Enhanced Autoregressive Integrated Moving Average Model for Anomaly Detection in Power Plant Operations DOI Open Access
Aliya Fahmi, Kazem Reza Kashyzadeh, Siamak Ghorbani

et al.

International journal of engineering. Transactions B: Applications, Journal Year: 2024, Volume and Issue: 37(8), P. 1691 - 1699

Published: Jan. 1, 2024

This study introduces an Enhanced Autoregressive Integrated Moving Average (E-ARIMA) model for anomaly detection in time-series data, using vibrations monitored by CA 202 accelerometers at the Kirkuk Gas Power Plant as a case study. The objective is to overcome limitations of traditional ARIMA models analyzing non-linear and dynamic nature industrial sensory data. novel proposed methodology includes data preparation through linear interpolation address dataset gaps, stationarity confirmation via Augmented Dickey-Fuller Test, optimization against Akaike Information Criterion, with specialized cross-validation technique. results show that E-ARIMA has superior performance compared conventional Seasonal (SARIMA) Vector models. In this regard, Mean Absolute Error (MAE), Squared (MSE), Root (RMSE) criteria were utilized evaluation. Finally, most important achievement research highlight enhanced predictive accuracy model, making it potent tool applications such machinery health monitoring, where early anomalies crucial prevent costly downtimes facilitate maintenance planning.

Language: Английский

Novel GA-Based DNN Architecture for Identifying the Failure Mode with High Accuracy and Analyzing Its Effects on the System DOI Creative Commons
Naeim Rezaeian,

Regina Gurina,

О. А. Салтыкова

et al.

Applied Sciences, Journal Year: 2024, Volume and Issue: 14(8), P. 3354 - 3354

Published: April 16, 2024

Symmetric data play an effective role in the risk assessment process, and, therefore, integrating symmetrical information using Failure Mode and Effects Analysis (FMEA) is essential implementing projects with big data. This proactive approach helps to quickly identify risks take measures address them. However, this task always time-consuming costly. On other hand, there need for expert field carry out process manually. Therefore, present study, authors propose a new methodology automatically manage through deep-learning technique. Moreover, due different nature of data, it not possible consider single neural network architecture all To overcome problem, Genetic Algorithm (GA) was employed find best hyperparameters. Finally, were processed predicted proposed without sending servers, i.e., external servers. The results analysis first risk, latency real-time processing, showed that can improve detection accuracy failure mode by 71.52%, 54.72%, 72.47%, 75.73% compared unique algorithm activation function Relu number neurons 32, respectively, related one, two, three, four hidden layers.

Language: Английский

Citations

12
A hybrid GRU and LSTM-based deep learning approach for multiclass structural damage identification using dynamic acceleration data DOI
T. K. Das, Shyamal Guchhait

Engineering Failure Analysis, Journal Year: 2025, Volume and Issue: unknown, P. 109259 - 109259

Published: Jan. 1, 2025

Language: Английский

Citations

1
A model mismatch method for gas turbine fault detection DOI
Junqi Luan, Shuying Li, Yunpeng Cao

et al.

Measurement, Journal Year: 2025, Volume and Issue: unknown, P. 116680 - 116680

Published: Jan. 1, 2025

Language: Английский

Citations

0
An Explainable AI approach for detecting failures in air pressure systems DOI
Shawqi Mohammed Farea, Mehmet Emin Mumcuoğlu, Mustafa Ünel

et al.

Engineering Failure Analysis, Journal Year: 2025, Volume and Issue: unknown, P. 109441 - 109441

Published: Feb. 1, 2025

Citations

0
Flight Anomaly Detection and Localization based on flight data fusion and Random Channel Masking DOI
Jie Zhong, Heng Zhang, Qiang Miao

et al.

Applied Soft Computing, Journal Year: 2025, Volume and Issue: unknown, P. 113023 - 113023

Published: March 1, 2025

Language: Английский

Citations

0
LSTM-Autoencoder Based Detection of Time-Series Noise Signals for Water Supply and Sewer Pipe Leakages DOI Open Access

Yungyeong Shin,

Kwang Yoon Na,

Si Eun Kim

et al.

