No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Motor Fault Diagnosis Using CNN Based Deep Learning Algorithm Considering Motor Rotating Speed
... The row (i), column (j), and layer (m) of the feature map are denoted by variables i, j, and m, respectively. The output of this operation is then standardized using a non-linear function (f ) as described in references [26,27]. After the feature map is extracted, it is passed through a pooling layer which shrinks the input and reduces computational load and memory usage. ...
... Finally, the input is classified using a fully connected layer and an activation function (e.g., SoftMax, ReLu, etc ). The SoftMax function layer calculates the probability of the input data belonging to the machine learning state labeled class [26,28]. The SoftMax function is frequently employed as the activation function in a multi-class classifier's output layer, with K denoting the number of classes. ...
This paper proposes a neural-network-based framework using Convolutional Neural Network and Long-Short Term Memory (CNN-LSTM) for detecting faults and recovering signals from Hall sensors in brushless DC motors. Hall sensors are critical components in determining the position and speed of motors, and faults in these sensors can disrupt their normal operation. Traditional fault-diagnosis methods, such as state-sensitive and transition-sensitive approaches, and fault-recovery methods, such as vector tracking observer, have been widely used in the industry but can be inflexible when applied to different models. The proposed fault diagnosis using the CNN-LSTM model was trained on the signal sequences of Hall sensors and can effectively distinguish between normal and faulty signals, achieving an accuracy of the fault-diagnosis system of around 99.3% for identifying the type of fault. Additionally, the proposed fault recovery using the CNN-LSTM model was trained on the signal sequences of Hall sensors and the output of the fault-detection system, achieving an efficiency of determining the position of the phase in the sequence of the Hall sensor signal at around 97%. This work has three main contributions: (1) a CNN-LSTM neural network structure is proposed to be implemented in both the fault-diagnosis and fault-recovery systems for efficient learning and feature extraction from the Hall sensor data. (2) The proposed fault-diagnosis system is equipped with a sensitive and accurate fault-diagnosis system that can achieve an accuracy exceeding 98%. (3) The proposed fault-recovery system is capable of recovering the position in the sequence states of the Hall sensors, achieving an accuracy of 95% or higher.
... To avoid the delicate and skillful parameter tuning, the other way is using adaptive learning rate methods, such as Adagrad [13], RMSprop [14], and Adam [15], in which only the initial learning rate needs to be predefined. However, as shown in our experiments and other studies [16][17][18], they are sensitive to the initial learning rate and each of them has its own effective intervals (refer to Figure 1). Usually, the setting of an initial learning rate is model-and dataset-dependent. ...
... This aims to find the optimal learning rate values and includes experimental performance analysis [18,32] and Bayesian optimization [33,34] based on mathematical theory. Specifically, [35] combines hyperband and Bayesian optimizations, additionally utilizing the history information of previous explored hyperparameter configurations to improve model utility. ...
Due to powerful data representation ability, deep learning has dramatically improved the state-of-the-art in many practical applications. However, the utility highly depends on fine-tuning of hyper-parameters, including learning rate, batch size, and network initialization. Although many first-order adaptive methods (e.g., Adam, Adagrad) have been proposed to adjust learning rate based on gradients, they are susceptible to the initial learning rate and network architecture. Therefore, the main challenge of using deep learning in practice is how to reduce the cost of tuning hyper-parameters. To address this, we propose a heuristic zeroth-order learning rate method, Adacomp, which adaptively adjusts the learning rate based only on values of the loss function. The main idea is that Adacomp penalizes large learning rates to ensure the convergence and compensates small learning rates to accelerate the training process. Therefore, Adacomp is robust to the initial learning rate. Extensive experiments, including comparison to six typically adaptive methods (Momentum, Adagrad, RMSprop, Adadelta, Adam, and Adamax) on several benchmark datasets for image classification tasks (MNIST, KMNIST, Fashion-MNIST, CIFAR-10, and CIFAR-100), were conducted. Experimental results show that Adacomp is not only robust to the initial learning rate but also to the network architecture, network initialization, and batch size.
... Its application leads to the development of prognostics, which allows for the estimation of the system's future health and the prediction of the remaining useful life of the system or system's components [5][6][7][8]. Motors are the backbone of industry; they start degrading due to different reasons such as long period of operation, variations of power supply, or harsh environment; which gradually lead to permanent damage [9][10][11]. Consequently, it becomes crucial to monitor the operation continuously. ...
... Strengths and drawbacks of DL models.(i) Can work with clean, balanced, and scaled data, regardless of the data type (ii) Integration in real-time systems is easy and allows one-dimensional data analysis (i) Requires a lot of tuning to work on dirty or unscaled data Allows for one-dimensional data analysis (ii) Combined optimization of all the layer parameters (i) Slower training than DBN and inefficient (ii) Combined optimization becomes impractical for large data CNN[10,83] (i) Fits well for multidimensional data analysis (ii) Enables for feature extraction from raw data (i) Complex architecture (ii) Requires large datasets and takes long training time (iii) Estimations of continuous data are poor RNN/LSTM/GRU ...
