Abstract One of the main functions of gas turbine monitoring is to estimate important unmeasured variables, for instance, thrust and power. Existing methods are too complex for an online monitoring system. Moreover, they do not extract diagnostic features from the estimated variables, making them unusable for diagnostics. Two of our previous studies began to address the problem of “light” algorithms for online estimation of unmeasured variables. The first study deals with models for unmeasured thermal boundary conditions of a turbine blade. These models allow an enhanced prediction of blade lifetime and are sufficiently simple to be used online. The second study introduces unmeasured variable deviations and proves their applicability. However, the algorithms developed were dependent on a specific engine and a specific variable. The present paper proposes a universal algorithm to estimate and monitor any unmeasured gas turbine variables. This algorithm is based on simple data-driven models and can be used in online monitoring systems. It is evaluated on real data of two different engines affected by compressor fouling. The results prove that the estimates of unmeasured variables are sufficiently accurate, and the deviations of these variables are good diagnostic features. Thus, the algorithm is ready for practical implementation.

中文翻译：

---

在线监测系统中燃气轮机未测量变量的估计

摘要 燃气轮机监测的主要功能之一是估计重要的未测量变量，例如推力和功率。现有方法对于在线监控系统来说过于复杂。此外，它们不会从估计变量中提取诊断特征，因此无法用于诊断。我们之前的两项研究开始解决用于在线估计未测量变量的“轻量级”算法问题。第一项研究涉及涡轮叶片未测量热边界条件的模型。这些模型可以增强对叶片寿命的预测，并且非常简单，可以在线使用。第二项研究介绍了未测量的变量偏差并证明了它们的适用性。然而，开发的算法取决于特定的引擎和特定的变量。本文提出了一种通用算法来估计和监控任何未测量的燃气轮机变量。该算法基于简单的数据驱动模型，可用于在线监控系统。它是根据受压缩机积垢影响的两种不同发动机的真实数据进行评估的。结果证明未测量变量的估计足够准确，这些变量的偏差是良好的诊断特征。因此，该算法已准备好进行实际实施。结果证明未测量变量的估计足够准确，这些变量的偏差是良好的诊断特征。因此，该算法已准备好进行实际实施。结果证明未测量变量的估计足够准确，这些变量的偏差是良好的诊断特征。因此，该算法已准备好进行实际实施。