No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Fault Detection and Severity Identification of Ball Bearings by Online Condition Monitoring
Abstract
The paper presents a fast, accurate and simple systematic approach for online condition monitoring and severity identification of ball bearings. This approach utilizes compact 1D convolutional neural networks (CNNs) to identify, quantify, and localize bearing damage. The proposed approach is verified experimentally under several single and multiple damage scenarios. The experimental results demonstrated that the proposed approach can achieve a high level of accuracy for damage detection, localization and quantification. Besides its real-time processing ability and superior robustness against the high-level noise presence, the compact and minimally-trained 1D CNNs in the core of the proposed approach can handle new damage scenarios with utmost accuracy.
... Abdeljaber et al. [11] presented a one-dimensional convolutional network for the detection, classification, and assessment of the severity of ball bearing faults. The proposed network was trained and then tested on scenarios involving multiple ball bearing faults simultaneously. ...
... As mentioned in the previous sections, CNNs are designed to exclusively operate on 2D data such as images and videos. To feed a CNN network with 1D data, like voice and data series, weather forecast data, vibration measurements, traffic flow, and electrocardiogram signals [33], different techniques, like reshaping, have been utilized to transform the 1D signal into a 2D representation [11]. However, 2D CNN networks, especially ones developed with deep architecture that have more than 1 M (usually above 10 M) parameters, exhibit high computational complexity [15]. ...
All kinds of vessels consist of dozens of complex machineries with rotating parts and electric motors that operate continuously in harsh environments with excess temperature, humidity, vibration, fatigue, and load. A breakdown or malfunction in one of these machineries can significantly impact a vessel’s operation and safety and, consequently, the safety of the crew and the environment. To maintain operational efficiency and seaworthiness, the shipping industry invests substantial resources in preventive maintenance and repairs. This study presents the economic and technical benefits of predictive maintenance over traditional preventive maintenance and repair by replacement approaches in the maritime domain. By leveraging modern technology and artificial intelligence, we can analyze the operating conditions of machinery by obtaining measurements either from sensors permanently installed on the machinery or by utilizing portable measuring instruments. This facilitates the early identification of potential damage, thereby enabling efficient strategizing for future maintenance and repair endeavors. In this paper, we propose and develop a convolutional neural network that is fed with raw vibration measurements acquired in a laboratory environment from the ball bearings of a motor. Then, we investigate whether the proposed network can accurately detect the functional state of ball bearings and categorize any possible failures present, contributing to improved maintenance practices in the shipping industry.
... However, 1-D convolutional layers process the input with a single operation, Sensors 2023, 23, 6223 7 of 15 therefore taking less time to learn features compared to recurrent layers that iterate over the time steps of the input [28]. Additionally, 1D-CNNs have been extensively deployed in classification models for sequences because of their advantages over 2D-CNNs, for instance, less-complicated configurations and hyperparameters, shallower architectures that make them easier to implement, smaller hardware setup, and low cost [29][30][31][32]. The 1D-CNN architecture employed in this research is illustrated in Figure 7. ...
... However, 1-D convolutional layers process the input with a single operation, therefore taking less time to learn features compared to recurrent layers that iterate over the time steps of the input [28]. Additionally, 1D-CNNs have been extensively deployed in classification models for sequences because of their advantages over 2D-CNNs, for instance, less-complicated configurations and hyperparameters, shallower architectures that make them easier to implement, smaller hardware setup, and low cost [29][30][31][32]. The 1D-CNN architecture employed in this research is illustrated in Figure 7. ...
In recent years, human activity recognition (HAR) has gained significant interest from researchers in the sports and fitness industries. In this study, the authors have proposed a cascaded method including two classifying stages to classify fitness exercises, utilizing a decision tree as the first stage and a one-dimension convolutional neural network as the second stage. The data acquisition was carried out by five participants performing exercises while wearing an inertial measurement unit sensor attached to a wristband on their wrists. However, only data acquired along the z-axis of the IMU accelerator was used as input to train and test the proposed model, to simplify the model and optimize the training time while still achieving good performance. To examine the efficiency of the proposed method, the authors compared the performance of the cascaded model and the conventional 1D-CNN model. The obtained results showed an overall improvement in the accuracy of exercise classification by the proposed model, which was approximately 92%, compared to 82.4% for the 1D-CNN model. In addition, the authors suggested and evaluated two methods to optimize the clustering outcome of the first stage in the cascaded model. This research demonstrates that the proposed model, with advantages in terms of training time and computational cost, is able to classify fitness workouts with high performance. Therefore, with further development, it can be applied in various real-time HAR applications.
... The low computational requirements of compact 1D CNNs make them ideal for low-cost and real-time applications [53][54][55][56][57][58][59][60]. ...
The main goal of this paper is to introduce a Motor Imagery (MI) classification system for electroencephalography (EEG) that is extremely precise. To achieve this goal, we propose using a feature-extracted deep one-dimension (1D) convolutional neural network (CNN) which provides a model that can be further improved through hyperheuristic multi-objective evolutionary search. We can improve the classification performance by training this deep CNN model with feature-extracted data from the Physionet MI dataset. We also present a semi-deep fine-tuning approach that can yield improvements with just four epochs. Our findings using the Physionet MI dataset illustrate that the approach we suggest surpasses most contemporary techniques used for classifying EEG signals. Our system is computationally efficient and can be trained using reliable EEG data for individual patients, allowing for accurate classification of their EEG records. Because of its straightforward and parameter-independent characteristics, our system is versatile and can be utilized with any EEG dataset.
... In the same direction, Abdeljaber et al [243] utilized compact convolutional neural networks for identifying, quantifying and localizing the damage in ball bearings. Experimental works were performed to verify the accuracy of the proposed approach. ...
Faults in rotating systems can cause significant damage to the machinery and can result in downtime and production losses. Hence, the timely detection and diagnosis of faults are very important for the smooth running of machines and the assurance of their safety and reliability. In view of this, a review of the literature has been presented in the article on the types of additive faults and their identification using conventional signal-based techniques and automated artificial intelligence techniques. Through a literature survey,
the faulty rigid and flexible rotor systems mounted on rolling element bearings, hydrodynamic bearings, and active magnetic bearings have been studied. The faults incorporated in this article are the additive fault types, in which the process is affected by adding process variables. The rotor unbalances, shaft or bearing misalignment, crack, internal damping, bow in the shaft, rotor-to-stator rub, and mechanical looseness are the classifications of additive faults. Additionally, understanding the rotor response through theoretical and experimental investigations influenced by the additive faults and its detection and diagnosis using vibration and current-induced signals is extremely important, and therefore the present paper briefly discusses this. Following the state of the art in the dynamic analysis and identification of multiple hazardous faults, the general remarks and future directions for further research have been suggested at the end of this article.
... A convolutional neural network (CNN) is a network architecture for deep learning that learns directly from two-dimensional image input. As an end-to-end diagnostic model, it is understood from the literature that it achieves a higher performance at a lower cost [36]. A one-dimensional CNN (1D-CNN) has a less complex network structure and lower computational complexity than a two-dimensional CNN (2D-CNN) and accepts one-dimensional raw data as input directly without any preprocessing. ...
Predictive maintenance (PdM) is implemented to efficiently manage maintenance schedules of machinery and equipment in manufacturing by predicting potential faults with advanced technologies such as sensors, data analysis, and machine learning algorithms. This paper introduces a study of different methodologies for automatically classifying the failures in PdM data. We first present the performance evaluation of fault classification performed by shallow machine learning (SML) methods such as Decision Trees, Support Vector Machines, k-Nearest Neighbors, and one-dimensional deep learning (DL) techniques like 1D-LeNet, 1D-AlexNet, and 1D-VGG16. Then, we apply normalization, which is a scaling technique in which features are shifted and rescaled in the dataset. We reapply classification algorithms to the normalized dataset and present the performance tables in comparison with the first results we obtained. Moreover, in contrast to existing studies in the literature, we generate balanced dataset groups by randomly selecting normal data and all faulty data for all fault types from the original dataset. The dataset groups are generated with 100 different repetitions, recording performance scores for each one and presenting the maximum scores. All methods utilized in the study are similarly employed on these groups. From these scores, the use of 1D-LeNet deep learning classifiers and feature normalization resulted in achieving the highest overall accuracy and F1-score performance of 98.50% and 98.32%, respectively. As a result, the goal of this study was to develop an efficient approach for automatic fault classification, leveraging data balance, and additionally, to provide an analysis of one-dimensional deep learning and shallow machine learning-based classification methods. In light of the experimentation and comparative analysis, this study successfully achieves its stated goal by demonstrating that one-dimensional deep learning and data balance collectively emerge as the optimal approach, offering good prediction accuracy.
... The integration of advanced machine learning technologies with vibration analysis has ushered in a new era for bearing Outer race Ball Inner race Cage fault diagnosis [1]. Specifically, leveraging Convolutional Neural Networks (CNN) was found to be a promising approach for detecting and identifying bearing defects at their early stages; thereby, enabling proactive maintenance strategies and ultimately reducing the economic and operational impact of unexpected failures [8], [9], [10]. Unfortunately, current solutions are usually designed to work in a very restricted setting and do not integrate the varying conditions and operations found in a real-life scenario, e.g., heavy noise and time-varying rotational speeds. ...
... Numerous methodologies have focused on the vibration signals to detect and identify the bearing faults. They can be classified as model-based methods [1], [2], [3], [4], signalprocessing approaches [5], [6], [7], [8], [9], [10], [11], [12], conventional machine learning (ML), and recent deep learning (DL) methods [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28]. Especially during the last decade, DL-based methods based on the vibration signal have increased tremendously. ...
Robust and real-time detection of faults has become an ultimate objective for predictive maintenance on rotating machinery. Vibration-based Deep Learning (DL) methodologies have become the
de facto
standard for bearing fault detection as they can produce state-of-the-art detection performances under certain conditions. Despite such particular focus on the vibration signal, the utilization of sound, on the other hand, has been widely neglected. As a result, no large-scale benchmark motor fault dataset exists with both sound and vibration data. The novel and significant contributions of this study can be summarized as follows. This study presents and publically shares the
Qatar University Dual-Machine Bearing Fault Benchmark
dataset (QU-DMBF), which encapsulates sound and vibration data from two different motors operating under 1080 working conditions. Then, we focus on the major limitations and drawbacks of vibration-based fault detection due to numerous installation and operational conditions. Finally, we propose the first DL approach for sound-based fault detection and perform comparative evaluations between the sound and vibration signals over the QU-DMBF dataset. A wide range of experimental results shows that the sound-based fault detection method is significantly more robust than its vibration-based counterpart, as it is entirely independent of the sensor location, cost-effective (requiring no sensor and sensor maintenance), and can achieve the same level of the best detection performance by its vibration-based counterpart. This study publicly shares the QU-DMBF dataset, the optimized source codes in PyTorch, and comparative evaluations with the research community.
... With the development of high-precision technology, the requirements for the regulation of various physical parameters of critical engineering parts are being tightened, encompassing aerospace, robotics, engineering vehicles, and numerous other application areas [1][2][3][4]. Among these, bearings are crucial to the regular operation and endurance of mechanical equipment since they are a crucial component of machinery. ...
Carbon quantum dots (CQDs) have already demonstrated their utility as lubricant additives, and non-contact temperature sensing based on CQDs offers considerable potential for condition monitoring in mechanical, electrical, and other fields, as well as lubrication-temperature multifunctional applications in lubricants. In this paper, we have successfully synthesized and designed high-brightness carbon quantum dots/polyvinyl alcohol (PVA) temperature sensor thin film and dispersions of CQDs in a liquid paraffin lubrication system. Based on fluorescence intensity and the fluorescence intensity ratio, the carbon quantum dot/PVA film exhibited exponential temperature-dependent properties with a wide applicability range, a high goodness of fit (R2 > 0.99), and high relative thermal sensitivity (relative sensitivities of 1.74% K−1 and 1.39% K−1 for fluorescence intensity and fluorescence intensity ratio, respectively). In addition, based on the fluorescence intensity, the CQDs exhibited a wide temperature range (20–90 °C), a high goodness of fit (R2 > 0.99), and higher sensitivity (2.84% K−1) in a liquid paraffin lubrication system, which reflects the temperature responsive properties of carbon quantum dots as additives in lubrication systems. These findings provide convenient and effective possibilities for the sensing and monitoring of carbon quantum dots and their multifunctional applications under lubrication systems.
... [52][53][54] Due to their ability to directly extract features from 1D signals and their advantages in terms of computational efficiency and accuracy, 1D-CNNs have garnered significant attention from the research community. Abdeljaber et al. 55 proposed a compact 1D-CNN method for online monitoring and assessing the severity of bearing damage, which exhibits low hardware requirements, high detection accuracy, and strong noise resistance. Sony et al. 56 augmented vibration time-series data using windows and introduced a hyperparameter-optimized 1D-CNN for classifying the augmented data, successfully achieving multi-class damage identification in a full-scale bridge. ...
Pipeline networks are crucial components of modern infrastructure, and ensuring their reliable operation is essential for sustainable development. The percussion-based methods are considered promising for detecting pipeline faults due to their avoidance of constant-contact sensors and ease of implementation. However, the majority of existing percussion-based methods suffer from limitations such as the requirement for manual feature extraction, as well as subpar noise resilience and adaptability. This paper introduces a one-dimensional convolutional bidirectional long short-term memory network with wide first-layer kernels for the classification of percussion-induced acoustic signals, thus achieving automatic identification of pipeline leakage and water deposit conditions. This approach directly extracts features from audio signals using wide first-layer convolutional kernels, eliminating the need for manual feature extraction. Additionally, it employs bidirectional long short-term memory to effectively capture long-term signal dependencies from both past and future contexts. To validate the effectiveness of the method, two case studies were conducted on three groups of pipes. The results show that the proposed method demonstrates superior noise resistance and adaptability compared to other methods, and it also exhibits strong applicability to other percussion signal datasets. Additionally, the impact of different first convolutional kernel sizes on the noise resistance and adaptive performance of the model was investigated, which provides robust guidance for the effective processing of percussion-induced acoustic signals.
... Fortunately, abnormal machine vibrations are largely a result of faulty bearings, and analyzing their characteristics facilitates diagnosing the bearing's health condition [4], [5], [6], [7]. Deep learning techniques, such as Convolutional Neural Networks (CNNs), are adequate to diagnose the bearing's condition by detecting irregularities in the vibration data [8], [9], [10], [11]. Nevertheless, prevailing solutions are usually tested under ideal conditions and often confined to fixed [12], [13], [14] or slowly increasing/decreasing rotational speeds [15], [16] that do not represent dynamic scenarios [4], i.e., time-varying changes. ...
Diagnosis of bearing faults is paramount to reducing maintenance costs and operational breakdowns. Bearing faults are primary contributors to machine vibrations, and analyzing their signal morphology offers insights into their health status. Unfortunately, existing approaches are optimized for controlled environments, neglecting realistic conditions such as time-varying rotational speeds and the vibration's non-stationary nature. This paper presents a fusion of time-frequency analysis and deep learning techniques to diagnose bearing faults under time-varying speeds and varying noise levels. First, we formulate the bearing fault-induced vibrations and discuss the link between their non-stationarity and the bearing's inherent and operational parameters. We also elucidate quadratic time-frequency distributions and validate their effectiveness in resolving distinctive dynamic patterns associated with different bearing faults. Based on this, we design a time-frequency convolutional neural network (TF-CNN) to diagnose various faults in rolling-element bearings. Our experimental findings undeniably demonstrate the superior performance of TF-CNN in comparison to recently developed techniques. They also assert its versatility in capturing fault-relevant non-stationary features that couple with speed changes and show its exceptional resilience to noise, consistently surpassing competing methods across various signal-to-noise ratios and performance metrics. Altogether, the TF-CNN achieves substantial accuracy improvements up to 15%, in severe noise conditions.
... With the development of artificial intelligence, the deep learning can automatically obtain the data features and does not rely on the expertise that is gradually accepted in the field of fault diagnosis [7], [8]. The convolutional neural network (CNN) is a classical method of deep learning, which has attracted much attention in recent years due to its powerful feature extraction capability. ...
The motor is the power source of a mechanical system, which often operates under the variable conditions make it prone to the failure, and the fault diagnosis is difficult. The classical deep learning network needs to design a deep and wide structure to enhance the feature extraction ability, but it will increase the computational cost. In order to realize the intelligent fault diagnosis of a motor under the variable conditions and take into account the efficiency and accuracy of the diagnosis technology, a new method based on the image data and the deep learning is proposed. Based on the ordinary gray image, a feature-enhanced gray texture image is obtained by the local binary pattern. Based on the standard CNN, a new Ada-act LMCNN intelligent fault diagnosis model is developed through the improvement of the adaptive activation function, multi-scale feature extraction and model lightweight. The experimental scheme with different speeds, loads and faults is designed for the rotor-bearing unit of a motor. The fault diagnosis effect of the proposed method is verified in the accelerated, accelerated on-load and mixed case. The results show that the proposed method performs well among the compared methods. The model parameters is 0.2M, the FLOPs is 91.8M, and the model size is 0.9MB. The fault diagnosis accuracy under the mixed case reaches 98.5%. This work provides an intelligent, lightweight, accurate and stable method for the motor fault diagnosis.
... Zhang et al. [30] used 1DCNN to extract the deep features of original signals by stacking convolution kernels. Abdeljaber et al. [31] integrated the feature extraction and classification of a traditional damage monitoring system into a single learner to maximize detection performance while optimizing features. Yang et al. [32] used 1DCNN for feature extraction in the detection model of oil and gas pipeline leakage. ...
Carbon monoxide (CO) is a toxic gas emitted during municipal solid waste incineration (MSWI). Its emission prediction is conducive to pollutant reduction and optimized control of MSWI. The variables of MSWI exhibit redundant and interdependent correlations with CO emissions. Furthermore, the mapping relationship is difficult to characterize. Therefore, the work proposed a CO emission prediction method based on reduced depth features and long short-term memory (LSTM) optimization. The particle design for reduced depth feature and LSTM optimization was initially developed—incorporating an adaptive threshold range for feature selection based on the inherent characteristics of modeling data. Secondly, the nonlinear depth features were extracted using ultra-one-dimensional convolution and subsequently fed into an LSTM model for prediction construction. The hyperparameters of the convolutional layer and LSTM were updated based on the loss function. The generalization performance of the model was used as the fitness function of the optimization. Finally, the particle swarm optimization (PSO) was used to adaptively reduce depth features and model’s hyperparameters. The rationality and effectiveness of the proposed method were validated using the benchmark dataset and CO dataset of MSWI. R² of the testing datasets for RB and CO were 0.9097 ± 3.64E-04 and 0.7636 ± 3.19E-03, respectively, by repeating 30 times.
... The integration of advanced machine learning technologies with vibration analysis has ushered in a new era for bearing Outer race Ball Inner race Cage fault diagnosis [1]. Specifically, leveraging Convolutional Neural Networks (CNN) was found to be a promising approach for detecting and identifying bearing defects at their early stages; thereby, enabling proactive maintenance strategies and ultimately reducing the economic and operational impact of unexpected failures [8], [9], [10]. Unfortunately, current solutions are usually designed to work in a very restricted setting and do not integrate the varying conditions and operations found in a real-life scenario, e.g., heavy noise and time-varying rotational speeds. ...
Detection of rolling-element bearing faults is crucial for implementing proactive maintenance strategies and for minimizing the economic and operational consequences of unexpected failures. However, many existing techniques are developed and tested under strictly controlled conditions, limiting their adaptability to the diverse and dynamic settings encountered in practical applications. This paper presents an efficient real-time convolutional neural network (CNN) for diagnosing multiple bearing faults under various noise levels and time-varying rotational speeds. Additionally, we propose a novel Fisher-based spectral separability analysis (SSA) method to elucidate the effectiveness of the designed CNN model. We conducted experiments on both healthy bearings and bearings afflicted with inner race, outer race, and roller ball faults. The experimental results show the superiority of our model over the current state-of-the-art approach in three folds: it achieves substantial accuracy gains of up to 15.8%, it is robust to noise with high performance across various signal-to-noise ratios, and it runs in real-time with processing durations five times less than acquisition. Additionally, by using the proposed SSA technique, we offer insights into the model's performance and underscore its effectiveness in tackling real-world challenges.
... 1D CNN applications have employed concise architectures, typically with one or two hidden CNN layers (S.Kiranyaz, Ince, Hamila, et al. 2015; S. Kiranyaz, Ince, and Gabbouj 2016; S. Kiranyaz, Ince, and Gabbouj 2017;Avci et al. 2018; Avcı, Abdeljaber, S. Kiranyaz, et al. 2017; Abdeljaber, Avcı, S. Kiranyaz, et al. 2017; Avcı, Abdeljaber, M. S. Kiranyaz, et al. 2018; Abdeljaber, Avcı, M. S. Kiranyaz, et al. 2018;Ince et al. 2016; S. Kiranyaz, Gastli, et al. 2019;Abdeljaber, Sassi, et al. 2019;Eren et al. 2019;Eren 2017). Conversely, nearly all 2D CNN applications have utilized more complex and extensive architectures. ...