The advent of deep learning (DL) has transformed diagnosis and prognosis techniques in industry. It has allowed tremendous progress in industrial diagnostics, has been playing a pivotal role in maintaining and sustaining Industry 4.0, and also paving the way for industry 5.0. It has become prevalent in the condition monitoring of industrial subsystems, a prime example being motors. Motors in various applications start deteriorating due to various reasons. Thus, the monitoring of their condition is of prime importance for sustaining the operation and maintaining efficiency. This paper presents a state-of-the-art review of DL-based condition monitoring for motors in terms of input data and feature processing techniques. Particularly, it reviews the application of various input features for effectiveness of DL models in motor condition monitoring in the sense of what problems are targeted using these feature processing techniques and how they are addressed. Furthermore, it discusses and reviews advances in DL-models, DL based diagnostic methods for motors, hybrid fault diagnostic techniques, points out important open challenges to these models, and signposts the prospective future directions for DL models. This review will assist researchers in identifying research gaps related to feature processing, so that, they may effectively contribute towards the implementation of DL models as applied to motor condition monitoring.
... In the case of deep networks, in addition to the basic parameters such as the number of neurones in layers or the number of layers, there are additional hyperparameters in the tuning process, e.g., the number of kernels in convolutional layers, kernel step, dropout rate, padding, and many others. In the case of networks with moderately complex structure, researchers decide to empirically select network parameters [42], but with increasing network complexity, empirical selection of parameters requires a lot of experience and intuition preceded by many years of research into the deep structures of neural networks in issues related to the diagnostics of electrical machines. In order to simplify the tuning of deep neural network models in the diagnosis of induction machines, the authors present an approach using the grid search method. ...
The objective of the investigation was to increase the effectiveness of damage detection in the stator of the squirrel-cage induction machine. The analysis aimed to enhance the operational trustworthiness of the squirrel-cage induction machine by employing nonintrusive diagnostic methods based on a current signal and modern artificial intelligence methods. The authors of the study introduced a diagnostic technique for identifying multiphase interturn short circuits of stator winding. These short circuits are one of the most common faults in induction machines. The proposed method focusses on deriving a diagnostic signal from the phase-current waveforms of the machine. The noninvasive nature of the diagnostic technique presented is attributed to the application of the field model of electromagnetic phenomena to determine the diagnostic signal. For this purpose, a field model of a squirrel-cage machine was developed. The waveforms of phase currents obtained from the field model were used as input into an elaborated machine failure neural classifier. A deep neural network was used to develop a neural classifier. The effectiveness of the developed classifier has been experimentally verified, and the obtained results have been presented, concluded, and discussed. The scientific novelty presented in the article is the presentation of research results on the use of a neural classifier to detect damage in all phases of the stator winding at an early stage of its appearance. The features of this type of damage are very difficult to observe in signal waveforms such as a phase current or torque.
... The essence of intelligent control lies in autonomy, mimicking human thought processes to achieve intelligent control. For instance, Han et al. [12] achieved precise observation of electromagnetic torque using a BP neural network. Chaoui and others [13] proposed a continuously adaptive RBF network for speed control of interior permanent-magnet synchronous motors, simplifying the control structure while enhancing control accuracy. ...
The permanent-magnet synchronous motor (PMSM), with the advantages of low energy consumption and stable operation, is considered a green power source to replace gasoline engines. Motor control is the core problem of the electric-drive system, so it is important to study the high-performance motor control algorithm. The traditional PMSM control strategy has problems such as torque pulsation, large overshoot, and parameters which are not easy to adjust. This work proposes a new model-predictive torque control (MPTC) based on multi-objective ranking for these issues. The Romberg observer was utilized to accurately estimate motor flux and torque across a wide range of speeds and ensure optimal performance of the MPTC. The optional voltage vectors were classified using graph theory. The model’s cost function was optimized and the control delay caused by hardware processing was compensated by a modified Euler method. A multi-objective ranking method was used to avoid the offline selection of MPTC weight coefficients. Additionally, one ranking method was used to reduce the complexity of the algorithm for multiple objectives. Based on the simulation results, the newly proposed MPTC method, when compared with traditional approaches, reduced the total harmonic distortion from 2.78% to 2.26%. Torque ripple decreased by approximately 58.4%, and the switching frequency was reduced by 3.05%, lowering the inverter’s switching losses. Therefore, the newly proposed MPTC had faster torque response, reduced computation time, and less torque pulsation, which further improved the dynamic performance of the permanent-magnet synchronous motor.
... Additionally, insufficient fault data is a great challenge in fault diagnosis modeling using ML. DT can generate faulty samples to address the unbalanced number of fault types instead of traditional modeling using over-sampling [120] and under-sampling [121] to improve model accuracy. With the advance of deep transfer learning, few-shot fault diagnosis can be achieved, while it still cannot address the unseen fault diagnosis challenge. ...
The recent advance of digital twin (DT) has greatly facilitated the development of predictive maintenance (PdM). DT for PdM enables accurate equipment status recognition and proactive fault prediction, enhancing reliability. This shift from reactive to proactive services optimizes maintenance schedules, minimizes downtime, and improves enterprise profitability and competitiveness. However, the research and application of DT for PdM are still in their infancy, probably because the role and function of machine learning (ML) in DT for PdM have not yet been fully investigated by the industry and academia. This paper focuses on a systematic review of the role of ML in DT for PdM and identifies, evaluates and analyses a clear and systematic approach to the published literature relevant to DT and PdM. Subsequently, the state-of-the-art applications of ML in various application areas of DT for PdM are introduced. Finally, the challenges and opportunities of ML for DT-PdM are revealed and discussed. The outcome of this paper can bring tangible benefits to the research and implementation of ML in DT-PdM.