Water, Journal Year: 2024, Volume and Issue: 16(18), P. 2631 - 2631

Published: Sept. 16, 2024

The efficient management of urban water distribution networks is crucial for public health and development. One the major challenges quick accurate detection leaks, which can lead to loss, infrastructure damage, environmental hazards. Many existing leak methods are ineffective, especially in complex aging pipeline networks. If these limitations not overcome, it result a chain failures, exacerbating increasing repair costs, causing shortages risks. issue further complicated by demand, climate change, population growth. Therefore, there an urgent need intelligent systems that overcome traditional methodologies leverage sophisticated data analysis machine learning technologies. In this study, we propose reliable advanced method detecting leaks pipes using framework based on Long Short-Term Memory (LSTM) combined with autoencoders. designed manage temporal dimension time-series enhanced ensemble techniques, making sensitive subtle signals indicating while robustly dealing noise signals. Through integration signal processing pattern recognition, learning-based model addresses problem, providing system enhances protection resource management. proposed approach greatly accuracy precision detection, essential contributions field offering promising prospects future sustainable strategies.

Language: Английский

Citations

3
Advancements in Gas Turbine Fault Detection: A Machine Learning Approach Based on the Temporal Convolutional Network–Autoencoder Model DOI Creative Commons
Al-Tekreeti Watban Khalid Fahmi, Kazem Reza Kashyzadeh, Siamak Ghorbani

et al.

Applied Sciences, Journal Year: 2024, Volume and Issue: 14(11), P. 4551 - 4551

Published: May 25, 2024

To tackle the complex challenges inherent in gas turbine fault diagnosis, this study uses powerful machine learning (ML) tools. For purpose, an advanced Temporal Convolutional Network (TCN)–Autoencoder model was presented to detect anomalies vibration data. By synergizing TCN capabilities and Multi-Head Attention (MHA) mechanisms, introduces a new approach that performs anomaly detection with high accuracy. train test proposed model, bespoke dataset of CA 202 accelerometers installed Kirkuk power plant used. The not only outperforms traditional GRU–Autoencoder, LSTM–Autoencoder, VAE models terms accuracy, but also shows Mean Squared Error (MSE = 1.447), Root (RMSE 1.193), Absolute (MAE 0.712). These results confirm effectiveness TCN–Autoencoder increasing predictive maintenance operational efficiency plants.

Language: Английский

Citations

2
Detecting APS failures using LSTM-AE and anomaly transformer enhanced with human expert analysis DOI
Mehmet Emin Mumcuoğlu, Shawqi Mohammed Farea, Mustafa Ünel

et al.

Engineering Failure Analysis, Journal Year: 2024, Volume and Issue: 165, P. 108811 - 108811

Published: Aug. 24, 2024

Language: Английский

Citations

2
Enhanced Autoregressive Integrated Moving Average Model for Anomaly Detection in Power Plant Operations DOI Open Access
Aliya Fahmi, Kazem Reza Kashyzadeh, Siamak Ghorbani

et al.

International journal of engineering. Transactions B: Applications, Journal Year: 2024, Volume and Issue: 37(8), P. 1691 - 1699

Published: Jan. 1, 2024

This study introduces an Enhanced Autoregressive Integrated Moving Average (E-ARIMA) model for anomaly detection in time-series data, using vibrations monitored by CA 202 accelerometers at the Kirkuk Gas Power Plant as a case study. The objective is to overcome limitations of traditional ARIMA models analyzing non-linear and dynamic nature industrial sensory data. novel proposed methodology includes data preparation through linear interpolation address dataset gaps, stationarity confirmation via Augmented Dickey-Fuller Test, optimization against Akaike Information Criterion, with specialized cross-validation technique. results show that E-ARIMA has superior performance compared conventional Seasonal (SARIMA) Vector models. In this regard, Mean Absolute Error (MAE), Squared (MSE), Root (RMSE) criteria were utilized evaluation. Finally, most important achievement research highlight enhanced predictive accuracy model, making it potent tool applications such machinery health monitoring, where early anomalies crucial prevent costly downtimes facilitate maintenance planning.

Language: Английский

Citations

0