Mohamed Mostafa Ramadan Ahmed

Analog circuits are highly susceptible to faults due to their complex composition; consequently, fault diagnosis is crucial for ensuring and enhancing their operational stability. This study proposes a fault diagnosis approach for a Sallen-Key filter circuit based on the analysis of frequency response signals. The method employs Discrete Wavelet Transform (DWT) and Ensemble Empirical Mode Decomposition (EEMD) techniques in conjunction with Autoencoder (AE) and Artificial Neural Network (ANN) models. Following signal analysis, AE models were trained on the steady-state conditions processed via DWT and EEMD (designated as AE-DWT and AE-EEMD). The system achieved a False Alarm Rate (FAR) of 1.99%, with fault detection performance further improved through the integration of the AE-DWT-EEMD approach. The fault detection stage is followed by the fault classification stage, which involves training an ANN model based on statistical features derived from latent feature learning. The proposed technique, specifically the ANN-EEMD model, outperforms the ANN-DWT model, achieving a False Alarm Rate (FAR) of 0% and detection and classification accuracy of 100%. The study utilized Monte Carlo analysis to generate a large dataset representing realistic testing conditions, comprising 201 normal cases and 5,228 abnormal cases (including short-circuit fault (SCF), soft fault1 (SFP), soft fault2 (SFN), and open-circuit fault OCF). The Sallen-Key filter model used in this study was simulated using Multisim, and the effectiveness of the proposed technique was verified and evaluated using MATLAB. The study demonstrated the technique's effectiveness and high reliability in fault detection and classification under realistic operating conditions.

Power systems, particularly distribution networks, are currently facing increasing challenges due to rising energy demand and the growing complexity of modern grids. The integration of Distributed Generators (DGs) within Microgrids (MGs) introduces additional fault current contributions, which significantly modify short-circuit levels. This change can result in miscoordination problems in existing Overcurrent Relay (OCR)-based protection systems, including both maloperation and delayed tripping. In addition, the increased fault current contribution from MGs may exceed the design limits of equipment such as circuit breakers (CBs), exposing them to higher electrical stress and increasing the risk of failure to operate properly. Consequently, these issues collectively degrade the overall reliability and security of the power system. To overcome these challenges, installing a Directional Fault Current Limiter (DFCL) between the microgrid and the upstream network is considered an effective mitigation approach. The DFCL helps in controlling fault current levels and improving protection coordination. However, determining the optimal impedance setting of the DFCL is a complex task, as it must simultaneously ensure proper coordination among OCRs and reduce short-circuit levels to enhance system reliability. In this context, this paper proposes an analytical methodology for determining the optimal DFCL impedance based on Thevenin equivalent impedance analysis. The proposed approach provides a straightforward impedance sizing procedure that restores overcurrent relay coordination following distributed generation integration while maintaining system reliability. Unlike optimization-based methods, the proposed analytical approach determines the required DFCL impedance directly from the equivalent Thevenin impedance without iterative computation, making it computationally efficient and suitable for real-time engineering application The methodology is validated under different operating scenarios involving synchronous generators, wind turbine generators, and photovoltaic systems. The results demonstrate that the proposed approach effectively restores relay coordination, limits excessive fault currents, and enhances the reliability of microgrid protection without requiring adaptive protection schemes or additional relay installation.

Overhead transmission lines, particularly the high-voltage direct current transmission lines (HVDC-TLs) employed in this study, are utilized for power transmission, which is one of the most crucial aspects of electrical power systems. Therefore, one of the most critical parts of guaranteeing the continuous and effective operation of these lines and reducing service interruptions for users is the very accurate detection of faults. A deep undercomplete autoencoder for protection based on sensor data (ISS) in HVDC-TL is presented in this work. This study presents a proposed technique for unsupervised fault detection and classification using unsupervised clustering with label-based validation. Under normal operating conditions, such as fluctuations in transmission line resistance and system noise, an autoencoder is trained on all expected HVDC system scenarios to reconstruct input signals. After applying the Discrete Wavelet Transform (DWT) technique to DC signals, a robust fault detection system. For fault classification, using the K-Means clustering algorithm, the proposed technique relies on receiving current signals from the ISS and extracting statistical, time-domain, and frequency-domain characteristics based on the relationships presented in this study. The proposed method presents an approach demonstrating how to integrate deep learning autoencoder characteristics with unsupervised assembly to protect HVDC systems. The study results show that the proposed method successfully distinguishes between normal operating conditions and fault conditions, resulting in high fault detection accuracy. The study presents a robust fault detection technique under varying normal and abnormal operating conditions.