... CNN based method is used for online fault detection of motors in industries as shown in [17]. High dimensional CNN performing both feature extraction and classification have been used in [18] to detect faults. Here, backpropagation algorithm has been used to update the filters at every epoch. ...
Single-phase induction motors (SPIM) are fairly efficiently employed in household devices and possess wide applications in the industrial system. Typically, current signature-analysis (CSA) based methods have been used for fault diagnosis, but not always been an efficient solution, especially for fractional kW SPIMs that may be replaced entirely if faulty. This paper presents a deep learning (DL) based fault detection scheme of a SPIM which is in-service in household devices. The fault detection is done employing a wavelet based deep dense convolutional neural network (W-D2CNN), where several wavelets are used as convolutional filters for convolution operation. The convoluted feature maps are further passed through a proposed nonlinearity, i.e., SL-ReLU, for richer feature extraction. The uniqueness of the proposed method lies in the feature learning ability, where upper-and lower-layer features of the dense block fuse via concatenation operation. Also, the SL-ReLU has been designed to combine logarithmic function, softsign, and ReLU operation. The designed SL-ReLU minimizes the probability of convoluted output and gradient explosion, due to which training has been improved along with feature learning ability. Comprehensive results reveal the efficacy of the proposed work in terms of performance metrics with state-of-the-art DL frameworks using the same dataset. The proposed work can also be extendable to multiphase-induction machines or any other electrical device working in any environment for fault identification.
... The stator, rotor, and bearing of one of the motors are artificially damaged to simulate the possible failure of the parts caused by the motor in the running state. Meanwhile, according to the six fault types set in the paper (the stator fault, rotor fault, bearing outer ring fault, bearing inner ring fault, bearing rolling body fault, and the synthetic fault of the front and rear bearings [21]), the faulty parts are, respectively, replaced to another motor that works normally, to collect data on the six fault states and normal states of the motor. Each fault type corresponds to its label, as shown in Table 3. ...
The motor is the primary impetus source of most mechanical equipment, and its failure will cause substantial economic losses and safety problems. Therefore, it is necessary to study online fault diagnosis techniques for motors, given the problems caused by shallow learning models or single-sensor fault analysis in previous motor fault diagnosis techniques, such as blurred fault features, inaccurate identification, and time and manpower consumption. In this paper, we proposed a model for motor fault diagnosis based on deep learning and multi-sensor information fusion. Firstly, a correlation adaptive weighting method is proposed in this paper, and it is used to integrate the collected multi-source homogeneous sensor information into multi-source heterogeneous sensor information through the data layer fusion. Secondly, the 1D-CNN is used to carry out feature extraction, feature layer fusion, and fault classification of multi-source heterogeneous information of the motor. Finally, the data of seven states (one healthy and six faulty) of the motor are collected by the motor drive test bench to realize the model’s training, testing, and verification. The experimental results show that the fault diagnosis accuracy of the model is 99.3%. Thus, this method has important practical implications for improving the accuracy of motor fault diagnosis further.
... There are many recent papers dealing with the AI-based diagnosis of emotors [1], [2], [3], [4], [5] and others. The authors describe the design of the ANN and provide the success rate of the network evaluation, still they either do not deal with on the edge implementation or mention that the integration is in progress. ...
... Another way is to use adaptive learning rate algorithms such as AdaGrad [20], Adam [21], and others that tune the gradient by applying a (diagonal) preprocessing matrix. However, as our experiments and other studies [22][23][24] indicate, these algorithms are susceptible to the initial learning rate, model, and dataset. This motivates us to design an adaptive algorithm that can reduce the cost of tuning and quickly converge in the over-parameterized models. ...
In recent years, deep learning has dramatically improved state of the art in many practical applications. However, this utility is highly dependent on fine-tuning of hyperparameters, including learning rate, batch size, and network initialization. Although many first-order adaptive gradient algorithms (e.g., Adam, AdaGrad) have been proposed to adjust the learning rate, they are vulnerable to the initial learning rate and network structure in the training over-parameterized models, especially in the dynamic online setting. Therefore, the main challenge of using deep learning in practice is how to reduce the cost of tuning hyperparameters. To address this problem, we integrate the adaptive strategy of Radhakrishnan et al. and the acceleration strategy of Ghadimi et al. to propose a fast adaptive online gradient algorithm, FAOGD. The adaptive strategy we adopt only adjusts the learning rate according to the historical gradient and training loss value, while the acceleration strategy is the heavy-ball momentum used to accelerate the training of deep models. The proposed FAOGD enjoys merit that there is no need to tune hyperparameters related to the learning rate, which thus saves much unnecessary computational overhead. It is also shown that FAOGD can obtain the regret bound of OT\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$O\left( {\sqrt{T} } \right) $$\end{document}, matching the Adam and AdaGrad using the empirical learning rate. Simulation results in the over-parameterized neural networks clearly show that FAOGD outperforms existing algorithms. Furthermore, FAOGD is also robust to network structures and batch size.
... For a relation between the motor speed and vibration signals, [5] proposes a CNN based deep learning approach for automatic motor fault diagnosis. In the same research line [6] establishes a comparison of fault motor diagnosis using RNN (Recurrent Neural Networks) and k-means in vibration analysis. ...
