Advanced Temporal Deep Learning Framework for Enhanced Predictive Modeling in Industrial Treatment Systems

Results in Engineering, Journal Year: 2025, Volume and Issue: unknown, P. 104158 - 104158

Published: Jan. 1, 2025

Language: Английский

---

Improved network anomaly detection system using optimized autoencoder − LSTM

Expert Systems with Applications, Journal Year: 2025, Volume and Issue: unknown, P. 126854 - 126854

Published: Feb. 1, 2025

Language: Английский

---

LSTM-Autoencoder Deep Learning Model for Anomaly Detection in Electric Motor

Energies, Journal Year: 2024, Volume and Issue: 17(10), P. 2340 - 2340

Published: May 13, 2024

Anomaly detection is the process of detecting unusual or unforeseen patterns events in data. Many factors, such as malfunctioning hardware, malevolent activities, modifications to data’s underlying distribution, might cause anomalies. One key factors anomaly balancing trade-off between sensitivity and specificity. Balancing these trade-offs requires careful tuning algorithm consideration specific domain application. Deep learning techniques’ applications, LSTMs (long short-term memory algorithms), which are autoencoders for an anomaly, have garnered increasing attention recent years. The main goal this work was develop solution electrical machine using LSTM-autoencoder deep model. focused on anomalies motor’s variation vibrations three axes: axial (X), radial (Y), tangential (Z), indicative potential faults failures. presented model a combination two architectures; LSTM layers were added autoencoder order leverage capacity handling large amounts temporal To prove efficiency, we will create regular Python programming language TensorFlow framework, compare its performance with our LSTM-based models be trained same database, evaluated primary points: training time, loss function, MSE Based obtained results, it clear that shows significantly smaller values compared autoencoder. On other hand, performs better than LSTM, comparing time. It appears then, presents superior although slower standard due complexity layers.

Language: Английский

---

Deep Learning Architecture for Detecting SQL Injection Attacks Based on RNN Autoencoder Model

Mathematics, Journal Year: 2023, Volume and Issue: 11(15), P. 3286 - 3286

Published: July 26, 2023

SQL injection attacks are one of the most common types on Web applications. These exploit vulnerabilities in an application’s database access mechanisms, allowing attackers to execute unauthorized queries. In this study, we propose architecture for detecting using a recurrent neural network autoencoder. The proposed was trained publicly available dataset attacks. Then, it compared with several other machine learning models, including ANN, CNN, decision tree, naive Bayes, SVM, random forest, and logistic regression models. experimental results showed that approach achieved accuracy 94% F1-score 92%, which demonstrate its effectiveness QL high comparison models covered study.

Language: Английский

---

Vibration data-driven anomaly detection in UAVs: A deep learning approach

Engineering Science and Technology an International Journal, Journal Year: 2024, Volume and Issue: 54, P. 101702 - 101702

Published: May 13, 2024

Unmanned Aerial Vehicles (UAVs) are employed for diverse applications, including aerial surveillance and package delivery. However, the occurrence of faults, especially propeller failures, poses significant risks to safe efficient operations. Detecting such faults at an early stage is critical avoiding catastrophic outcomes ensuring reliability lifespan UAVs. To address this crucial need, study proposes a novel approach monitoring vibration signals using wavelet scattering long short-term memory (LSTM) autoencoder network. The LSTM can learn temporal patterns from input signals, whereas capture dynamics interactions various frequency components signals. First, deliberate modification was made one blades DJI M600 multi-rotor UAV deliberately induce vibration. proposed network then evaluated on acquired signal MTi-G-700 IMU. results showed that warning were generated all axes before failures occurred. Notably, earliest warnings obtained y-axis data within 100 s, while first z-axis recognized 130 s later. failure occurred roughly 280 s. experimental findings indicate method accurately detect anomalies could potentially lead failure.

Language: Английский

---

Transformer Core Fault Diagnosis via Current Signal Analysis with Pearson Correlation Feature Selection

Electronics, Journal Year: 2024, Volume and Issue: 13(5), P. 926 - 926

Published: Feb. 29, 2024

The role of transformers in power distribution is crucial, as their reliable operation essential for maintaining the electrical grid’s stability. Single-phase are highly versatile, making them suitable various applications requiring precise voltage control and isolation. In this study, we investigated fault diagnosis a 1 kVA single-phase transformer core subjected to induced faults. Our diagnostic approach involved using combination advanced signal processing techniques, such fast Fourier transform (FFT) Hilbert (HT), analyze current signals. analysis aimed differentiate characterize unique signatures associated with each type, utilizing statistical feature selection based on Pearson correlation machine learning classifier. results showed significant improvements all metrics classifier models, particularly k-nearest neighbor (KNN) algorithm, 83.89% accuracy computational cost 0.2963 s. For future studies, our focus will be deep models improve effectiveness proposed method.

