Explore the vast range of titles available.

The damaging effects of corona faults have made them a major concern in metal-clad switchgear, requiring extreme caution during operation. Corona faults are also the primary cause of flashovers in medium-voltage metal-clad electrical equipment. The root cause of this issue is an electrical breakdown of the air due to electrical stress and poor air quality within the switchgear. Without proper preventative measures, a flashover can occur, resulting in serious harm to workers and equipment. As a result, detecting corona faults in switchgear and preventing electrical stress buildup in switches is critical. Recent years have seen the successful use of Deep Learning (DL) applications for corona and non-corona detection, owing to their autonomous feature learning capability. This paper systematically analyzes three deep learning techniques, namely 1D-CNN, LSTM, and 1D-CNN-LSTM hybrid models, to identify the most effective model for detecting corona faults. The hybrid 1D-CNN-LSTM model is deemed the best due to its high accuracy in both the time and frequency domains. This model analyzes the sound waves generated in switchgear to detect faults. The study examines model performance in both the time and frequency domains. In the time domain analysis (TDA), 1D-CNN achieved success rates of 98%, 98.4%, and 93.9%, while LSTM obtained success rates of 97.3%, 98.4%, and 92.4%. The most suitable model, the 1D-CNN-LSTM, achieved success rates of 99.3%, 98.4%, and 98.4% in differentiating corona and non-corona cases during training, validation, and testing. In the frequency domain analysis (FDA), 1D-CNN achieved success rates of 100%, 95.8%, and 95.8%, while LSTM obtained success rates of 100%, 100%, and 100%. The 1D-CNN-LSTM model achieved a 100%, 100%, and 100% success rate during training, validation, and testing. Hence, the developed algorithms achieved high performance in identifying corona faults in switchgear, particularly the 1D-CNN-LSTM model due to its accuracy in detecting corona faults in both the time and frequency domains.

Real‐time foot pressure monitoring using wearable smart systems, with comprehensive foot health monitoring and analysis, can enhance quality of life and prevent foot‐related diseases. However, traditional smart insole solutions that rely on basic data analysis methods of manual feature extraction are limited to real‐time plantar pressure mapping and gait analysis, failing to meet the diverse needs of users for comprehensive foot healthcare. To address this, we propose a deep learning‐enabled smart insole system comprising a plantar pressure sensing insole, portable circuit board, deep learning and data analysis blocks, and software interface. The capacitive sensing insole can map both static and dynamic plantar pressure with a wide range over 500 kPa and excellent sensitivity. Statistical tools are used to analyze long‐term foot pressure usage data, providing indicators for early prevention of foot diseases and key data labels for deep learning algorithms to uncover insights into the relationship between plantar pressure patterns and foot issues. Additionally, a segmentation method assisted deep learning model is implemented for exercise‐fatigue recognition as a proof of concept, achieving a high classification accuracy of 95%. The system also demonstrates various foot healthcare applications, including daily activity statistics, exercise injury avoidance, and diabetic foot ulcer prevention. This paper proposes a deep‐learning enabled wearable smart insole system. With a highly sensitive capacitive pressure sensing insole and deep‐learning enabled data analysis process, it provides long‐term data analysis, early prevention indicators, and deep insights into the relationship between plantar pressure and foot issues. Various foot‐healthcare applications are proven, including daily statistics, exercise injury avoidance, and diabetic foot ulcer prevention.

Fault diagnosis (FD) plays a vital role in building a smart factory regarding system reliability improvement and cost reduction. Recent deep learning-based methods have been applied for FD and have obtained excellent performance. However, most of them require sufficient historical labeled data to train the model which is difficult and sometimes not available. Moreover, the big size model increases the difficulties for real-time FD. Therefore, this article proposed a domain adaptive and lightweight framework for FD based on a one-dimension convolutional neural network (1D-CNN). Particularly, 1D-CNN is designed with a structure of autoencoder to extract the rich, robust hidden features with less noise from source and target data. The extracted features are processed by correlation alignment (CORAL) to minimize domain shifts. Thus, the proposed method could learn robust and domain-invariance features from raw signals without any historical labeled target domain data for FD. We designed, trained, and tested the proposed method on CRWU bearing data sets. The sufficient comparative analysis confirmed its effectiveness for FD.