This paper presents an approach for automatic anomaly detection through vibration analysis based on machine learning algorithms. The study focuses on induction motors in a predictive maintenance context, but can be applied to other domains. Vibration analysis is an important diagnostic tool in industrial data analysis to predict anomalies caused by equipment defects or in its use, allowing to increase its lifetime. It is not a new technique and is widely used in the industry, however with the Industry 4.0 paradigm and the need to digitize any process, it gains relevance to automatic fault detection. The Isolation Forest algorithm is implemented to detect anomalies in vibration datasets measured in an experimental apparatus composed of an induction motor and a coupling system with shaft alignment/misalignment capabilities. The results show that it is possible to detect anomalies automatically with a high level of precision and accuracy.KeywordsIndustry 4.0Anomaly detectionIsolation forestVibration analysisBigML
... In pipeline MFL inspection, MFL data analysis can be divided into three parts: data preprocessing, defect identification, and profile inversion [9]. In the defect identification process, defect MFL signals are detected among a large amount of MFL measurements [10]. ...
Magnetic flux leakage (MFL) inspection, one of the nondestructive testing methods, has been widely applied in pipeline maintenance. In pipeline MFL data processing, defect identification is a crucial step, which aims at measuring the locations of defect MFL signals in MFL heat maps. MFL signals collected from the pipeline are not ideal, containing noise and interference. In this case, measuring the locations of defect MFL signals, especially the locations of weak defect MFL signals, is a challenge. To address this challenge, an enhancement process is required to coordinate with the identification process. In this article, two separated processes, enhancement and identification, are integrated into a single-stage framework, aiming at improving the defect identification performance by strengthening the differences between defect signal areas and pipe wall signal areas. The proposed framework can enhance the defect areas purposefully and ignore the noise and interference in nondefect areas, which promotes the measuring effect for locations of defect MFL signals. In the proposed method, an enhancement module is constructed to upsample the MFL heat maps, and the resolution of feature maps in the framework is increased to
$288 \times 600$
. A novel loss function is designed, and the gray value contrast between defect signal areas and pipe wall signal areas in MFL heat maps is enhanced from 10 to 48 approximately through task-oriented joint training. The proposed method achieves 0.967 average precision (AP), and the identification accuracy is improved to 97.3%. In addition, the average deviations of identified defect signals are reduced by around 2–9 mm, and the uncertainties are reduced to 0.27–0.35 mm. The experiment results validate the superiority of the proposed framework in industrial applications.
... The achieved classification accuracy was 98%, 98%, and 100% for the IMs under normal, rotor fault, and bearing fault conditions. A CNN based method to diagnose faults in IMs taking into account the motor speed (Han, Choi, Hong & Kim, 2019). The vibration signal was used as input to a CNN. ...
Purpose: Developments in Industry 4.0 technologies and Artificial Intelligence (AI) have enabled data-driven manufacturing. Predictive maintenance (PdM) has therefore become the prominent approach for fault detection and diagnosis (FD/D) of induction motors (IMs). The maintenance and early FD/D of IMs are critical processes, considering that they constitute the main power source in the industrial production environment. Machine learning (ML) methods have enhanced the performance and reliability of PdM. Various deep learning (DL) based FD/D methods have emerged in recent years, providing automatic feature engineering and learning and thereby alleviating drawbacks of traditional ML based methods. This paper presents a comprehensive survey of ML and DL based FD/D methods of IMs that have emerged since 2015. An overview of the main DL architectures used for this purpose is also presented. A discussion of the recent trends is given as well as future directions for research.
Design/methodology/approach: A comprehensive survey has been carried out through all available publication databases using related keywords. Classification of the reviewed works has been done according to the main ML and DL techniques and algorithms.
Findings: DL based PdM methods have been mainly introduced and implemented for IM fault diagnosis in recent years. Novel DL FD/D methods are based on single DL techniques as well as hybrid techniques. DL methods have also been used for signal preprocessing and moreover, have been combined with traditional ML algorithms to enhance the FD/D performance in feature engineering. Publicly available datasets have been mostly used to test the performance of the developed methods, however industrial datasets should become available as well. Multi-agent system (MAS) based PdM employing ML classifiers has been explored. Several methods have investigated multiple IM faults, however, the presence of multiple faults occurring simultaneously has rarely been investigated.
Originality/value: The paper presents a comprehensive review of the recent advances in PdM of IMs based on ML and DL methods that have emerged since 2015.
... Wavelet packet decomposition was applied in combination with SVM to distinguish different types of bearing faults in [40,41]. In [42], A deep learning algorithm was developed for motor fault diagnosis that also keeps in considering motor speed parameters. ...
Accurate and early detection of machine faults is an important step in the preventive maintenance of industrial enterprises. It is essential to avoid unexpected downtime as well as to ensure the reliability of equipment and safety of humans. In the case of rotating machines, significant information about machine’s health and condition is present in the spectrum of its vibration signal. This work proposes a fault detection system of rotating machines using vibration signal analysis. First, a dataset of 3-dimensional vibration signals is acquired from large induction motors representing healthy and faulty states. The signal conditioning is performed using empirical mode decomposition technique. Next, multi-domain feature extraction is done to obtain various combinations of most discriminant temporal and spectral features from the denoised signals. Finally, the classification step is performed with various kernel settings of multiple classifiers including support vector machines, K-nearest neighbors, decision tree and linear discriminant analysis. The classification results demonstrate that a hybrid combination of time and spectral features, classified using support vector machines with Gaussian kernel achieves the best performance with 98.2% accuracy, 96.6% sensitivity, 100% specificity and 1.8% error rate.