As technology continues to advance and become more integrated in the oil and gas industry, a vast amount of data is now prevalent across various scientific disciplines, providing new opportunities to gain insightful and actionable information. The convergence of digital transformation with the physics of fluid flow through porous media and pipelines has driven the advancement and application of machine learning (ML) techniques to extract further value from this data. As a result, digital transformation and its associated machine-learning applications have become a new area of scientific investigation. The transformation of brownfields into digital oilfields can aid in energy production by accomplishing various objectives, including increased operational efficiency, production optimization, collaboration, data integration, decision support, and workflow automation. This work aims to present a framework of these applications, specifically through the implementation of virtual sensing, predictive analytics using predictive maintenance on production hydraulic systems (with a focus on electrical submersible pumps), and prescriptive analytics for production optimization in steam and waterflooding projects. In terms of virtual sensing, the accurate estimation of multi-phase flow rates is crucial for monitoring and improving production processes. This study presents a data-driven approach for calculating multi-phase flow rates using sensor measurements located in electrical submersible pumped wells. An exhaustive exploratory data analysis is conducted, including a univariate study of the target outputs (liquid rate and water cut), a multivariate study of the relationships between inputs and outputs, and data grouping based on principal component projections and clustering algorithms. Feature prioritization experiments are performed to identify the most influential parameters in the prediction of flow rates. Model comparison is done using the mean absolute error, mean squared error and coefficient of determination. The results indicate that the CNN-LSTM network architecture is particularly effective in time series analysis for ESP sensor data, as the 1D-CNN layers are capable of extracting features and generating informative representations of time series data automatically. Subsequently, the study presented herein a methodology for implementing predictive maintenance on artificial lift systems, specifically regarding the maintenance of Electrical Submersible Pumps (ESPs). Conventional maintenance practices for ESPs require extensive resources and manpower and are often initiated through reactive monitoring of multivariate sensor data. To address this issue, the study employs the use of principal component analysis (PCA) and extreme gradient boosting trees (XGBoost) to analyze real-time sensor data and predict potential failures in ESPs. PCA is utilized as an unsupervised technique and its output is further processed by the XGBoost model for prediction of system status. The resulting predictive model has been shown to provide signals of potential failures up to seven days in advance, with an F1 score greater than 0.71 on the test set. In addition to the data-driven modeling approach, The present study also in- corporates model-free reinforcement learning (RL) algorithms to aid in decision-making in production optimization. The task of determining the optimal injection strategy poses challenges due to the complexity of the underlying dynamics, including nonlinear formulation, temporal variations, and reservoir heterogeneity. To tackle these challenges, the problem was reformulated as a Markov decision process and RL algorithms were employed to determine actions that maximize production yield. The results of the study demonstrate that the RL agent was able to significantly enhance the net present value (NPV) by continuously interacting with the environment and iteratively refining the dynamic process through multiple episodes. This showcases the potential for RL algorithms to provide effective and efficient solutions for complex optimization problems in the production domain. In conclusion, this study represents an original contribution to the field of data-driven applications in subsurface energy systems. It proposes a data-driven method for determining multi-phase flow rates in electrical submersible pumped (ESP) wells utilizing sensor measurements. The methodology includes conducting exploratory data analysis, conducting experiments to prioritize features, and evaluating models based on mean absolute error, mean squared error, and coefficient of determination. The findings indicate that a convolutional neural network-long short-term memory (CNN-LSTM) network is an effective approach for time series analysis in ESPs. In addition, the study implements principal component analysis (PCA) and extreme gradient boosting trees (XGBoost) to perform predictive maintenance on ESPs and anticipate potential failures up to a seven-day horizon. Furthermore, the study applies model-free reinforcement learning (RL) algorithms to aid decision-making in production optimization and enhance net present value (NPV).
... Recently, 1D CNNs has attracted a lot of attention due to their advantages over 2-D Convolutional Neural Networks (2D CNN) such as less complicated computation under analogous network, configuration, and hyperparameters; shallower architectures that are easier to train and execute; low hardware setup and low cost, which is suitable for a variety of applications. [22][23][24][25] Like the conventional 2D CNNs, the input layer of the 1D CNN is a passive layer that receives the 1D signal. Because each set consists of three time series, the size of the input layer is 3. Overall, the 1D CNN model architecture employed in this study is illustrated in Figure 6. ...
In this study, the authors proposed a method to fabricate a resistive stretch textile sensor from polyester spandex (PET/SP) fabric and commercial single-walled carbon nanotube (SWCNT). In addition, we designed and trained a one-dimension convolutional neural network to classify four resistance workouts, which employed data acquired from the proposed sensor as the input. To figure out the most appropriate PET/SP sample for the deep learning application, we investigated morphologies and characterization of three samples in distinct conditions of the coating process. Data acquired from the proposed sensor illustrated the significant difference between activated and non-activated muscle groups in each specific exercise. With the PET/SP sample which met the requirements of the application, after 100 epochs, the deep learning model achieved 97.2% training accuracy and 90% test accuracy. This study demonstrates that the SWCNT-coated PET/SP stretch textile sensor can be utilized effectively to track the activity of forearm muscles during resistance training. Other than that, the proposed 1D-CNN, with the advantage of training time and computational cost, is able to classify time series data with high performance and thus can be applied widely in various deep learning applications, especially in the healthcare and sports industries.
... For the dynamic process modeling, dynamic expansion methods, time-series-based correlation analyses, dynamic inner approaches, and the state-space model methods are gradually proposed [11], [12], [13]. Generally speaking, at present, various improved methods and traditional basic models are not independent of each other, and more and more methods combine them to make comprehensive use of the advantages of different methods, so that better online monitoring and fault diagnosis are achieved [15], [16]. However, the shallow network structure adopted by these methods cannot mine complex deep features well, and the expression ability of complex manufacturing process data is insufficient. ...
Large-scale manufacturing processes are usually made up of multiple interrelated and distributed continuously subprocesses, which are transmitted and connected by information and quality flow. The characteristics of long processes, quality heritability between subprocesses, and dynamic-coupled variables bring severe challenges to conventional quality-related fault diagnosis. Against this background, a novel distributed diagnosis framework for quality-related faults is proposed in this article. First, the sequential manufacturing process is decomposed into multiple subprocesses based on mechanism knowledge. Second, a novel dual-attention quality-driven autoencoder method is designed as the model for local fault diagnosis. Deep nonlinear features are extracted under quality supervision; meanwhile, the dynamic information and the different correlations among variables are also considered. Then, based on the tandem structure of the manufacturing process, multiple dual-attention quality-driven autoencoder models corresponding to each subprocess are constructed and stacked into a distributed model. Bayesian inference is used to build global monitoring statistics. Moreover, after faults occur, the intervariable attention weights are achieved to identify faulty variables. Finally, the effectiveness and advantages of the proposed framework are demonstrated via a practical large-scale sequential manufacturing process, the hot strip mill process.
... Abnormal vibration is an indicator of system failure; hence, many fault detection and prognosis studies are conducted based on vibration-indicated data collected through suitable sensors, mainly accelerometers [20][21][22]. The advent of deep learning models, especially 1D CNN, is found to be accurate in extracting the unique features in order to classify and assess the severity of anomalies, and they are used for real-time applications due to their simple structure, low computational complexity, and easy deployability [23]. ...
An automated Condition Monitoring (CM) and real-time controlling framework is essential for outdoor mobile robots to ensure the robot’s health and operational safety. This work presents a novel Artificial Intelligence (AI)-enabled CM and vibrotactile haptic-feedback-based real-time control framework suitable for deploying mobile robots in dynamic outdoor environments. It encompasses two sections: developing a 1D Convolutional Neural Network (1D CNN) model for predicting system degradation and terrain flaws threshold classes and a vibrotactile haptic feedback system design enabling a remote operator to control the robot as per predicted class feedback in real-time. As vibration is an indicator of failure, we identified and separated system- and terrain-induced vibration threshold levels suitable for CM of outdoor robots into nine classes, namely Safe, moderately safe system-generated, and moderately safe terrain-induced affected by left, right, and both wheels, as well as severe classes such as unsafe system-generated and unsafe terrain-induced affected by left, right, and both wheels. The vibration-indicated data for each class are modelled based on two sensor data: an Inertial Measurement Unit (IMU) sensor for the change in linear and angular motion and a current sensor for the change in current consumption at each wheel motor. A wearable novel vibrotactile haptic feedback device architecture is presented with left and right vibration modules configured with unique haptic feedback patterns corresponding to each abnormal vibration threshold class. The proposed haptic-feedback-based CM framework and real-time remote controlling are validated with three field case studies using an in-house-developed outdoor robot, resulting in a threshold class prediction accuracy of 91.1% and an effectiveness that, by minimising the traversal through undesired terrain features, is four times better than the usual practice.
... The advantage of the 1D CNN model has also been verified in previous works [18,19], showing a better accuracy and inference time compared to other common approaches such as long short-term memory (LSTM), CNN-LSTM, multilayer perceptron (MLP), and support vector machine (SVM). The type of sensor generally used for collecting vibration data in various systems are micro-electromechanical systems (MEMS) or piezoelectric accelerometers [20,[29][30][31]. The studies described above show that vibrationbased CM research has been scarcely conducted for wheeled mobile autonomous robots, especially for cleaning applications. ...
Autonomous mobile cleaning robots are ubiquitous today and have a vast market need. Current studies are mainly focused on autonomous cleaning performances, and there exists a research gap on monitoring the robot’s health and safety. Vibration is a key indicator of system deterioration or external factors causing accelerated degradation or threats. Hence, this work proposes an artificial intelligence (AI)-enabled automated condition monitoring (CM) framework using two heterogeneous sensor datasets to predict the sources of anomalous vibration in mobile robots with high accuracy. This allows triggering proper maintenance or corrective actions based on the condition of the robot’s health or workspace, easing condition-based maintenance (CbM). Anomalous vibration sources are classified as induced by uneven Terrain, Collision with obstacles, loose Assembly, and unbalanced Structure, which causes accelerated system deterioration or potential hazards. Here, an unexplored heterogeneous sensor dataset using inertial measurement unit (IMU) and current sensors is proposed for effective recognition across different vibration classes, resulting in higher-accuracy prediction. A simple-structured 1D convolutional neural network (1D CNN) is developed for training and real-time prediction. A 2D CbM map is generated by fusing the predicted classes in real time on an occupancy grid map of the workspace to monitor the conditions of the robot and workspace remotely. The evaluation test results of the proposed method show that the usage of heterogeneous sensors performs significantly more accurately (98.4%) than previous studies, which used IMU (92.2%) and camera (93.8%) sensors individually. Also, this model is comparatively fast, fit for the environment, and ideal for real-time applications in mobile robots based on field trial validations, enhancing mobile robots’ productivity and operational safety.
... Such signals may be decomposed into sub-bands in many scales for 1D CNN to learn to "extract" specific features that can be used for classification tasks, such as bearings fault detection [39], paritne-specific electrocardiogram (ECG) classification [40], and other damage detection tasks for structural health monitoring for civil, mechanical and aerospace engineering [41,42]. In 1D CNN, feature extraction and classification are fused into one process, leading to reduction in computational complexities [43]. 1D CNN also has lower hardware requirement and, hence, lower costs compared to 2D CNN, for which training normally requires special hardware setup [44]. ...
