Publication: Predictive modelling of a gas turbine In combined cycle operations using Statistical data mining techniques.
No Thumbnail Available
Date
2025-09
Authors
Ahmed Abed, Ahmed Ibrahim
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Mechanical vibrations in rotating equipment, particularly in gas turbine (GT) power units, often result from issues like misalignment, wear, unbalance, and bearing faults. As vibration monitoring is crucial, online systems are commonly deployed. A review of existing research indicates that 93.2% of studies focus on individual components rather than holistic GT unit monitoring. This study aims to develop predictive models for assessing the influence of physical parameters on vibration levels, active power output, and exhaust gas temperature (EGT) in the SGT-PAC 2000E unit's generator, turbine, and compressor. Using 1488 data instances collected over four months via the SPPA-T3000 online monitoring system, the study analyzed 24 physical parameters. Statistical analyses (including P-values and correlation coefficients) were conducted in Excel, while data pre-processing, model training, and validation were performed using WEKA. Key objectives included identifying critical GT features influencing system performance, visualizing vibration trends via color-coded charts, and developing predictive models using linear regression and machine learning algorithms.Results showed that ambient conditions (temperature, pressure, humidity) had negligible impact on GT parameters. Mechanical vibration velocities were consistent across paired components. The study produced 36 predictive models: 4 vibration velocity, 4 displacement, and 1 active power model per month. Velocity models achieved r > 0.99, displacement models had r > 0.70 across all components, and power models reached r > 0.98. EGT prediction using machine learning showed that the random forest algorithm yielded the highest accuracy.