... Wavelet packet decomposition was applied in combination with SVM to distinguish different types of bearing faults in [40,41]. In [42], A deep learning algorithm was developed for motor fault diagnosis that also keeps in considering motor speed parameters. ...
Accurate and early detection of machine faults is an important step in the preventive maintenance of industrial enterprises. It is essential to avoid unexpected downtime as well as to ensure the reliability of equipment and safety of humans. In the case of rotating machines, significant information about machine's health and condition is present in the spectrum of its vibration signal. This work proposes a fault detection system of rotating machines using vibration signal analysis. First, a dataset of 3-dimensional vibration signals is acquired from large induction motors representing healthy and faulty states. The signal conditioning is performed using empirical mode decomposition technique. Next, multi-domain feature extraction is done to obtain various combinations of most discriminant temporal and spectral features from the denoised signals. Finally, the classification step is performed with various kernel settings of multiple classifiers including support vector machines, K-nearest neighbors, decision tree and linear discriminant analysis. The classification results demonstrate that a hybrid combination of time and spectral features, classified using support vector machines with Gaussian kernel achieves the best performance with 98.2% accuracy, 96.6% sensitivity, 100% specificity and 1.8% error rate.
... The feature set from motor data has been extracted and can be classified employing CNN. In small scale industries, CNN based deep learning can be used for online motor fault diagnosis [17]. Sometimes, complex algorithms of convolutional discriminative feature-based learning with backpropagation (BP)-based artificial neural network has also been used for precise machine fault diagnosis [18]. ...
... Snoek et al. [27] applied the pseudo-Bayesian neural network to tune the learning rate of another deep neural network, but the learning rate is fixed in all iterations. Han et al. [28] investigated the methods to tune the learning rate, the batch size, the number of epochs, and the momentum coefficient of the CNN methods for motor fault classification. Saufi et al. [29] studied a review on HPO for DL applications, and pointed out several potential algorithms for the learning rate selection. ...
Fault classification is vital in smart manufacturing, and Convolutional Neural Network (CNN) has been widely applied in fault classification. But the performance of CNN heavily depends on its learning rate. As the default setting on learning rate cannot guarantee its performance, the learning rate tuning process becomes essential. However, the traditional learning rate tuning methods either costmuch time consumption or rely on the experts’ experiences, so it is a considerable barrier for the users. To overcome this drawback, this paper proposes a CNN with automatic learning rate scheduler (AutoLR-CNN) for fault classification. Firstly, the Long Short-Term Memory (LSTM) is used to extract the features of the past loss of CNN. Then, an agent based on Deep Deterministic Policy Gradient (DDPG) is trained to automatically control the learning rate for CNN online. Thirdly, the double CNN structure is developed to enhance the stability of the proposed method. The proposed AutoLR-CNN is tested on two famous bearing datasets and a practical bearing dataset on wind turbine. The results of AutoLR-CNN are superior to six common used baseline learning rate schedulers in Tensorflow. AutoLR-CNN is also compared with other reported machine learning and deep learning methods. The results show that AutoLR-CNN has achieved the state-of-the-art performance in fault classification.
Due to the open-circuit fault of neutral point clamped (NPC) inverter having the characteristics of solid concealment and multiple fault modes, it is necessary to study an accurate and efficient fault diagnosis method to detect and locate the fault in time. However, the existing methods are either based on artificial analytical modeling or signal analysis, which is challenging to model and needs better portability or the model needs to be simplified to implement online applications. Therefore, this paper proposes a data-driven fault diagnosis and high-performance online inference method. The diagnosis method simplifies the traditional labeling method by designing labels and takes one-dimensional depth separable convolution (1D-DSC) and global max pooling (GMP) to replace conventional one-dimensional convolution and local pooling called 1D-DSCNN-GMP. Then, the model is compressed and optimized based on the TensorRT framework. Experiments show that compared with the existing methods, the proposed method can reduce the number of model parameters by more than 90%. The average online inference time of the model after compression and deployment to the Edge Computing Board (ECB) is about 1 ms, only 5% fundamental period, and can maintain 100% diagnostic accuracy, with better online application potential.
