Scholarlink

<p>Gas turbines serve as essential assets in the natural gas transportation sector, where they are widely used for power generation. Despite their importance, these complex rotating machines are inherently susceptible to various operational issues. Such issues not only compromise reliability and efficiency, but also contribute to long-term performance degradation. To address these challenges, modern gas turbines are typically equipped with fault monitoring systems. These systems play a crucial role in minimizing the risk of failure, reducing the frequency of breakdowns, and supporting uninterrupted operation. In this study, a novel fault detection strategy is introduced for a twin-shaft gas turbine. The proposed approach integrates an adaptive neuro-fuzzy inference system with subtractive clustering hybridization (SC_ANFIS). Leveraging real-time processing of 50,000 samples of operational data, the method achieves a fault detection accuracy of 93.8%. Experimental evaluations confirm that the SC_ANFIS framework not only detects faults with high precision, but also effectively classifies fault types. This capability significantly reduces the occurrence of unplanned turbine shutdowns. Moreover, by implementing automatic residual thresholding, the system successfully decreases the false alarm rate to 41%. These results highlight the promise of subtractive clustering hybridization as a robust enhancement to gas turbine fault diagnosis.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Gas turbine fault’s detection using neuro-fuzzy adaptive inference system combined with subtractive clustering hybridization

  • 期刊论文

  • 发表日期：

    2025年9月19日

查看全文
  • Tarek Idris Bisker,
  • Nadji Hadroug,
  • Ahmed Hafaifa,
  • Abdelhamid Iratni,
  • Ahmed Saïd Nouri,
  • Ilhami Colak

摘要

Gas turbines serve as essential assets in the natural gas transportation sector, where they are widely used for power generation. Despite their importance, these complex rotating machines are inherently susceptible to various operational issues. Such issues not only compromise reliability and efficiency, but also contribute to long-term performance degradation. To address these challenges, modern gas turbines are typically equipped with fault monitoring systems. These systems play a crucial role in minimizing the risk of failure, reducing the frequency of breakdowns, and supporting uninterrupted operation. In this study, a novel fault detection strategy is introduced for a twin-shaft gas turbine. The proposed approach integrates an adaptive neuro-fuzzy inference system with subtractive clustering hybridization (SC_ANFIS). Leveraging real-time processing of 50,000 samples of operational data, the method achieves a fault detection accuracy of 93.8%. Experimental evaluations confirm that the SC_ANFIS framework not only detects faults with high precision, but also effectively classifies fault types. This capability significantly reduces the occurrence of unplanned turbine shutdowns. Moreover, by implementing automatic residual thresholding, the system successfully decreases the false alarm rate to 41%. These results highlight the promise of subtractive clustering hybridization as a robust enhancement to gas turbine fault diagnosis.