Subsea power cables are critical assets for electrical transmission and distribution networks, and highly relevant to regional, national, and international energy security and decarbonization given the growth in offshore renewable energy generation. Existing condition monitoring techniques are restricted to highly constrained online monitoring systems that only prioritize internal failure modes, representing only 30% of cable failure mechanisms, and has limited capacity to provide precursor indicators of such failures or damages. To overcome these limitations, we propose an innovative fusion prognostics approach that can provide the in situ integrity analysis of the subsea cable. In this paper, we developed low-frequency wide-band sonar (LFWBS) technology to collect acoustic response data from different subsea power cable sample types, with different inner structure configurations, and collate signatures from induced physical failure modes as to obtain integrity data at various cable degradation levels. We demonstrate how a machine learning approach, e.g., SVM, KNN, BP, and CNN algorithms, can be used for integrity analysis under a hybrid, holistic condition monitoring framework. The results of data analysis demonstrate the ability to distinguish subsea cables by differences of 5 mm in diameter and cable types, as well as achieving an overall 95%+ accuracy rate to detect different cable degradation stages. We also present a tailored, hybrid prognostic and health management solution for subsea cables, for cable remaining useful life (RUL) prediction. Our findings addresses a clear capability and knowledge gap in evaluating and forecasting subsea cable RUL. Thus, supporting a more advanced asset management and planning capability for critical subsea power cables.
... The vibration data for fault detection and prognosis of equipment is typically measured using micro-electro-mechanical systems (MEMS) or piezoelectric accelerometers [20][21][22][23], mainly for bearings, motors, and machinery. Accelerometers are also used for terrain classification in outdoor mobile robot applications, for instance, classifying indoor floor, paving, asphalt, gravel, boule court, and grass in [24], as well as in [25] for a four-wheeled planetary exploration rover to classify sand, gravel, and clay. ...
An automated condition monitoring (CM) framework is essential for outdoor mobile robots to trigger prompt maintenance and corrective actions based on the level of system deterioration and outdoor uneven terrain feature states. Vibration indicates system failures and terrain abnormalities in mobile robots; hence, five vibration threshold classes for CM in outdoor mobile robots were identified, considering both vibration source system deterioration and uneven terrain. This study proposes a novel CM approach for outdoor mobile robots using a 3D LiDAR, employed here instead of its usual use as a navigation sensor, by developing an algorithm to extract the vibration-indicated data based on the point cloud, assuring low computational costs without losing vibration characteristics. The algorithm computes cuboids for two prominent clusters in every point cloud frame and sets motion points at the corners and centroid of the cuboid. The three-dimensional vector displacement of these points over consecutive point cloud frames, which corresponds to the vibration-affected clusters, are compiled as vibration indication data for each threshold class. A simply structured 1D Convolutional Neural Network (1D CNN)-based vibration threshold prediction model is proposed for fast, accurate, and real-time application. Finally, a threshold class mapping framework is developed which fuses the predicted threshold classes on the 3D occupancy map of the workspace, generating a 3D CbM map in real time, fostering a Condition-based Maintenance (CbM) strategy. The offline evaluation test results show an average accuracy of vibration threshold classes of 89.6% and consistent accuracy during real-time field case studies of 89%. The test outcomes validate that the proposed 3D-LiDAR-based CM framework is suitable for outdoor mobile robots, assuring the robot’s health and operational safety.
... In contrast, for 2D CNN, the use of GPU is mandatory. Recent studies proved that with limited labeled data and high-variation 1D data, the 1D CNN showed superior performance [93][94][95][96][97][98][99][100][101][102][103][104][105]. ...
Heart disease is a significant public health problem, and early detection is crucial for effective treatment and management. Conventional and noninvasive techniques are cumbersome, time-consuming, inconvenient, expensive, and unsuitable for frequent measurement or diagnosis. With the advance of artificial intelligence (AI), new invasive techniques emerging in research are detecting heart conditions using machine learning (ML) and deep learning (DL). Machine learning models have been used with the publicly available dataset from the internet about heart health; in contrast, deep learning techniques have recently been applied to analyze electrocardiograms (ECG) or similar vital data to detect heart diseases. Significant limitations of these datasets are their small size regarding the number of patients and features and the fact that many are imbalanced datasets. Furthermore, the trained models must be more reliable and accurate in medical settings. This study proposes a hybrid one-dimensional convolutional neural network (1D CNN), which uses a large dataset accumulated from online survey data and selected features using feature selection algorithms. The 1D CNN proved to show better accuracy compared to contemporary machine learning algorithms and artificial neural networks. The non-coronary heart disease (no-CHD) and CHD validation data showed an accuracy of 80.1% and 76.9%, respectively. The model was compared with an artificial neural network, random forest, AdaBoost, and a support vector machine. Overall, 1D CNN proved to show better performance in terms of accuracy, false negative rates, and false positive rates. Similar strategies were applied for four more heart conditions, and the analysis proved that using the hybrid 1D CNN produced better accuracy.
... In particular, 1D CNNs have received considerable attention because they are useful for classification and have a shallow architecture that is easier to train and run. With lower hardware setups and costs, they are suitable for various applications such as recognizing interesting patterns [42][43][44][45]. In this study, we used a 1D CNN to divide the four categories of standing, walking, power walking, and running from time-series data [46,47]. ...
Research on healthcare and body monitoring has increased in recent years, with respiratory data being one of the most important factors. Respiratory measurements can help prevent diseases and recognize movements. Therefore, in this study, we measured respiratory data using a capacitance-based sensor garment with conductive electrodes. To determine the most stable measurement frequency, we conducted experiments using a porous Eco-flex and selected 45 kHz as the most stable frequency. Next, we trained a 1D convolutional neural network (CNN) model, which is a type of deep learning model, to classify the respiratory data according to four movements (standing, walking, fast walking, and running) using one input. The final test accuracy for classification was >95%. Therefore, the sensor garment developed in this study can measure respiratory data for four movements and classify them using deep learning, making it a versatile wearable in the form of a textile. We expect that this method will advance in various healthcare fields.
... On the other hand, different artificial neural network methods have been established based on convolutional neural networks (CNN) to identify bearing faults and improve the accuracy of fault diagnosis, such as MACCNN [53], ADCNN [54], MT-1DCNN [55], 1-D CNN [56], CNN-GRU [57], etc., or an intelligent fault diagnosis model (GL-mRMR-SVM) based on support vector machine (SVM) and feature fusion and feature selection [58], a support tensor machine (STM) [59], etc., to establish a new fault identification method to improve the accuracy of mainshaft bearing fault diagnosis. Wang Rui [60] proposed a deep convolutional neural network that combines residual blocks and channel attention mechanisms for bearing fault diagnosis. ...
Aeroengine mainshaft bearings are key components in modern aeroengines, and their main functions are to support the rotation of the main shaft of the aeroengine in harsh environments, such as high temperature, heavy load, high speed and oil break; reduce the friction coefficient during the high-speed rotation of the main shaft; and reliably ensure the rotation accuracy and power transmission of the aeroengine’s main shaft during operation. The manufacture of aeroengine mainshaft bearings requires complex processes and precise machining to ensure high performance and reliability, and how to intelligently complete the production and manufacture of mainshaft bearings and ensure the strength and accuracy of the bearings, quickly distinguish the fault types of the bearings and efficiently calculate, analyze and predict the life of the bearings are the current research hotspots. Therefore, building a high-fidelity and computationally efficient digital twin life cycle of aeroengine mainshaft bearings is a valuable solution. This paper summarizes the key manufacturing technology, manufacturing mode and manufacturing process based on digital twins in the life cycle of aeroengine mainshaft bearings, including the metallurgical process, heat treatment process and grinding process of aeroengine mainshaft bearings. It presents a fault diagnosis and life analysis of mainshaft bearings of aeroengines, discussing the key technologies and research directions of the life cycle of mainshaft bearings based on digital twins.
... Zhang et al [25] proposed a CNN model with a first-layer wide kernel, which has robust fault diagnosis performance. In [26], an online condition monitoring method based on compact onedimensional CNN was used to locate and quantify the damage of bearings, which can identify single and multiple damages with good real-time processing capability and noise immunity. Chen et al [27] proposed a fault diagnosis model that combines CNN and long short-term memory (LSTM) network, where two CNNs as a feature extractor to extract fault features in the original vibration signal and LSTM network used for fault identification. ...
The domain adaptation methods have good performance in solving the distribution discrepancy of vibration signals of rolling bearings under variable conditions, but without considering the alignment of different categories. To this end, a new dual adversarial domain adaptation (2ADA) mechanism for feature intra-category is proposed and a fault diagnosis model based on 2ADA is built in this paper. The method effectively uses category information to achieve category awareness, and avoids misclassification at the fuzzy decision boundary. In the training process, the multiple-kernel maximum mean discrepancy is used to reduce the discrepancy and perform a global alignment. The category-level alignment is performed when 2ADA is activated, which due to obtain more comprehensive domain adaptation performance and improve the accuracy of fault classification. The results of fault diagnosis experiments on the CWRU bearing dataset and the rotating machinery fault platform dataset demonstrate that, the diagnosis accuracy of the proposed method is improved by up to 15.46% and 5.75% on tasks with high domain shift when compared with CNN method, which verifies the effectiveness of the method.
... The traditional Convolution Neural Network (CNN) has been widely used in image classification tasks (Krizhevsky et al., 2012), but it has disadvantages in dealing with onedimensional data (Kiranyaz et al., 2021). Then, One Dimension-Convolution Neural Network (1D-CNN) was chosen to detect abnormal ECG (Abdeljaber et al., 2018a; and to judge motor and bearing faults (Abdeljaber et al., 2018b;Ince et al., 2016;Zhang et al., 2018) for its advantage of effectively learning one-dimensional time series and short-term correlation predicting (Yu and Koltun, 2015). Its good performance in time prediction is even comparable to RNNs (Bai et al., 2018). ...
Highlights
To predict soil temperature, a new deep learning model called 1D-CNN-MLP is proposed, which has higher accuracy or faster convergence compared with MLP or LSTM.
Convolutional neural network part in the model could extract and calculate transmission of soil temperature.
Using the non-sequential data of several soil temperature layers combined with the model, we can predict other temperature layers.
The model can greatly reduce the difficulty and cost of soil temperature measurement.
Abstract. Soil temperature plays an important role in agriculture. In order to achieve cost reduction in the sensor arrangement when monitoring soil temperature, a novel model called 1D-CNN-MLP (One dimensional convolutional neural network-Multilayer perceptron) was proposed for soil temperature prediction. Meteorological data and soil temperature data on different soil layers collected for the 2018~2021 period from a weather station in Yangling, China, were used for calculation in our work. Our model was evaluated using statistical measures of MSE (Mean Square error). The model parameters with high operation efficiency and high accuracy are obtained, and the training result records much lower error than MLP (multilayer perceptron) and faster convergence than LSTM (long short-term memory) with an MSE of 0.288 x 10&-3. The 1D-CNN (One-dimensional convolutional neural network) part of the model is used to reveal and extrapolate the law of how soil temperature propagates in different soil layers. In the case where only three layers of soil temperature data are known, the characteristic temperature layer depths of 10 cm, 15 cm, and 40 cm, are selected to place sensors and obtain the best prediction effect of soil temperature at different depths of 5 to 160 cm with a RMSE (Root mean squared error) of 1.988?. The model may help users with improved and economical soil temperature prediction and control, thus boosting crop yield. Ultimately, we found the model has a relatively poor performance in the accuracy of deep soil temperature prediction when only three layers of soil temperature data are known, and it is suggested that the model can be further optimized in terms of kernel parameter setting, data composition, and the variation law of deep soil temperature. Keywords: 1D-CNN, MLP, Soil temperature prediction.
... Various data formats can be utilized as input data for ML and DL. First, the raw data are directly acquired from proximities and used as input data [9,10,21]. The orbit depicts two-dimensional data, while it does not contain time information. ...
Fault diagnosis is important in rotor systems because severe damage can occur during the operation of systems under harsh conditions. The advancements in machine learning and deep learning have led to enhanced performance of classification. Two important elements of fault diagnosis using machine learning are data preprocessing and model structure. Multi–class classification is used to classify faults into different single types, whereas multi–label classification classifies faults into compound types. It is valuable to focus on the capability of detecting compound faults because multiple faults can exist simultaneously. Diagnosis of untrained compound faults is also a merit. In this study, input data were first preprocessed with short–time Fourier transform. Then, a model was built for classification of the state of the system based on multi–output classification. Finally, the proposed model was evaluated based on its performance and robustness for classification of compound faults. This study proposes an effective model based on multi–output classification, which can be trained using only single fault data for the classification of compound faults and confirms the robustness of the model to changes in unbalance.
... Condition monitoring based on vibration could be used to identify mechanical faults on induction motors, more specifically, bearing faults [11], [12], [13]. There are at least four methods found in the literature to identify these kinds of faults: i) vibration frequency analysis; ii) envelope analysis; iii) shorted time Fourier transformation; and iv) empirical mode decomposition. ...
Advanced Technologies for Realizing Sustainable Development Goals: 5G, AI, Big Data, Blockchain, and Industry 4.0 Applications explores the intersection of cutting-edge technologies and their role in achieving the United Nations Sustainable Development Goals (SDGs). This book covers diverse topics, including energy-efficient cities, smart healthcare systems, blockchain for social empowerment, and sustainable agriculture. It explores the impact of 5G, AI, machine learning, and cybersecurity on smart cities, industry, and healthcare, providing valuable insights for sustainable development. Key Features: - Highlights the role of advanced technologies like 5G, AI, and blockchain in achieving SDGs - Provides case studies on smart cities, healthcare, and agriculture - Examines emerging issues in cybersecurity and sustainability - Offers insights into Industry 4.0 tools and their applications This book is essential for those seeking to understand how emerging technologies can drive global sustainability efforts.
Autonomous damage identification of submerged structure-foundation systems is challenging due to the difficulty of acquiring damage-induced system responses for training deep learning models. In this study, a novel approach integrating pseudo-damage simulation and convolutional neural network (CNN) deep learning is proposed for damage identification in the submerged structure-foundation system. Pseudo-damage simulation is a technique to generate equivalent damage conditions in inaccessible submerged sub-systems for training deep learning models. The following approaches are implemented to achieve the objective. Firstly, a scheme of pseudo-damage simulation for 1-D CNN deep learning is designed for the caisson-foundation system. Secondly, a vibration monitoring method using pseudo-wave-impulse excitations is designed for the caisson-foundation system. Thirdly, 1-D CNN models are trained for individual caisson units to predict the location and size of foundation damage by vibration signals out of a series of pseudo-damage cases. The 1-D CNN models demonstrate accurate performance in handling untrained scenarios. Experimental results validate the effectiveness of the proposed approach in achieving high accuracy for identification of the foundation damage.
Advancements in drone technology make them important in many areas. military, industry, and disaster The efficacy of a drone's communication systems can be greatly impacted by temperature fluctuations, either from environmental conditions or mechanical problems in the drone's construction. This study gives an analysis and computational model of the impact of temperature on the performance of drone communication. Utilizing a one-dimensional convolutional neural network, we aim to forecast the signal-to-noise ratio (SNR) and received signal strength indicator (RSSI). Following the initial stage of dataset creation in the drone laboratory, proceed to reprocess the dataset and divide it into a 70% training set and a 30% testing set. Subsequently, a graphical user interface (GUI) was developed using MATLAB App Designer to enhance user friendliness. The outcome suggests that the efficiency of the drone communication system declines with rising temperatures. Using 1DCNN is our contribution to this work; other studies depend only on simulation to assess performance. One benefit of 1DCNN is that the impact may be evaluated by automatically extracting important features from the input dataset. Using 1DCNN is our special addition to this project; other research evaluate the UAV communication system's effectiveness only through simulation. We propose in this work to optimize system characteristics for improved performance, including power transfer, by adding a feedback loop between the CNN result and the communication system. Furthermore, we investigate how different weather conditions, such wind and rain, affect UAV communication systems.
The statistical relationship between sensor signature features and lubricant solid particle contamination conditions in a spherical roller bearing has been investigated in this study. The influence of particle size and concentration of solid contaminants in lubricant on the RMS parameter of time-domain acoustic emission, vibration, and sound sensor signals are examined. Machine learning algorithms are trained with time domain statistical features derived from sensor signatures to predict the lubricant conditions. Decision trees, bagging tree ensembles, and support vector machines are used to build ML models. Decision Tree models are built using classification and regression tree algorithms with three distinct split criteria, namely gini, towing, and maximum deviance. A bagged tree ensemble model is constructed using the decision tree as a base learner. In the support vector machine, kernel tricking is done to optimize the classification boundaries. Models built using Acoustic emission signature features predict lubricant conditions with better accuracy compared to models constructed using sound and vibration signature features. Feature-level fusion approach is implemented by combining the vibration, sound, and acoustic emission features at the feature level to improve the prediction power of machine learning models. The bagged tree ensemble and support vector machine models, which are trained using fused features, predict lubricant conditions in spherical roller bearings with an accuracy of around 99%.
Induction motors often operate in variable conditions, and the data distribution of diverse conditions is different. To solve the issue of low accuracy of diagnosis methods caused by large differences in data distribution, an intelligent fault diagnosis method of induction motor rotor-bearing system is proposed. The method combines convolutional neural network with adaptive batch normalization, and the gray texture images transformed from vibration signals are used as the input samples. A total of thirteen conditions including constant and time-varying loads are designed to verify the performance of the method. The average accuracy of the model for thirteen conditions of test set reaches 98.7 %, and the average accuracy for eight unknown conditions of test set is 99.0 %. The proposed method can extract features that are not affected by the variable load conditions, facilitate the transfer from no-load to on-load and known to unknown conditions.
The vibration signal is a typical non-stationary signal, making it challenging to use traditional time-frequency analysis techniques for fault diagnosis. Therefore, this work investigates the processing of vibration signals and proposes a deep learning method based on processed signals for the fault diagnosis of ball bearings. In this work, the fault diagnosis is formulated as an image classification problem and solved with deep learning networks. The intrinsic mode functions (IMFs), converted from the vibration signals in the time domain, are then transformed into symmetrized dot pattern (SDP) images. In order to increase classification accuracy, the SDP parameters in this study are chosen by optimizing image similarity. The feasibility and accuracy of the proposed approach are examined experimentally.
Robust and real-time detection of faults has become an ultimate objective for predictive maintenance on rotating machinery. Studies have indicated that around half of all motor failures are attributed to bearing faults alone. Vibration-based Deep Learning (DL) methodologies have become the de facto standard for bearing fault detection as they can produce state-of-the-art detection performances under certain conditions. Despite such particular focus on the vibration signal, the utilization of sound, on the other hand, has been neglected whilst only a few studies have been proposed during the last two decades, all of which were based on a conventional ML approach. One major reason is the lack of a benchmark dataset providing a large volume of both vibration and sound data over several working conditions for different machines and sensor locations. In this study, we address this need by presenting the new benchmark Qatar University Dual-Machine Bearing Fault Benchmark dataset (QU-DMBF), which encapsulates sound and vibration data from two different motors operating under 1080 working conditions overall. Then we draw the focus on the major limitations and drawbacks of vibration-based fault detection due to numerous installation and operational conditions. Finally, we propose the first DL approach for sound-based fault detection and perform comparative evaluations between the sound and vibration signals over the QU-DMBF dataset. A wide range of experimental results shows that the sound-based fault detection method is significantly more robust than its vibration-based counterpart, as it is entirely independent of the sensor location, cost-effective (requiring no sensor and sensor maintenance), and can achieve the same level of the best detection performance by its vibration-based counterpart. With this study, the QU-DMBF dataset, the optimized source codes in PyTorch, and comparative evaluations are now publicly shared.
In this paper, we delve into the evolving landscape of vibration-based structural damage detection (SDD) methodologies, emphasizing the pivotal role civil structures play in society's wellbeing and progress. While the significance of monitoring the resilience, durability, and overall health of these structures remains paramount, the methodology employed is continually evolving. Our focus encompasses not just the transformation brought by the advent of artificial intelligence but also the nuanced challenges and future directions that emerge from this integration. We shed light on the inherent nonlinearities civil engineering structures face, the limitations of current validation metrics, and the conundrums introduced by inverse analysis. Highlighting machine learning's (ML) transformative role, we discuss how techniques such as artificial neural networks and support vector machine's have expanded the SDD's scope. Deep learning's (DL) contributions, especially the innovative capabilities of convolutional neural network in raw data feature extraction, are elaborated upon, juxtaposed with the potential pitfalls, like data overfitting. We propose future avenues for the field, such as blending undamaged real-world data with simulated damage scenarios and a tilt towards unsupervised algorithms. By synthesizing these insights, our review offers an updated perspective on the amalgamation of traditional SDD techniques with ML and DL, underlining their potential in fostering more robust civil infrastructures.
Online fault detection through continuous vibration monitoring is crucial to prevent potentially catastrophic accidents throughout the service life of bearings. However, the high complexity and dynamics of actual vibration environments make it challenging, especially for incipient faults. This paper develops an integrated framework for this issue via spectrum sparsity measure and dynamic alarm thresholds. First, a new spectrum sparsity measure is proposed to establish a spectrum sparsity index (SSI), which is modeled as a noisy sequence to obtain denoised SSI (DSSI) using the generalized total variation algorithm. SSI is achieved by extracting the enhanced envelope spectrum (EES), and further, 1.5-dimensional EES (1.5DEES) is defined for explicit fault identification. Then, a dynamic alarm threshold setting strategy is designed using the generalized extreme value distribution (GEVD) to fit real-time updated SSIs. Finally, an integrated framework is constructed using DSSI as the observation line, SSI-GEVD-based dynamic alarm thresholds for alarming anomalies, and 1.5DEES for immediately identifying fault type. Two life-cycle experimental signals of rolling bearings and one real-world monitoring signal from a high-speed shaft bearing are studied to validate the proposed framework. Results demonstrate that it can effectively achieve online fault detection with superior performance in the lowest false and missed alarms.