This paper presents a novel method for diagnosing compressor faults using the Graph Attention Network (GAT). Specifically, we address the challenge of analyzing multivariate time series data generated by compressors. Our proposed method consists of three main steps. Firstly, we construct a temporal graph for each variable of the multivariate time series using the Limited Penetrable Visibility Graph (LPVG) to capture the temporal dependencies within each variable. Subsequently, we feed these temporal graphs into the GAT to obtain a representation of each variable. Secondly, we leverage the inter-variable dependencies by constructing an adaptive inter-variable graph, where each node represents a variable, and the node feature is the previously obtained variable representation. We then input this inter-variable graph into another GAT to further capture the dependencies between variables. Finally, we use the output from the GAT to train a classifier for fault diagnosis, resulting in an end-to-end model. Our proposed method outperforms existing techniques in diagnosing faults in a real dataset.KeywordsCompressor Fault DiagnosisMultivariate Time SeriesLimited Penetrable Visibility GraphGraph Attention Network
Induction Motor (IM) is the most important prime mover in the automotive industry and has great potential in Electric Vehicles (EVs). IM failure is one of the most prevalent reasons for EV failure. This paper presents a fault diagnosis method for IM to enhance the efficiency, performance, and availability of EV and to reduce its maintenance costs. Firstly, current and vibration signatures were acquired at varying speed conditions from four IMs having different fault conditions. The acquired signatures were decomposed using Hilbert Transform (HT) and further converted into the time-frequency domain using Constant-Q Transform (CQT). This time-frequency data were utilized for training the Machine Learning (ML) and Deep Learning (DL) model. A comparative analysis was done in terms of classification accuracies given by the ML and DL models. Eventually, the model performance was also studied for both current and vibration signatures. The experimental finding showed that the DL model has better potential than ML models for IM fault diagnosis under varying operating conditions.
In the fault diagnosis of the existing permanent magnet synchronous motor, the characteristics of the fault are often extracted based on the vibration acceleration signal. The acquisition of vibration acceleration requires the additional installation of expensive sensors, and the diagnosis effect is greatly affected by the surrounding environment. On the other hand, at this time, due to the influence of the load side, the fault characteristics of the bearing and other faults are easily submerged, resulting in a diagnosis failure or even a misdiagnosis. This paper proposes a bearing fault diagnosis method based on the motor speed signal. Combined with the CNN, the feature extraction and analysis of the rotational speed signal are carried out, and an improved algorithm is proposed by combining the artificially selected eigenvalues in the frequency domain. The experimental results show that this method can still complete the diagnosis of bearing faults well in the presence of misalignment fault interference, which shows the potential of deep learning technology represented by CNN in the use of rotational speed signals to diagnose various types of motor faults and provide experimental and theoretical basis for it.
Light food refers to healthy and nutritious food that has the characteristics of low calorie, low fat, and high fiber. Light food has been favored by the public, especially by the young generation in recent years. Moreover, affected by the COVID-19 epidemic, consumers’ awareness of a healthy diet has been improved to a certain extent. As both take-out and in-place orders for light food are growing rapidly, there are massive customer reviews left on the Meituan platform. However, massive, multi-dimensional unstructured data has not yet been fully explored. This research aims to explore the customers’ focal points and sentiment polarity of the overall comments and to investigate whether there exist differences of these two aspects before and after the COVID-19. A total of 6968 light food customer reviews on the Meituan platform were crawled and finally used for data analysis. This research first conducted the fine-grained sentiment analysis and classification of the light food customer reviews via the SnowNLP technique. In addition, LDA topic modeling was used to analyze positive and negative topics of customer reviews. The experimental results were visualized and the research showed that the SnowNLP technique and LDA topic modeling achieve high performance in extracting the customers’ sentiments and focal points, which provides theoretical and data support for light food businesses to improve customer service. This research contributes to the existing research on LDA modeling and light food customer review analysis. Several practical and feasible suggestions are further provided for managers in the light food industry.
In order to meet the requirements of multi-parameter and real-time fault diagnosis of China VI vehicle emission standards diesel engine air system, this paper focuses on the research of the diesel engine air system fault detection and diagnosis method based on deep learning to improve the operational safety and fault diagnosis efficiency. In this paper, the air system fault detection and diagnosis are completed based on the AE model and the CNN model, combined with the actual operation of the diesel engine. Among them, the AE model successfully detected all real-time operation faults, and the false detection rate on the health data was 0.1162%; the CNN model obtained a 90.77% fault diagnosis accuracy rate on the test set of the real-time operation data set. The results show that the model has high accuracy in the diagnosis of diesel engine faults, which is of great significance to the application of deep learning based on big data processing.
Analysis of 1-D vibration signals is the most common method used for safety analysis and health monitoring of rotary machines. How to effectively extract features involved in 1-D sequence data is crucial for the accuracy of real-time fault diagnosis. This chapter aims to develop a more effective means of extracting useful features potentially involved in 1-D vibration signals. First, an improved parallel long short-term memory called PeLSTM is designed by adding a peephole connection before each forget gate to prevent useless information transferring in the cell. It not only can solve the memory bottleneck problem of traditional long short-term memory for long sequence but also can make full use of all possible information helpful for feature extraction. Second, a fusion network with a new training mechanism is designed to fuse features extracted from PeLSTM and the convolutional neural network, respectively. The fusion network can incorporate a 2-D screenshot image into comprehensive feature extraction. It can provide a more accurate fault diagnosis result since the 2-D screenshot image is another form of expression for a 1-D vibration sequence involving additional trend and locality information. Finally, the real-time 2-D screenshot image is fed into the convolutional neural network to secure a real-time online diagnosis, which is the primary requirement of the engineers in health monitoring. Validity of the proposed method is verified by fault diagnosis for rolling bearing and gearbox.
