Loading...

Identification of Possible Failures in Online Data of Gas Turbines based on Information Fusion using Artificial Intelligence

Haghighat Shoar, Amin | 2025

0 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 58063 (58)
  4. University: Sharif University of Technology, International Campus, Kish Island
  5. Department: Science and Engineering
  6. Advisor(s): Behzad, Mehdi; Mohammadi, Somayeh
  7. Abstract:
  8. Gas turbines' safe operation depends on the monitoring of performance thresholds and equipment limitations. This is accomplished through online data monitoring using sensors, with the collected data processed by monitoring software supplied by turbine manufacturers to extract information on operating conditions. The thresholds are usually set during manufacturing under standard conditions and saved in the control software. As the turbine ages and undergoes maintenance, these thresholds and operating conditions may change, requiring expert analysis to determine safe operating limits. Technological advancements have enabled the processing of operational data to establish patterns and create data-driven models for detecting and classifying faults. Artificial intelligence (AI) algorithms, with their nonlinear approximation capabilities and proficiency in large-scale data processing and abstract information extraction, offer reliable and dynamic methods for real-time performance monitoring and robustness. Due to the vastness and complexity of gas turbine operational data and the fact that the data labeling process in supervised machine learning algorithms for long-life turbines is usually a time-consuming process with detection errors, this research uses an unsupervised machine learning method with the K-Means algorithm in order to overcome this limitation. To validate the effectiveness of this approach, three supervised machine learning models feedforward neural networks, convolutional neural networks, and support vector machines are also implemented. To analyze the operating conditions and detect the healthy or unhealthy state of the gas turbine by using machine learning algorithms, various parameters including air intake, fuel conditions, oil temperature, bearing temperature, vibration, turbine shaft speed, and exhaust temperature, which were recorded by the operators over a year, were collected and, after converting them into digital data in an Excel table, were entered into the algorithms as CSV. The results obtained show that the k-means clustering method allows the identification of patterns and outliers based only on the inherent structure of the data and has acceptable accuracy in detecting anomalies in gas turbine performance parameters, so that the accuracy in the training was 93% and in the test was 89%. This achievement demonstrates the significant impact of unsupervised machine learning techniques in the field of gas turbine monitoring, which increases the reliability, efficiency, and cost-effectiveness of gas turbine operation
  9. Keywords:
  10. Gas Turbines ; Condition Monitoring ; Artificial Intelligence ; Fault Diagnosis ; Data Fusion ; Prognostics Failure

 Digital Object List

 Bookmark

No TOC