Operator attention failure due to mental fatigue during extended equipment operations is a common cause of equipment-related accidents that result in catastrophic injuries and fatalities. As a result, tracking operators' mental fatigue is critical to reducing equipment-related accidents on construction sites. Previously, several strategies aimed at recognizing mental fatigue with adequate accuracy, such as machine learning utilizing EEG-based wearable sensing systems, have been proposed. However, the ability to track operators’ mental fatigue for its implementation on an actual construction site is still an issue. For instance, the mobility and systemic instability of EEG sensors necessitate their application in laboratory settings rather than on actual construction sites. Furthermore, while the machine learning classifiers achieved acceptable accuracy, their input is limited to manually developed EEG features, which may compromise the models’ performance on real construction sites. Accordingly, the current research proposes the viability of a construction site strategy that uses flexible headband-based sensors for acquiring raw EEG data and deep learning networks to recognize operators' mental fatigue. To serve this purpose, a one-hour excavator operation by fifteen operators was conducted on a construction site. The NASA-TLX score was used as the ground truth of mental fatigue, and brain activity patterns were recorded using a wearable EEG sensor. The raw EEG data was then used to develop deep learning-based classification models. Finally, the performance of deep learning models, i.e., long short-term memory, bidirectional LSTM, and one-dimensional convolutional networks, was investigated using accuracy, precision, recall, specificity, and an F1-score. The findings indicate that the Bi-LSTM model outperforms the other deep learning models with a high accuracy of 99.941% and F1-score between 99.917% and 99.993%. These findings demonstrate the feasibility of applying the Bi-LSTM model and contribute to wearable sensor-based mental fatigue recognition and classification, thus enhancing on-site health and safety operations.
Electrocardiogram (ECG) signal, which is composite of multiple segments such as P-wave, QRS complex and T-wave, plays a crucial role in the treatment of cardiovascular disease. For an analysis of cardiac diagnosis, it is required that clinicians scan the ECG signal for QRS complex or R-peaks (the highest peak of the QRS complex) detection, which relies on their expertise and takes enormous time. In order to provide more realistic treatment of cardiovascular diseases, so many computer-based techniques detecting R-peaks/QRS complex in the ECG signal with noises and different characteristics have been actively developed in research article for many years. Moreover, researchers have created various data sets for R-peaks/QRS complex detection. Although R-peaks/QRS complex detection is one of the notable research areas with so many computer-based techniques and ECG data sets, no comprehensive review paper have been published recently. The main aim of this study is to present a wide range of computer-based techniques proposed for detection of R-peaks/QRS complex. First of all, in this study, computer-based techniques proposed in the literature for the detection of R-peaks/QRS complex and their stages are introduced in detail. The generalization ability of these techniques is investigated deeply. Details are given about the most preferred ECG data sets produced in the literature for the analysis of computer-based techniques. Finally, the performances of computer-based techniques and generalization abilities are analyzed for each data set created in the literature by giving the results of the evaluation metrics.
This study presents a novel data-driven approach for calculating multiphase flow rates in electrical submersible pumped wells. Traditional methods for estimating flow rates at test separators fail to identify production trends and require additional costs for maintenance. In response, virtual flow metering (VFM) has emerged as an attractive research area in the oil and gas industry. This study introduces a robust workflow utilizing symbolic regression, extreme gradient boosted trees, and a deep learning model that includes a pipeline of convolutional neural network (CNN) layers and long short-term memory algorithm (LSTM) layers to predict liquid rate and water cut in real time based on pump sensors’ data. The novelty of this approach lies in offering a cost-effective and accurate alternative to the usage of multiphase physical flow meters and production testing. Additionally, the study provides insights into the potential of data-driven methods for VFM in electrical submersible pumped wells, highlighting the effectiveness of the proposed approach. Overall, this study contributes to the field by introducing a new, data-driven method for accurately predicting multiphase flow rates in real time, thereby providing a valuable tool for monitoring and optimizing production processes in the oil and gas industry.
Defects in different positions and degrees in pile foundations will affect the building structure’s safety and the foundation’s bearing capacity. The efficiency and accuracy of using traditional methods to identify multi-defect types of pile foundations are very low, so finding suitable methods to improve their related indicators for pile foundation safety and engineering applications is necessary. In this paper, under the condition of secondary development of finite element software ABAQUS to obtain the time-domain signal database of six kinds of multi-defect pile foundations, a multi-defect type identification method of pile foundations based on two-channel convolutional neural network (TC-CNN) and low-strain pile integrity test (LSPIT) is proposed. Firstly, simulated time-domain signals of the dynamic measurements that match the experimental results performed wavelet packet denoising. Secondly, the 1D time-domain signals before and after denoising and the corresponding 2D wavelet time–frequency maps are inputs to retain more data information and prevent overfitting. Finally, TC-CNN achieved the multi-defect type identification of concrete piles. Compared with the single-channel convolutional neural network, this method can effectively fuse 1D and 2D features, extract more potential features, and make the classification accuracy reach 99.17%.
Early detection of tooth cracks is crucial for effective condition-based monitoring and decision making. The scope of this work was to bring more insight into the vibration behavior of spur gears in the presence of single and multiple simultaneous tooth cracks. The investigation was conducted in both time and frequency domains. A finite element analysis was performed to determine the variation in stiffness with respect to the angular position for different combinations of crack lengths. A simplified nonlinear lumped parameter model of a one-stage gearbox with six degrees of freedom was then developed to simulate the vibration response of faulty external spur gears. Four different multiple-crack scenarios were proposed and studied. The performances of various statistical fault detection indicators were considered and investigated. The simulation results obtained via MATLAB indicated that, as the severity of a single crack increases, the values of the time domain statistical indicators increase also, but at different rates. Moreover, the number of cracks was found to have a negative effect on the values of all the performance indicators, except for the RMS. The number and amplitude of the sidebands in the frequency spectrum were also considered, while assessing the severity of the faults in each scenario. It was observed that, in the case of consecutive tooth cracks, the number of spectrum peaks and the number of cracks were consistent in the frequency range of 4-5 kHz. The main finding of this study was that the peak spectral amplitude was the most sensitive indicator of the number and severity of cracks.
Structural damage detection has been an interdisciplinary area of interest for various engineering fields. While the available damage detection methods have been in the process of adapting machine learning concepts, most machine learning based methods extract "hand-crafted" features which are fixed and manually selected in advance. Their performance varies significantly among various patterns of data depending on the particular structure under analysis. Convolutional neural networks (CNNs), on the other hand, can fuse and simultaneously optimize two major sets of an assessment task (feature extraction and classification) into a single learning block during the training phase. This ability not only provides an improved classification performance but also yields a superior computational efficiency. 1D CNNs have recently achieved state-of-the-art performance in vibration-based structural damage detection; however, it has been reported that the training of the CNNs requires significant amount of measurements especially in large structures. In order to overcome this limitation, this paper presents an enhanced CNN-based approach that requires only two measurement sets regardless of the size of the structure. This approach is verified using the experimental data of the Phase II benchmark problem of structural health monitoring which had been introduced by IASC-ASCE Structural Health Monitoring Task Group. As a result, it is shown that the enhanced CNN-based approach successfully estimated the actual amount of damage for the nine damage scenarios of the benchmark study.
Each year more than 7 million people die from cardiac arrhythmias. Yet no robust solution exists today to detect such heart anomalies right at the moment they occur. The purpose of this study was to design a personalized health monitoring system that can detect early occurrences of arrhythmias from an individual’s electrocardiogram (ECG) signal. We first modelled the common causes of arrhythmias in the signal domain as a degradation of normal ECG beats to abnormal beats. Using the degradation models, we performed abnormal beat synthesis which created potential abnormal beats from the average normal beat of the individual. Finally, a Convolutional Neural Network (CNN) was trained using real normal and synthesized abnormal beats. As a personalized classifier, the trained CNN can monitor ECG beats in real time for arrhythmia detection. Over 34 patients’ ECG records with a total of 63,341 ECG beats from the MIT-BIH arrhythmia benchmark database, we have shown that the probability of detecting one or more abnormal ECG beats among the first three occurrences is higher than 99.4% with a very low false-alarm rate.
Intelligent machine health monitoring and fault diagnosis are becoming increasingly important for modern manufacturing industries. Current fault diagnosis approaches mostly depend on expert-designed features for building prediction models. In this paper, we proposed IDSCNN, a novel bearing fault diagnosis algorithm based on ensemble deep convolutional neural networks and an improved Dempster–Shafer theory based evidence fusion. The convolutional neural networks take the root mean square (RMS) maps from the FFT (Fast Fourier Transformation) features of the vibration signals from two sensors as inputs. The improved D-S evidence theory is implemented via distance matrix from evidences and modified Gini Index. Extensive evaluations of the IDSCNN on the Case Western Reserve Dataset showed that our IDSCNN algorithm can achieve better fault diagnosis performance than existing machine learning methods by fusing complementary or conflicting evidences from different models and sensors and adapting to different load conditions.
Bearing faults are the biggest single source of motor failures. Artificial Neural Networks (ANNs) and other decision support systems are widely used for early detection of bearing faults. The typical decision support systems require feature extraction and classification as two distinct phases. Extracting fixed features each time may require a significant computational cost preventing their use in real-time applications. Furthermore, the selected features for the classification phase may not represent the most optimal choice. In this paper, the use of 1D Convolutional Neural Networks (CNNs) is proposed for a fast and accurate bearing fault detection system. The feature extraction and classification phases of the bearing fault detection are combined into a single learning body with the implementation of 1D CNN. The raw vibration data (signal) is fed into the proposed system as input eliminating the need for running a separate feature extraction algorithm each time vibration data is analyzed for classification. Implementation of 1D CNNs results in more efficient systems in terms of computational complexity. The classification performance of the proposed system with real bearing data demonstrates that the reduced computational complexity is achieved without a compromise in fault detection accuracy.
Most of the classical structural damage detection systems involve two processes, feature extraction and feature classification. Usually, the feature extraction process requires large computational effort which prevent the application of the classical methods in real-time structural health monitoring applications. Furthermore, in many cases, the hand-crafted features extracted by the classical methods fail to accurately characterize the acquired signal, resulting in poor classification performance. In an attempt to overcome these issues, this paper presents a novel, fast and accurate structural damage detection and localization system utilizing one dimensional convolutional neural networks (CNNs) arguably for the first time in SHM applications. The proposed method is capable of extracting optimal damage-sensitive features automatically from the raw acceleration signals, allowing it to be used for real-time damage detection. This paper presents the preliminary experiments conducted to verify the proposed CNN-based approach.
