Implementation of Vibrations Faults Monitoring and Detection on Gas Turbine System Based on the Support Vector Machine Approach.

Autor: Hadroug, Nadji, Iratni, Abdelhamid, Hafaifa, Ahmed, Alili, Bachir, Colak, Ilhami
Předmět:
Zdroj: Journal of Vibration Engineering & Technologies; Mar2024, Vol. 12 Issue 3, p2877-2902, 26p
Abstrakt: Purpose: Gas turbines play a critical role in the gas and hydrocarbon industry, but they are prone to failures and malfunctions that can impact their performance and safety. Therefore, it is crucial to implement effective measures for detecting, locating, and identifying these faults, as well as monitoring the operational state of gas turbines. This study proposes a fault detection approach using support vector machines (SVMs) for vibration monitoring of gas turbines. The objective is to develop and validate a reliable and efficient fault diagnosis system capable of estimating the operating state of the gas turbine in normal and degraded modes, while improving the accuracy and speed of fault detection. Methods: This paper applies the SVM method to vibration monitoring of a GE MS5002 gas turbine. The SVM method automatically extracts information about the turbine's operation from its input/output data and classifies faults caused by vibrations. Furthermore, the SVM method establishes appropriate models for turbine instability phenomena with minimal estimation errors. The optimal fault detection function for the gas turbine is determined by utilizing various kernel functions in the SVM method. The SVM-based solutions provide a robust monitoring strategy and performance indicators for vibration diagnosis of the gas turbine. Results: The results demonstrate the effectiveness of the SVM method in monitoring and detecting vibration faults in gas turbines, making it a practical diagnostic indicator. Experimental tests of the kernel-SVM method exhibit superior performance compared to other techniques, achieving fault detection efficiency rates of 99.44% to 99.96% for bearings. This confirms that the kernel SVM fault detection algorithm has a minimal prediction error rate in identifying vibration faults in gas turbines. Conclusions: This study presents the development of a SVMs approach for monitoring and detecting vibration defects in GE MS5002 turbine bearings, based on an experimental study using input/output operating data. The proposed fault diagnosis approach is validated through robustness tests conducted in real-time, specifically for the detection of vibration faults affecting the four bearings. This approach establishes diagnostic elements with improved performance and ensures the continuous and stable monitoring of the examined turbine components. [ABSTRACT FROM AUTHOR]
Databáze: Complementary Index
Externí odkaz: