Abstract:Developing a maintenance schedule for different parts of a power plant can help to prevent serious system damage while also having a direct effect on reducing maintenance time and costs. Moreover, the proposed procedure can provide a better understanding of the health state of the system. The occurrence of problems (anomalous behavior) in power plants usually takes time and happens gradually. Predicting the possibility of anomalous behavior arising can therefore result in the prediction of a failure. In this paper, the authors tried to propose a new method for reducing the volume of input data by extracting the latent layer output of an Autoencoder, which was then applied to identify features. We achieved this by keeping relevant features and learning the anomalous behavior's encoded representation. Furthermore, the latent layer output from the encoder was used to train a Deep Long-Short-Term Memory (DLSTM) model with three hidden layers for fault occurrence prediction by considering an anomaly detection approach. To this end, a threshold approach was used to determine the possibility of a fault occurring in the turbine, with data collected from the Kirkuk gas turbine power plant. Piezoelectric accelerometer CA 202 was utilized as a sensor to collect data in industry. Since most of the recorded data were normal and the amount of abnormal data was small, various techniques, i.e., flipping and jittering, were employed to create synthetic abnormal data and augment the applicable dataset. In summary, according to the proposed methodology, the system is modelled as a time series classification problem to mimic the anomaly detection problem. The results showed that the accuracy of the trained model for detecting turbine vibration as a system abnormality is over 86%. Also, the implementation of the proposed prediction model on the experimental dataset yielded satisfactory results for detecting and predicting the occurrence of faults in Kirkuk gas turbine power plant.

Gas Turbine Anomaly Detection under Time-Varying Operation Conditions Based on Spectra Alignment and Self-Adaptive Normalization

Vibration Anomaly Detection of Wind Turbine Based on Temporal Convolutional Network and Support Vector Data Description

Real-time Novelty Detection of an Industrial Gas Turbine Using Performance Deviation Model and Extreme Function Theory

An Intelligent Anomaly Detection Method for Rotating Machinery Based on Vibration Vectors

Construction of Data-Driven Performance Digital Twin for a Real-World Gas Turbine Anomaly Detection Considering Uncertainty

Uncertainty‐aware Nuclear Power Turbine Vibration Fault Diagnosis Method Integrating Machine Learning and Heuristic Algorithm

Wind Turbine Fault Detection with Multimodule Feature Extraction Network and Adaptive Strategy

Long Short-Term Memory Network-Based Normal Pattern Group for Fault Detection of Three-Shaft Marine Gas Turbine

Anomaly detection and diagnosis for wind turbines using long short-term memory-based stacked denoising autoencoders and XGBoost

Fault detection in the gas turbine of the Kirkuk power plant: An anomaly detection approach using DLSTM-Autoencoder

Deep Multiscale Soft-Threshold Support Vector Data Description for Enhanced Heavy-Duty Gas Turbine Generator Sets' Anomaly Detection

Steam Turbine Anomaly Detection: An Unsupervised Learning Approach Using Enhanced Long Short-Term Memory Variational Autoencoder

Learning Multi-Pattern Normalities in the Frequency Domain for Efficient Time Series Anomaly Detection

Fault Detection of Wind Turbine Generator Bearing Using Attention-Based Neural Networks and Voting-Based Strategy

Performance-based fault diagnosis of a gas turbine engine using an integrated support vector machine and artificial neural network method

Aero-engine gas path anomaly monitoring based on unsupervised structure and deep slow feature representation

Anomaly Detection of Power Plant Equipment Using Long Short-Term Memory Based Autoencoder Neural Network

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

Multi-mode non-Gaussian variational autoencoder network with missing sources for anomaly detection of complex electromechanical equipment

Anomaly Detection for Wind Turbines Using Long Short-Term Memory-Based Variational Autoencoder Wasserstein Generation Adversarial Network under Semi-Supervised Training

Bearing fault diagnosis using weakly supervised long short-term memory