Periodic vibration signals captured by the accelerometers carry rich information for bearing fault diagnosis. Existing methods mostly rely on hand-crafted time-consuming preprocessing of data to acquire suitable features. In this paper, we use an easy and effective method to transform the 1-D temporal vibration signal into a 2-D image. With the signal image, convolutional Neural Network (CNN) is used to train the raw vibration data. As powerful feature extractor and classifier for image recognition, CNN can learn to acquire features most suitable for the classification task by being trained. With the image format of vibration signals, the neuron in fully-connected layer of CNN can see farther and capture the periodic feature of signals. According to the results of the experiments, when fed in enough training samples, the proposed method outperforms other common methods. The proposed method can also be applied to solve intelligent diagnosis problems of other machine systems.
Structural health monitoring (SHM) and vibration-based structural damage detection have been a continuous interest for civil, mechanical and aerospace engineers over the decades. Early and meticulous damage detection has always been one of the principal objectives of SHM applications. The performance of a classical damage detection system predominantly depends on the choice of the features and the classifier. While the fixed and hand-crafted features may either be a sub-optimal choice for a particular structure or fail to achieve the same level of performance on another structure, they usually require a large computation power which may hinder their usage for real-time structural damage detection. This paper presents a novel, fast and accurate structural damage detection system using 1D Convolutional Neural Networks (CNNs) that has an inherent adaptive design to fuse both feature extraction and classification blocks into a single and compact learning body. The proposed method performs vibration-based damage detection and localization of the damage in real-time. The advantage of this approach is its ability to extract optimal damage-sensitive features automatically from the raw acceleration signals. Large-scale experiments conducted on a grandstand simulator revealed an outstanding performance and verified the computational efficiency of the proposed real-time damage detection method.
http://www.sciencedirect.com/science/article/pii/S0022460X16306204
Early detection of the motor faults is essential and Artificial Neural Networks (ANNs) are widely used for this purpose. The typical systems usually encapsulate two distinct blocks: feature extraction and classification. Such fixed and hand-crafted features may be a sub-optimal choice and require a significant computational cost that will prevent their usage for realtime applications. In this paper, we propose a fast and accurate motor condition monitoring and early fault detection system using 1D Convolutional Neural Networks (CNNs) that has an inherent adaptive design to fuse the feature extraction and classification phases of the motor fault detection into a single learning body. The proposed approach is directly applicable to the raw data (signal) and thus eliminates the need for a separate feature extraction algorithm resulting in more efficient systems in terms of both speed and hardware. Experimental results obtained using real motor data demonstrate the effectiveness of the proposed method for real-time motor condition monitoring.
Rolling element bearings are the most crucial part of any rotating machines. The failures of bearing without warning will result catastrophic consequences in many situations. Therefore condition monitoring of bearing is very important. In this paper, artificial intelligence techniques are used to predict and analyses the bearing faults. Experiments were carried out on rolling bearing having localized defects on the various bearing components for wide range of speed and vibration signals were stored. Condition monitoring systems is divided in two important part one feature extraction and second diagnosis through extracted features. Daubechies wavelet is popular for smoothing of signals so, it is chosen for reducing the background noise from vibration signal. Kurtosis, RMS, Creast factor and Peak difference as suitable time domains features are extracted from decompose time velocity signals. Back propagation multilayer neural network was train and tested by 369 pre-treated normliesed features. Support vector machine is also used for the same data for predicting bearing faults. Finally, it is found that Support vector machine techniques gives better results over ANN.
The aim of this paper is to analyze the sensitivity of fault scalar indicators extracted from time domain signals to bearing damage manifested through an increase in size and in the number of localized defects. Six scalar indicators are considered: Peak, RMS, Crest factor, Kurtosis, Impulse factor, and Shape factor. A new software application, called BEAT (BEAring Toolbox), was developed in order to simulate bearing vibratory response to the excitations produced by localized defects. The predictability of the simulation model has already been confirmed by previous comparisons with the results of experiments performed on a bearing test rig. Simulation results show that these time indicators can be used for the early prediction of a fault during the initial stages of degradation. However, they become less sensitive as the damage increases and becomes very severe. Two new descriptors called TALAF and THIKAT, which combine conventional descriptors, are developed in order to improve diagnosis up to the point where the ultimate signs of catastrophic failure are observed, to diagnose the severity of degradation in four stages, and to help managers schedule their maintenance operations.
We propose a fast and accurate patient-specific electrocardiogram (ECG) classification and monitoring system using an adaptive implementation of 1D Convolutional Neural Networks (CNNs) that can fuse feature extraction and classification into a unified learner. In this way, a dedicated CNN will be trained for each patient by using relatively small common and patient-specific training data and thus it can also be used to classify long ECG records such as Holter registers in a fast and accurate manner. Alternatively, such a solution can conveniently be used for real-time ECG monitoring and early alert system on a lightweight wearable device. The experimental results demonstrate that the proposed system achieves a superior classification performance for the detection of ventricular ectopic beats (VEB) and supraventricular ectopic beats (SVEB).
This paper presents a fast and accurate patient-specific electrocardiogram (ECG) classification and monitoring system.
An adaptive implementation of 1D Convolutional Neural Networks (CNNs) is inherently used to fuse the two major blocks of the ECG classification into a single learning body: feature extraction and classification. Therefore, for each patient an individual and simple CNN will be trained by using relatively small common and patient-specific training data, and thus such a patient-specific feature extraction ability can further improve the classification performance. Since this also negates the necessity to extract hand-crafted manual features, once a dedicated CNN is trained for a particular patient, it can solely be used to classify possibly long ECG data stream in a fast and accurate manner or alternatively, such a solution can conveniently be used for real-time ECG monitoring and early alert system on a light-weight wearable device.
The results over the MIT-BIH arrhythmia benchmark database demonstrate that the proposed solution achieves a superior classification performance than most of the state-of-the-art methods for the detection of ventricular ectopic beats (VEB) and supraventricular ectopic beats (SVEB).
Besides the speed and computational efficiency achieved, once a dedicated CNN is trained for an individual patient, it can solely be used to classify his/her long ECG records such as Holter registers in a fast and accurate manner.
Due to its simple and parameter invariant nature, the proposed system is highly generic and thus applicable to any ECG dataset.
Condition monitoring and fault diagnosis of rolling element bearings (REBs) are at present very important to ensure the steadiness of industrial and domestic machinery. According to the non-stationary and non-linear characteristics of REB vibration signals, feature extraction method is based on empirical mode decomposition (EMD) energy entropy in this paper. A mathematical analysis to select the most significant intrinsic mode functions (IMFs) is presented. Therefore, the chosen features are used to train an artificial neural network (ANN) to classify bearings defects. Experimental results indicated that the proposed method based on run-to-failure vibration signals can reliably categorize bearing defects. Using a proposed health index (HI), REB degradations are perfectly detected with different defect types and severities. Experimental results consist in continuously evaluating the condition of the monitored bearing and thereby detect online the severity of the defect successfully. This paper shows potential application of ANN as effective tool for automatic bearing performance degradation assessment without human intervention.
This paper addresses Rolling Contact Fatigue (RCF) occurs due to the result of cyclic stress developed during operation and mechanism that involve in fretting failure of rolling element bearing. As bearing raceways of non-rotating rolling element bearings exposed to vibration or sliding oscillation false Brinelling occurs. Bearing surface due to false Brinelling tends to damage within a short period, due to cavities created on the bearing raceway. Recommendation towards enhancement of bearing life is also suggested.
This work aims to develop the theoretical fundamentals and numerical details of new software, dedicated to the simulation of the dynamic behavior of rotating ball bearings in the presence of localized surface defects. In this article, the generation of vibration by a point defect in a rolling element bearing is modeled as a function of the rotation of the bearing, of the distribution of the load in the bearing, of the bearing structure elasticity, of the oil film characteristics, and of the transfer path between the bearing and the transducer. The numerical model is developed with the assumption that the dynamic behavior of the bearing can be represented by a coupled three-degree-of-freedom system, after which the governing equations of the simulation model are solved using computer simulation techniques. A new application, called BEAT (BEAring Toolbox), was developed in order to simulate bearings' vibratory response to the excitations produced by localized defects. By adding a noisy response due to the sliding friction occurring between the moving parts to the impulsive response caused by localized defects, the BEAT software is able to provide realistic results, similar to those produced by a sensor during experimental measurements.
This paper introduces a method to detect a fault associated with critical components/subsystems of an engineered system. It is required, in this case, to detect the fault condition as early as possible, with specified degree of confidence and a prescribed false alarm rate. Innovative features of the enabling technologies include a Bayesian estimation algorithm called particle filtering, which employs features or condition indicators derived from sensor data in combination with simple models of the system's degrading state to detect a deviation or discrepancy between a baseline (no-fault) distribution and its current counterpart. The scheme requires a fault progression model describing the degrading state of the system in the operation. A generic model based on fatigue analysis is provided and its parameters adaptation is discussed in detail. The scheme provides the probability of abnormal condition and the presence of a fault is confirmed for a given confidence level. The efficacy of the proposed approach is illustrated with data acquired from bearings typically found on aircraft and monitored via a properly instrumented test rig.
A common practice to gain invariant features in object recognition models is to aggregate multiple low-level features over a small neighborhood. However, the differences between those models makes a comparison of the properties of different aggregation functions hard. Our aim is to gain insight into different functions by directly comparing them on a fixed architecture for several common object recognition tasks. Empirical results show that a maximum pooling operation significantly outperforms subsampling operations. Despite their shift-invariant properties, overlapping pooling windows are no significant improvement over nonoverlapping pooling windows. By applying this knowledge, we achieve state-of-the-art error rates of 4.57% on the NORB normalized-uniform dataset and 5.6% on the NORB jittered-cluttered dataset.
A study is presented to compare the performance of bearing fault detection using three types of artificial neural networks (ANNs), namely, multilayer perceptron (MLP), radial basis function (RBF) network, and probabilistic neural network (PNN). The time domain vibration signals of a rotating machine with normal and defective bearings are processed for feature extraction. The extracted features from original and preprocessed signals are used as inputs to all three ANN classifiers: MLP, RBF, and PNN for two-class (normal or fault) recognition. The characteristic parameters like number of nodes in the hidden layer of MLP and the width of RBF, in case of RBF and PNN along with the selection of input features, are optimized using genetic algorithms (GA). For each trial, the ANNs are trained with a subset of the experimental data for known machine conditions. The ANNs are tested using the remaining set of data. The procedure is illustrated using the experimental vibration data of a rotating machine with and without bearing faults. The results show the relative effectiveness of three classifiers in detection of the bearing condition.