Accurate Earth orientation parameter (EOP) predictions are needed for many applications, e.g., for the tracking and navigation of interplanetary spacecraft missions. One of the most difficult parameters to forecast is the length of day (LOD), which represents the variation in the Earth’s rotation rate since it is primarily affected by the torques associated with changes in atmospheric circulation. In this study, a new-generation time-series prediction algorithm is developed. The one-dimensional convolutional neural network (1D CNN), which is one of the deep learning methods, is introduced to model and predict the LOD using the IERS EOP 14 C04 and axial Z component of the atmospheric angular momentum (AAM), which was taken from the German Research Centre for Geosciences (GFZ) since it is strongly correlated with the LOD changes. The prediction procedure operates as follows: first, we detrend the LOD and Z-component series using the LS method, then, we obtain the residual series of each one to be used in the 1D CNN prediction algorithm. Finally, we analyze the results before and after introducing the AAM function. The results prove the potential of the proposed method as an optimal algorithm to successfully reconstruct and predict the LOD for up to 7 days.

Road surface conditions significantly impact driving safety and maintenance costs. Especially in connected and automated vehicles (CAVs), the road surface type recognition is critical for environmental perception. Traditional road surface recognition methods face limitations in feature extraction, so an improved one-dimensional convolutional neural network (1D-CNN) algorithm was proposed based on the VGG16 architecture. A vibration signal acquisition system was developed to efficiently acquire high-quality vehicle vibration signals. The optimized 1D-CNN algorithm model contains only 101.6 k parameters, significantly reducing computational cost and training time while maintaining high accuracy. Data augmentation, Adam optimization algorithm and L2 regularization were integrated to enhance generalization capabilities and suppress overfitting. On public datasets and actual vehicles tests, recognition accuracy rate reached 99.3% and 99.4%, respectively, substantially outperforming conventional methods. The algorithm also exhibited strong adaptability to different data sources. The research findings have implications for the accurate and efficient identification of road surfaces.

In the field of sensors, in areas such as industrial, clinical, or environment, it is common to find one dimensional (1D) formatted data (e.g., electrocardiogram, temperature, power consumption). A very promising technique for modelling this information is the use of One Dimensional Convolutional Neural Networks (1D CNN), which introduces a new challenge, namely how to define the best architecture for a 1D CNN. This manuscript addresses the concept of One Dimensional Neural Architecture Search (1D NAS), an approach that automates the search for the best combination of Neuronal Networks hyperparameters (model architecture), including both structural and training hyperparameters, for optimising 1D CNNs. This work includes the implementation of search processes for 1D CNN architectures based on five strategies: greedy, random, Bayesian, hyperband, and genetic approaches to perform, collect, and analyse the results obtained by each strategy scenario. For the analysis, we conducted 125 experiments, followed by a thorough evaluation from multiple perspectives, including the best-performing model in terms of accuracy, consistency, variability, total running time, and computational resource consumption. Finally, by presenting the optimised 1D CNN architecture, the results for the manuscript’s research question (a real-life clinical case) were provided.

Emotions play a crucial role in human thoughts, cognitive processes, and decision-making. EEG has become a widely utilized tool in emotion recognition due to its high temporal resolution, real-time monitoring capabilities, portability, and cost-effectiveness. In this paper, we propose a novel end-to-end emotion recognition method from EEG signals, called MSDCGTNet, which is based on the Multi-Scale Dynamic 1D CNN and the Gated Transformer. First, the Multi-Scale Dynamic CNN is used to extract complex spatial and spectral features from raw EEG signals, which not only avoids information loss but also reduces computational costs associated with the time-frequency conversion of signals. Then, the Gated Transformer Encoder is utilized to capture global dependencies of EEG signals. This encoder focuses on specific regions of the input sequence while reducing computational resources through parallel processing with the improved multi-head self-attention mechanisms. Third, the Temporal Convolution Network is used to extract temporal features from the EEG signals. Finally, the extracted abstract features are fed into a classification module for emotion recognition. The proposed method was evaluated on three publicly available datasets: DEAP, SEED, and SEED_IV. Experimental results demonstrate the high accuracy and efficiency of the proposed method for emotion recognition. This approach proves to be robust and suitable for various practical applications. By addressing challenges posed by existing methods, the proposed method provides a valuable and effective solution for the field of Brain-Computer Interface (BCI).