Language: Английский

---

Anomaly Detection Methods for Industrial Applications: A Comparative Study

Electronics, Journal Year: 2023, Volume and Issue: 12(18), P. 3971 - 3971

Published: Sept. 20, 2023

Anomaly detection (AD) algorithms can be instrumental in industrial scenarios to enhance the of potentially serious problems at a very early stage. Of course, “Industry 4.0” revolution is fostering implementation intelligent data-driven decisions industry based on increasingly efficient machine learning (ML) algorithms. Most well-known AD methods use supervised approach focusing fault classification. They assume availability labeled data for both normal and anomalous classes. However, many environments, set instances more challenging obtain than data. Hence, this work implements an unsupervised two different using typical benchmark bearing-fault dataset. The first method relies manual extraction vibration metrics provided as input ML algorithm. second one deep (DL) approach, automatically latent representation from raw performance demonstrate that approaches distinguish state bearing faulty. DL methodology proves higher accuracy rate recognizing faults better ability provide information about size.

Language: Английский

---

Anomaly Detection in Gas Turbines Using Outlet Energy Analysis with Cluster-Based Matrix Profile

Energies, Journal Year: 2024, Volume and Issue: 17(3), P. 653 - 653

Published: Jan. 30, 2024

Gas turbines play a key role in generating power. It is really important that they work efficiently, safely, and reliably. However, their performance can be adversely affected by factors such as component wear, vibrations, temperature fluctuations, often leading to abnormal patterns indicative of potential failures. As result, anomaly detection has become an area active research. Matrix Profile (MP) methods have emerged promising solution for identifying significant deviations time series data from normal operational patterns. While most existing MP focus on vibration analysis gas turbines, this paper introduces novel approach using the outlet power signal. This modified approach, termed Cluster-based (CMP) analysis, facilitates identification subsequent within turbine engine system. Significantly, CMP not only accelerates processing speed, but also provides user-friendly support information operators. The experimental results real-world demonstrate effectiveness our early anomalies system

Language: Английский

---

A Mold Damage Monitoring Algorithm for Power Metallurgy Molding Machines Using Bidirectional Long Short-Term Memory on an Internet of Things Platform

Sensors, Journal Year: 2025, Volume and Issue: 25(7), P. 2143 - 2143

Published: March 28, 2025

In this paper, an analysis and monitoring algorithm is proposed for mold health evaluation using vibration data. Two inertial measurement units (IMUs) embedded system are first used to acquire data from a powder metallurgy molding machine. These collected on Internet of Things (IoT) platform the Message Queueing Telemetry Transport (MQTT) protocol. For analysis, signal Z axis segmented label contact section upper middle molds, corresponding stamping friction X, Y, axes extracted. Using only historical normal stamping, Bidirectional Long Short-Term Memory (Bi-LSTM) model with attention mechanism trained predict vibrations several minutes in advance. By comparing predicted observed at current time, mean square errors (MSEs) calculated evaluate status mold. Several ablation experiments were conducted assess performance model. The average MSE values samples abnormal smaller than 0.5 larger 1.0, respectively. experimental results confirm that prediction indicators can effectively notify operators An early warning damage was successfully implemented, enhancing predictive maintenance.

Language: Английский

---

A Comparison of Deep Learning Algorithms for Anomaly Detection in Discrete Mechanical Systems

Applied Sciences, Journal Year: 2023, Volume and Issue: 13(9), P. 5683 - 5683

Published: May 5, 2023

The application of intelligent systems for structural health monitoring is investigated. A change in the nominal configuration can be related to a defect that has monitored before it reaches critical condition. Evidently, ability automatically detect changes structure very attractive feature. When there no prior knowledge on system, deep learning models could effectively and enhance capability determining damage location. However, acquisition data damaged structures not always practical. In this paper, two approaches, physics-informed autoencoder simple data-driven autoencoder, are applied test rig consisting small four-storey building model. Modifications system simulated by changing stiffness springs. Both machine algorithms outperform traditional approach based an experimental modal analysis. Moreover, increased potential neural networks locate confirmed.

Language: Английский

---