Due to the increasing dependence on clean energy, grid-connected solar power plants have become an essential component of modern electrical grids. As reliance on these systems grows, safeguarding them against faults has become critical, making the development of more advanced and reliable techniques for fault detection and classification a key requirement for safe operation. Traditional methods are often inefficient in terms of speed and reliability. Consequently, this study presents a proposed method for fault identification and classification in grid-connected photovoltaic systems. The approach is based on continuous acquisition of AC signals from the measurement point (IGrid), followed by signal processing using Discrete Wavelet Transform (DWT), which is employed to train an Autoencoder (AE) under steady-state operating conditions. Continuous training of AC signals under normal operation forms the basis of this work. After completing the AE training phase and determining the square error signal, DWT is applied again. The fault identification process is then performed in three stages: the first stage is based on calculating the Pearson coefficient, the second stage relies on determining the Signal-to-Noise Ratio (SNR), where positive values indicate stable operation and negative values indicate instability, and the third stage, known as the Performance Response Evaluation Signal (PRES), determines threshold boundaries based on three-phase AC signals, ensuring robust fault detection. The fault classification phase is subsequently carried out using Artificial Neural Network (ANN) training, divided into two stages based on pitch detection as a fault signature to extract training features. MATLAB R2024a is used to simulate the system and validate the effectiveness of the proposed method. The results confirm that the technique is a robust and reliable algorithm for fault detection and classification under various operating conditions.

The integration of intelligent machine health monitoring systems (MHMS) into rotating machinery and industrial equipment has accelerated due to the growing demand for productivity, dependability, and predictive maintenance in contemporary production systems. Deep learning (DL), one of the new methods, has proven to be quite effective at directly extracting discriminative features from high‐dimensional vibration and acoustic signals, doing away with the requirement for human feature engineering. The DL approaches used for vibration‐based defect diagnosis and prognostics of rotating machinery are critically and methodically reviewed in this research. Signal properties, fault modes, and experimental settings are thoroughly investigated in representative public benchmark datasets, such as those from Case Western Reserve University, Paderborn University, IMS, and Pronostia. Major DL architectures, convolutional neural networks (CNN), autoencoders (AE), deep belief networks (DBN), recurrent neural networks (RNN/LSTM), and hybrid CNN–LSTM frameworks are compared in terms of feature learning mechanisms, robustness to noise, computational complexity, and suitability for real‐time industrial deployment. In addition, the paper also looks at recent developments in few‐shot and meta‐learning techniques that deal with limited fault samples and data imbalance in real‐world mechanical systems. From the standpoint of mechanical engineering, the difficulties of domain shift, interpretability, computational limitations, and edge implementation are severely examined. Lastly, a real‐time deployment‐oriented Few‐Shot based MS‐1D‐CNN framework deployed in an industrial environment with a focus on scalability, robustness, and integration with industrial control systems is described for real‐time intelligent MHMS in Industry 4.0 scenarios. Consolidated insights and future research prospects for creating dependable, flexible, and computationally effective DL solutions for rotating machinery health management are provided by this review.

With increasing population growth, energy demand has become a fundamental necessity, leading to the expansion of power generation and transmission systems. Due to the long distances between generation sites and load centers, HVDC transmission lines have emerged as an optimal solution. However, their widespread use increases their exposure to critical faults. This paper proposes a low-complexity technique for fault detection, classification, and location in HVDC transmission lines using single-ended current measurements at the rectifier side.The proposed method introduces a novel feature, SHY, defined as the hypotenuse of a triangle formed by the steady-state current (adjacent side) and the deviation between the measured sample and steady-state value (opposite side). Under normal conditions, SHY approaches zero, while it slightly deviates in the presence of noise and significantly changes during faults.For fault classification, an ANFIS-based model is trained using extracted features (SHY and AX) to distinguish between internal DC faults, external AC faults (at rectifier or inverter sides), and open-circuit faults. Following classification, fault location is determined using the maximum current value, exploiting the near-linear relationship of fault current data. Linear regression is applied, and accuracy is enhanced through ANFIS-based data linearization and segmentation of the transmission line into protection zones.The proposed technique is validated under noisy conditions, and its classification and localization performance is tested with fault resistance up to 300 Ω. The HVDC system is modeled in PSCAD, while MATLAB is used for validation, demonstrating the method’s effectiveness and reliability for protection applications.