Rainfall–runoff modelling is complicated due to numerous complex interactions and feedback in the water cycle among precipitation and evapotranspiration processes, and also geophysical characteristics. Consequently, the lack of geophysical characteristics such as soil properties leads to difficulties in developing physical and analytical models when traditional statistical methods cannot simulate rainfall–runoff accurately. Machine learning techniques with data-driven methods, which can capture the nonlinear relationship between prediction and predictors, have been rapidly developed in the last decades and have many applications in the field of water resources. This study attempts to develop a novel 1D convolutional neural network (CNN), a deep learning technique, with a ReLU activation function for rainfall–runoff modelling. The modelling paradigm includes applying two convolutional filters in parallel to separate time series, which allows for the fast processing of data and the exploitation of the correlation structure between the multivariate time series. The developed modelling framework is evaluated with measured data at Chau Doc and Can Tho hydro-meteorological stations in the Vietnamese Mekong Delta. The proposed model results are compared with simulations of long short-term memory (LSTM) and traditional models. Both CNN and LSTM have better performance than the traditional models, and the statistical performance of the CNN model is slightly better than the LSTM results. We demonstrate that the convolutional network is suitable for regression-type problems and can effectively learn dependencies in and between the series without the need for a long historical time series, is a time-efficient and easy to implement alternative to recurrent-type networks and tends to outperform linear and recurrent models.

Hardware-in-the-Loop (HIL) has been recommended by ISO 26262 as an essential test bench for determining the safety and reliability characteristics of automotive software systems (ASSs). However, due to the complexity and the huge amount of data recorded by the HIL platform during the testing process, the conventional data analysis methods used for detecting and classifying faults based on the human expert are not realizable. Therefore, the development of effective means based on the historical data set is required to analyze the records of the testing process in an efficient manner. Even though data-driven fault diagnosis is superior to other approaches, selecting the appropriate technique from the wide range of Deep Learning (DL) techniques is challenging. Moreover, the training data containing the automotive faults are rare and considered highly confidential by the automotive industry. Using hybrid DL techniques, this study proposes a novel intelligent fault detection and classification (FDC) model to be utilized during the V-cycle development process, i.e., the system integration testing phase. To this end, an HIL-based real-time fault injection framework is used to generate faulty data without altering the original system model. In addition, a combination of the Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) is employed to build the model structure. In this study, eight types of sensor faults are considered to cover the most common potential faults in the signals of ASSs. As a case study, a gasoline engine system model is used to demonstrate the capabilities and advantages of the proposed method and to verify the performance of the model. The results prove that the proposed method shows better detection and classification performance compared to other standalone DL methods. Specifically, the overall detection accuracies of the proposed structure in terms of precision, recall and F1-score are 98.86%, 98.90% and 98.88%, respectively. For classification, the experimental results also demonstrate the superiority under unseen test data with an average accuracy of 98.8%.

Smart devices with sensors now enable continuous measurement of activities of daily living. Accordingly, various human activity recognition (HAR) experiments have been carried out, aiming to convert the measures taken from smart devices into physical activity types. HAR can be applied in many research areas, such as health assessment, environmentally supported living systems, sports, exercise, and security systems. The HAR process can also detect activity-based anomalies in daily life for elderly people. Thus, this study focused on sensor-based activity recognition, and we developed a new 1D-CNN-based deep learning approach to detect human activities. We evaluated our model using raw accelerometer and gyroscope sensor data on three public datasets: UCI-HAPT, WISDM, and PAMAP2. Parameter optimization was employed to define the model’s architecture and fine-tune the final design’s hyper-parameters. We applied 6, 7, and 12 classes of activity recognition to the UCI-HAPT dataset and obtained accuracy rates of 98%, 96.9%, and 94.8%, respectively. We also achieved an accuracy rate of 97.8% and 90.27% on the WISDM and PAMAP2 datasets, respectively. Moreover, we investigated the impact of using each sensor data individually, and the results show that our model achieved better results using both sensor data concurrently.