Hydraulic machinery systems are a widely used machine found in hydropower stations. As a result, it is vital that such machinery is monitored, diagnosed, maintained, or replaced prior to failing to reduce downtime and labor costs. Currently, most researchers have begun to investigate condition-based maintenance or a predictive maintenance strategy instead of a time-based maintenance strategy. This chapter reviews the state of the art in diagnostics and prognostics pertaining to hydraulic machinery systems. Attention is given to detailing the application status of sensor detection technology, cavitation research, intelligent evaluation and diagnosis technology, and prognostics research, among others, used by researchers in the main areas of diagnostics and prognostics.
The need to manufacture more competitive equipment, together with the emergence of the digital technologies from the so-called Industry 4.0, have changed many paradigms of the industrial sector. Presently, the trend has shifted to massively acquire operational data, which can be processed to extract really valuable information with the help of Machine Learning or Deep Learning techniques. As a result, classical Condition Monitoring methodologies, such as model- and signal-based ones are being overcome by data-driven approaches. Therefore, the current paper provides a review of these data-driven active supervision strategies implemented in electric drives for fault detection and diagnosis (FDD). Hence, first, an overview of the main FDD methods is presented. Then, some basic guidelines to implement the Machine Learning workflow on which most data-driven strategies are based, are explained. In addition, finally, the review of scientific articles related to the topic is provided, together with a discussion which tries to identify the main research gaps and opportunities
Fault diagnosis of various rotating equipment plays a significant role in industries as it guarantees safety, reliability and prevents breakdown and loss of any source of energy. Early identification is a fundamental aspect for diagnosing the faults which saves both time and costs and in fact it avoids perilous conditions. Investigations are being carried out for intelligent fault diagnosis using machine learning approaches. This article analyses various machine learning approaches used for fault diagnosis of rotating equipment. In addition to this, a detailed study of different machine learning strategies which are incorporated on various rotating equipment in the context of fault diagnosis is also carried out. Mainly, the benefits and advance patterns of deep neural network which are applied to multiple components for fault diagnosis are inspected in this study. Finally, different algorithms are proposed to propagate the quality of fault diagnosis and the conceivable research ideas of applying machine learning approaches on various rotating equipment are condensed in this article.
Analysis of one-dimensional vibration signals is the most common method used for safety analysis and health monitoring of rotary machines. How to effectively extract features involved in one-dimensional sequence data is crucial for the accuracy of real-time fault diagnosis. This article aims to develop more effective means of extracting useful features potentially involved in one-dimensional vibration signals. First, an improved parallel long short-term memory called parallel long short-term memory with peephole is designed by adding a peephole connection before each forget gate to prevent useless information transferring in the cell. It can not only solve the memory bottleneck problem of traditional long short-term memory for long sequence but also can make full use of all possible information helpful for feature extraction. Second, a fusion network with new training mechanism is designed to fuse features extracted from parallel long short-term memory with peephole and convolutional neural network, respectively. The fusion network can incorporate two-dimensional screenshot image into comprehensive feature extraction. It can provide more accurate fault diagnosis result since two-dimensional screenshot image is another form of expression for one-dimensional vibration sequence involving additional trend and locality information. Finally, real-time two-dimensional screenshot image is fed into convolutional neural network to secure a real-time online diagnosis which is the primary requirement of the engineers in health monitoring. Validity of the proposed method is verified by fault diagnosis for rolling bearing and gearbox.
Hyperparameter optimization is a very difficult problem in developing deep learning algorithms. In this paper, a genetic algorithm was applied to solve this problem. The accuracy and the verification time were considered by conducting a fitness evaluation. The algorithm was evaluated by using a simple model that has a single convolution layer and a single fully connected layer. A model with three layers was used. The MNIST dataset and a motor fault diagnosis dataset were used to train the algorithm. The results show that the method is useful for reducing the training time.
Feedforward neural networks trained by error backpropagation are examples of nonparametric regression estimators. We present a tutorial on nonparametric inference and its relation to neural networks, and we use the statistical viewpoint to highlight strengths and weaknesses of neural models. We illustrate the main points with some recognition experiments involving artificial data as well as handwritten numerals. In way of conclusion, we suggest that current-generation feedforward neural networks are largely inadequate for difficult problems in machine perception and machine learning, regardless of parallel-versus-serial hardware or other implementation issues. Furthermore, we suggest that the fundamental challenges in neural modeling are about representation rather than learning per se. This last point is supported by additional experiments with handwritten numerals.
In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively.
Deep neural nets with a large number of parameters are very powerful machine learning systems. However, overfitting is a serious problem in such networks. Large networks are also slow to use, making it difficult to deal with overfitting by combining the predictions of many different large neural nets at test time. Dropout is a technique for addressing this problem. The key idea is to randomly drop units (along with their connections) from the neural network during training. This prevents units from co-adapting too much. During training, dropout samples from an exponential number of different "thinned" networks. At test time, it is easy to approximate the effect of averaging the predictions of all these thinned networks by simply using a single unthinned network that has smaller weights. This significantly reduces overfitting and gives major improvements over other regularization methods. We show that dropout improves the performance of neural networks on supervised learning tasks in vision, speech recognition, document classification and computational biology, obtaining state-of-the-art results on many benchmark data sets. © 2014 Nitish Srivastava, Geoffrey Hinton, Alex Krizhevsky, Ilya Sutskever and Ruslan Salakhutdinov.
This paper presents a novel hybrid model for fault detection and classification of motor bearing. In the proposed model, permutation entropy (PE) of the vibration signal is calculated to detect the malfunctions of the bearing. If the bearing has faults, the vibration signal is decomposed into a set of intrinsic mode functions (IMFs) by ensemble empirical mode decomposition (EEMD). The PE values of the first several IMFs (IMF-PE) are calculate to reveal the multi-scale intrinsic characteristics of the vibration signal. Then, support vector machines (SVM) optimized by inter-cluster distance (ICD) in the feature space (ICDSVM) is used to classify the fault type as well as fault severity. Finally, the proposed model is fully evaluated by experiments and comparative studies. The results demonstrate its effectiveness and robustness for motor bearing fault detection and classification.
Vibration inspection of
electro-mechanical components and systems is an
important tool for automated reliable online as well as
post-process production quality
assurance. Considering that bad electromotor samples are
very rare in the production line, we propose a novel
automated fault detection method named “Tilear”, based on
Deep Belief Networks (DBNs) training only with good electromotor
samples. Tilear consctructs an
auto-encoder with DBNs, aiming to
reconstruct the inputs as closely as possible. Tilear is
structured in two parts: training and
decision-making. During
training, Tilear is trained only with informative features
extracted from preprocessed vibration signals of good
electromotors, which enables the trained Tilear only to know
how to reconstruct good electromotor vibration signal
features. In the decision-making
part, comparing the recorded signal from test
electromotor and the Tilear reconstructed signal, allows to
measure how well a recording from a test electromotor matches the Tilear model
learned from good electromotors. A reliable decision can
be made.
In this work we investigate the effect of the convolutional network depth on
its accuracy in the large-scale image recognition setting. Our main
contribution is a thorough evaluation of networks of increasing depth, which
shows that a significant improvement on the prior-art configurations can be
achieved by pushing the depth to 16-19 weight layers. These findings were the
basis of our ImageNet Challenge 2014 submission, where our team secured the
first and the second places in the localisation and classification tracks
respectively.
Grid search and manual search are the most widely used strategies for hyper-parameter optimization. This paper shows empirically and theoretically that randomly chosen trials are more efficient for hyper-parameter optimization than trials on a grid. Empirical evidence comes from a comparison with a large previous study that used grid search and manual search to configure neural networks and deep belief networks. Compared with neural networks configured by a pure grid search, we find that random search over the same domain is able to find models that are as good or better within a small fraction of the computation time. Granting random search the same computational budget, random search finds better models by effectively searching a larger, less promising configuration space. Compared with deep belief networks configured by a thoughtful combination of manual search and grid search, purely random search over the same 32-dimensional configuration space found statistically equal performance on four of seven data sets, and superior performance on one of seven. A Gaussian process analysis of the function from hyper-parameters to validation set performance reveals that for most data sets only a few of the hyper-parameters really matter, but that different hyper-parameters are important on different data sets. This phenomenon makes grid search a poor choice for configuring algorithms for new data sets. Our analysis casts some light on why recent "High Throughput" methods achieve surprising success--they appear to search through a large number of hyper-parameters because most hyper-parameters do not matter much. We anticipate that growing interest in large hierarchical models will place an increasing burden on techniques for hyper-parameter optimization; this work shows that random search is a natural baseline against which to judge progress in the development of adaptive (sequential) hyper-parameter optimization algorithms.
Many intelligent learning methods have been successfully applied in gearbox fault diagnosis. Among them, self-organizing maps (SOMs) have been used effectively as they preserve the topological relationships of data. However, the structures of data clusters learned by SOMs may not be apparent and their shapes are often distorted. This paper presents a semisupervised diagnosis method based on a distance-preserving SOM for machine-fault detection and classification, which can also be used to visualize the SOM learning results directly. An experimental study performed on a gearbox and bearings indicated that the developed approach is effective in detecting incipient gear-pitting failure and classifying different bearing defects and levels of ball-bearing defects.
Machine learning algorithms frequently require careful tuning of model
hyperparameters, regularization terms, and optimization parameters.
Unfortunately, this tuning is often a "black art" that requires expert
experience, unwritten rules of thumb, or sometimes brute-force search. Much
more appealing is the idea of developing automatic approaches which can
optimize the performance of a given learning algorithm to the task at hand. In
this work, we consider the automatic tuning problem within the framework of
Bayesian optimization, in which a learning algorithm's generalization
performance is modeled as a sample from a Gaussian process (GP). The tractable
posterior distribution induced by the GP leads to efficient use of the
information gathered by previous experiments, enabling optimal choices about
what parameters to try next. Here we show how the effects of the Gaussian
process prior and the associated inference procedure can have a large impact on
the success or failure of Bayesian optimization. We show that thoughtful
choices can lead to results that exceed expert-level performance in tuning
machine learning algorithms. We also describe new algorithms that take into
account the variable cost (duration) of learning experiments and that can
leverage the presence of multiple cores for parallel experimentation. We show
that these proposed algorithms improve on previous automatic procedures and can
reach or surpass human expert-level optimization on a diverse set of
contemporary algorithms including latent Dirichlet allocation, structured SVMs
and convolutional neural networks.
Data Acquisition System for Real Time Motor Fault Diagnosis and Deep Learning Process
- D J Choi
- J H Han
- H S Kim
- M J Choi
- C W Kwon
- S K Hong
Data Acquisition System for Real Time Motor Fault Diagnosis and Deep Learning Process
- choi