This paper introduces a new bearing fault detection and diagnosis
scheme based on hidden Markov modeling (HMM) of vibration signals. First
features are extracted from amplitude demodulated vibration signals
obtained from both normal and faulty bearings. The features are based on
the reflection coefficients of the polynomial transfer function of the
autoregressive model of the vibration signal. These features are then
used to train HMMs to represent various bearing conditions. The
technique allows for online detection of faults by monitoring the
probabilities of the pre-trained HMM for the normal case. It also allows
for the diagnosis of the fault by the HMM that gives the highest
probability. The new scheme was tested with experimental data collected
from drive end ball bearing of an induction motor (Reliance Electric 2HP
IQPreAlert) driven mechanical system
A multiscale Canny edge detection is equivalent to finding the local maxima of a wavelet transform. The authors study the properties of multiscale edges through the wavelet theory. For pattern recognition, one often needs to discriminate different types of edges. They show that the evolution of wavelet local maxima across scales characterize the local shape of irregular structures. Numerical descriptors of edge types are derived. The completeness of a multiscale edge representation is also studied. The authors describe an algorithm that reconstructs a close approximation of 1-D and 2-D signals from their multiscale edges. For images, the reconstruction errors are below visual sensitivity. As an application, a compact image coding algorithm that selects important edges and compresses the image data by factors over 30 has been implemented
Bearing faults are one of the major causes of motor failures. The bearing defects induce vibration, resulting in the modulation of the stator current. In this paper, the stator current is analyzed via wavelet packet decomposition to detect bearing defects. The proposed method enables the analysis of frequency bands that can accommodate the rotational speed dependence of the bearing defect frequencies. The wavelet packet decomposition also provides a better treatment of nonstationary stator current than currently used Fourier techniques.
Unlike traditional machine learning techniques, convolutional neural networks (CNNs), one of deep learning methods, automate the feature extraction process required for an effective classification. In general, CNN based bearing fault diagnosis analyzes raw signals to classify the localized faults. However, bearings are subjected to non-stationary speeds due to various operating conditions, and thus, CNN cannot determine optimal features of the various conditions while analyzing raw signals, reducing classification accuracy. In this paper, we propose a pre-processing step to improve the performance of the CNN based fault diagnosis by extracting envelope spectrums (ES) on the raw signals. As ES demodulates the signals to provide the information inherent in defect frequency of faults and its variations to non-stationary speeds, CNN can learn to extract distinctive features to diagnose bearing defects effectively. The proposed method is evaluated on acoustic emission based low speed bearing data. The trained CNN model is tested on data with different revolutions per minute (RPM), and it achieves the classification accuracy greater than 94.8%.
In recent years, intelligent fault diagnosis algorithms using machine learning technique have achieved much success. However, due to the fact that in real world industrial applications, the working load is changing all the time and noise from the working environment is inevitable, degradation of the performance of intelligent fault diagnosis methods is very serious. In this paper, a new model based on deep learning is proposed to address the problem. Our contributions of include: First, we proposed an end-to-end method that takes raw temporal signals as inputs and thus doesn’t need any time consuming denoising preprocessing. The model can achieve pretty high accuracy under noisy environment. Second, the model does not rely on any domain adaptation algorithm or require information of the target domain. It can achieve high accuracy when working load is changed. To understand the proposed model, we will visualize the learned features, and try to analyze the reasons behind the high performance of the model.
In this paper, we propose a new bearing fault diagnosis method without the feature extraction, based on Convolutional Neural Network (CNN). The 1-D vibration signal is converted to 2-D data called vibration image. Then, the vibration images are fed into the CNN for bearing fault classification. Experiments are carried out with bearing data from the Case Western Reserve University Bearing Fault Database and its result are compared with the results of other methods to show the effectiveness of the proposed algorithm.
In this paper, a nonlinear quality-related fault detection approach is proposed based on kernel least squares (KLS) model. The major novelty of the proposed method is that it utilizes KLS model to exploit the entire correlation between feature and output matrices. Firstly, it uses a nonlinear projection function to map original pro- cess variables into feature space in which the correlation between feature and output matrices is realized by means of KLS. Then the feature matrix is decomposed into two orthogonal parts by singular value decomposition (SVD) and the statistics for each part are determined appropriately for the purpose of quality-related fault detection. Compared with existing kernel partial least squares (KPLS) based approaches, the proposed new method has the following obvious advantages: (1) it extracts the full correlation in- formation of feature matrix while KPLS-based approaches only use the partial correlation of several selected latent variables, therefore it is more stable than the existing ones; (2) it omits the iterative computation of KPLS model and the determination of the number of latent variables, therefore it is more efficient in engineering implementation; (3) it only uses two statistics to determine the type of fault while most of the KPLS-based approaches need four, therefore it has a more simple diagnosis logic. For simulation verification, a widely used literature example and an industrial benchmark are utilized to evaluate the performance of the proposed method.
This article considers the issue of quality-related process monitoring. A novel kernel direct decomposition (KDD) algorithm is proposed and a KDD-based nonlinear qualityrelated fault detection approach is designed. The proposed KDD algorithm first maps original process variables into feature space to deal with the nonlinearities among these variables. Feature matrix is then directly decomposed into two orthogonal parts according to its full correlation with output matrix without building any regression model. Compared with conventional nonlinear methods, the KDD-based approach has the following advantages: (1) it is simpler in design as it omits the steps of constructing a regression model like kernel partial least squares (KPLS); (2) its performance is more stable because it extracts the full correlation information of feature matrix unlike KPLS-based methods which only use the partial correlation information of several selected latent variables; and (3) it has a simpler diagnosis logic since it only uses two statistics to determine the type of fault while most existing methods need four. Simulations on a literature example and a simulated industrial process are used to demonstrate the advantages of the new method.
A study is presented to compare the performance of bearing fault detection using three types of artificial neural networks (ANNs), namely, multilayer perceptron (MLP), radial basis function (RBF) network, and probabilistic neural network (PNN). The time domain vibration signals of a rotating machine with normal and defective bearings are processed for feature extraction. The extracted features from original and preprocessed signals are used as inputs to all three ANN classifiers: MLP, RBF, and PNN for two-class (normal or fault) recognition. The characteristic parameters like number of nodes in the hidden layer of MLP and the width of RBF, in case of RBF and PNN along with the selection of input features, are optimized using genetic algorithms (GA). For each trial, the ANNs are trained with a subset of the experimental data for known machine conditions. The ANNs are tested using the remaining set of data. The procedure is illustrated using the experimental vibration data of a rotating machine with and without bearing faults. The results show the relative effectiveness of three classifiers in detection of the bearing condition.
Traditional artificial methods and intelligence-based methods of classifying and diagnosing various mechanical faults with high accuracy by extracting effective features from vibration data, such as support vector machines and back propagation neural networks, have been widely investigated. However, the problems of extracting features automatically without significantly increasing the demand for machinery expertise and maximizing accuracy without overcomplicating machine structure have to date remained unsolved. Therefore, a novel hierarchical learning rate adaptive deep convolution neural network based on an improved algorithm was proposed in this study, and its use to diagnose bearing faults and determine their severity was investigated. To test the effectiveness of the proposed method, an experiment was conducted with bearing-fault data samples obtained from a test rig. The method achieved a satisfactory performance in terms of both fault-pattern recognition and fault-size evaluation. In addition, comparison revealed that the improved algorithm is well suited to the fault-diagnosis model, and that the proposed method is superior to other existing methods.
Extracting damage-sensitive features plays an important role in all structural health monitoring applications, as it determines the metrics on which to base decision-making with regard to operation, maintenance, damage state, and so on. This article adopts the widely employed frequency response function, both its magnitude and phase, as the selected feature source and demonstrates how the damage types and locations are able to be classified by means of Bayesian recursive confidence updating. The features are estimated from the in situ acquired vibration data on a rotating machinery test-bed, and the probabilistic models that quantify feature uncertainty are the likelihood functions in a Bayesian framework, which informs the most plausible decisions based on the collected evidence. The damage classification effort in this article specifically calculates the posterior probability, considering the prior and likelihood of data observations; posterior probabilities are then fed back as prior probabilities in the next iteration as new test data are observed. There are three ball-bearing damage conditions applied to the rotary machine test-bed, and the correct model representing the correct damage types will be selected by the model with the maximum posterior confidence. Classification via posterior probability is shown in this article to outperform traditional likelihood evaluations, and the Bayesian recursive implementation distinguishes all three conditions in this work.
Bearing degradation is the most common source of faults in electrical machines. In this context, this work presents a novel monitoring scheme applied to diagnose bearing faults. Apart from detecting local defects, i.e., single-point ball and raceway faults, it takes also into account the detection of distributed defects, such as roughness. The development of diagnosis methodologies considering both kinds of bearing faults is, nowadays, subject of concern in fault diagnosis of electrical machines. First, the method analyzes the most significant statistical-time features calculated from vibration signal. Then, it uses a variant of the curvilinear component analysis, a nonlinear manifold learning technique, for compression and visualization of the feature behavior. It allows interpreting the underlying physical phenomenon. This technique has demonstrated to be a very powerful and promising tool in the diagnosis area. Finally, a hierarchical neural network structure is used to perform the classification stage. The effectiveness of this condition-monitoring scheme has been verified by experimental results obtained from different operating conditions.
This paper demonstrates a novel and cost-effective approach for diagnosis and prognosis of bearing faults in small and medium size induction motors. Even though, many researchers dealt with the bearing fault diagnosis of induction motors by using traditional and soft computing approaches, the application of these techniques for predicting the remaining life time of electrical equipment is not seen much in the literature. Moreover, individual artificial intelligence (AI) techniques suffer from their own drawbacks, which can overcome by forming a hybrid approach combining the advantages of each technique. Hence, in this paper an attempt has been made to combine neural networks and fuzzy logic and forming a fuzzy back propagation (fuzzy BP) network for identifying the present condition of the bearing and estimate the remaining useful time of the motor. The results obtained from fuzzy BP network are compared with the neural network, which show that the hybrid approach is well suitable for assessing the present condition of the bearing and the time available for the replacement of the bearing.
The present work was performed in the frame of cooperation between Baumuller Nurnberg and the University of Siegen. This paper addresses the detection of damages of the bearings in servo motors and presents new results of the industrial application of a new developed diagnosis method for detecting such based on frequency response analysis. In the experimental work a permanent magnet synchronous machine (PMSM) is used and two different cases are considered: first, only the machine under no-load condition is investigated, afterwards the driving machine was installed in a two-mass-system simulating real conditions. The experimental results show that the method for damage detection works in both cases. In contrast to former publications the present paper shows results obtained on bearing faults in motors coming from the field to the repair shop. The measurement of the required signals (current and speed) can be accomplished during operation of the plant in closed loop speed control. The system is excited by pseudo random binary test-signals (PRBS). The deviations between the frequency response obtained during the commissioning of the plant and the curve measured under fault condition serve as indicators for the damage.