcome this challenge, this paper utilizes the ultrasound signal to monitor health condition of tail rotor bearing, and proposes an automatic feature extraction and fusion method. In our proposed method, Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) is first used decompose the ultrasound signal into several intrinsic mode functions (IMFs) containing different fault-related information. Then, we design an attention-based weight module to dynamically assign different weights to each IMFs. Finally, the weighted IMFs are fed to a CNN-based feature extraction module to perform feature learning and classification. A Unmanned Aerial vehicle (UAV) rotor simulation bench has been setup to validate the performance of our proposed method. Experimental results illustrate that the proposed approach can effectively decompose and extract the discriminative fault features of Unmanned Aerial vehicle (UAV) tail rotor bearing and achieved superior diagnosis performance.

In view of the fact that the random noise interferes with the characteristic extraction of a rolling bearing fault signal, a new method of fault feature extraction is proposed based on the combination of the generalized S transform and singular value decomposition (SVD). Firstly, the 2D time–frequency spectrum bearing fault signal is obtained by applying the generalized S transform, and the time–frequency spectrum matrix is used as the objective matrix of SVD to solve the singular values. Then the K-means clustering algorithm is used to classify the singular value sequence, and the singular values for reconstruction are determined. Finally, the de-noised matrix is carried out the generalized S inversion transform to get the de-noised fault signal, and the power spectrum is calculated to finish the fault diagnosis. By analyzing the simulated signal and the actual bearing fault data, results show that the proposed method can effectively identify typical faults of rolling bearings and improve the diagnosis effect of rolling bearing faults. And it provides a new way to realize the fault diagnosis of rolling bearings under noise.

Traditional bearing sensory diagnostic include touching and hearing rely on personal experience, and for more complex system are unable to meet the needs of equipment fault diagnosis. The research on bearing fault diagnosis is developing significantly. Bearings are used in rotating machinery and most machinery failures are caused by bearing failures. The fault diagnosis of bearings is an important research area. The core of bearing fault diagnosis is the pattern recognition of fault features. The key of pattern recognition is to develop a reasonable classifier. Intelligent pattern recognition has been developed such as principal components, support vector machine, neural network. In this study, a bearing fault diagnosis based on exploring pattern recognition is proposed. The key to pattern recognition is to design a significant classifier. A number of features from bearing vibration of normal and fault bearing are extracted and processed using principal components of correlation matrix. Plot of principal components shows the visualization of normal and fault bearing and the classifier is chosen subjectively. The principal components exploration will be confirmed using least squares support vector machine. The parameter of support vector machine estimated using heuristic optimization particle swarm optimization. The proposed method can be applied in the detection of faults of bearing

The success of arrhythmia classification tasks with Machine Learning (ML) algorithms is based on the handcrafting extraction of features from Electrocardiography (ECG) signals. However, feature extraction is a time-consuming trial-and-error approach. Deep Neural Network (DNN) algorithms bypass the process of handcrafting feature extraction since the algorithm extracts the features automatically in their hidden layers. However, it is important to have access to a balanced dataset for algorithm training. In this exploratory research study, we will compare the evaluation metrics among Convolutional Neural Networks (1D-CNN) and Support Vector Machines (SVM) using a dataset based on the merged public ECG signals database TNMG and CINC17 databases. Results: Both algorithms showed good performance using the new, merged ECG database. For evaluation metrics, the 1D-CNN algorithm has a precision of 93.04%, an accuracy of 93.07%, a recall of 93.20%, and an F1-score of 93.05%. The SVM classifier (λ = 10, C = 10 × 109) achieved the best classification metrics with two combined, handcrafted feature extraction methods: Wavelet transforms and R-peak Interval features, which achieved an overall precision of 89.04%, accuracy of 92.00%, recall of 94.20%, and F1-score of 91.54%. As an unique input feature and SVM (λ=10,C=100), wavelet transforms achieved precision, accuracy, recall, and F1-score metrics of 86.15%, 85.33%, 81.16%, and 83.58%. Conclusion: Researchers face a challenge in finding a broad dataset to evaluate ML models. One way to solve this problem, especially for deep learning models, is to combine several public datasets to increase the amount of data. The SVM and 1D-CNN algorithms showed positive results with the merge of databases, showing similar F1-score, precision, and recall during arrhythmia classification. Despite the favorable results for both of them, it should be considered that in the SVM, feature selection is a time-consuming trial-and-error process; meanwhile, CNN algorithms can reduce the workload significantly. The disadvantage of CNN algorithms is that it has a higher computational processing cost; moreover, in the absence of access to powerful computational processing, the SVM can be a reliable solution.

Support Vector Machine (SVM) is a machine learning algorithm based on the Statistical Learning Theory (SLT), which can get good classification effects with a few learning samples. SVM represents a new approach to pattern classification and has been shown to be particularly successful in many fields such as image identification and face recognition. It also provides us with a new method to develop intelligent fault diagnosis. This paper presents a SVM based approach for fault diagnosis of rolling bearings. Experimentation with vibration signals of bearing was conducted. The vibration signals acquired from the bearings were directly used in the calculating without the preprocessing of extracting its features. Compared with the Artificial Neural Network (ANN) based method, the SVM based method has desirable advantages. Also a multi-faults SVM classifier based on binary classifiers is constructed for gear faults in this paper. Other experiments with gear fault samples showed that the multi-fault SVM classifier has good classification ability and high efficiency, and suitable for on line diagnosis for mechanical system.