Abstract

At present, steam turbine manufacturers are interested in delivering diagnostic equipment systems for steam turbine units (STU) together with automatic control systems. These diagnostic systems should be adapted to a specific turbine size (or modification). The diagnostic system adaptation depends on the algorithm type used to detect defects. The following types of algorithms can be singled out: computational ones determining the defect using a computational model; expert ones based on probabilistic methods (using the summation of weighting coefficients of malfunction symptoms and the Bayes theorem); and digital ones based on the phenomenological relationships among process parameters, equipment state parameters, and other indicators obtained over the test period of equipment’s operation. The computational algorithms can be integrated into the turbine APCS and executed online (such as control system algorithms and steam admission system algorithms) or be used for handling postoperative tasks (such as diagnostics of auxiliary equipment or solving technical and economic problems). Expert algorithms are employed when development of a computational model involves great difficulties (e.g., in diagnosing a turbine thermal-expansion monitoring system or vibration-monitoring system). Digital algorithms are not related to the type of tasks, to specific equipment, or its features or operating modes. The degree of definiteness of a detected defect depends on the type of algorithm. The need to adapt diagnostic systems to other equipment involves not only the development of new diagnostic algorithms but also the formalization of the discourse. By this is meant the selection and justification of a standard state (prototype), i.e., a model state, to be used as a reference for development of diagnostic symptoms and a digital description of defect symptoms determined from the measurements and expert observations and opinions required for diagnosing.

中文翻译：

---

汽轮机设备诊断算法在火电厂特定条件下的适应性调整

摘要

当前，汽轮机制造商对提供汽轮机单元（STU）诊断设备系统以及自动控制系统感兴趣。这些诊断系统应适合特定的涡轮机尺寸（或修改）。诊断系统的适应性取决于用于检测缺陷的算法类型。可以选择以下类型的算法：使用计算模型确定缺陷的计算算法；基于概率方法的专家模型（使用故障症状的加权系数和贝叶斯定理的总和）；基于过程参数，设备状态参数以及在设备运行测试期间获得的其他指标之间的现象学关系而建立的数字化指标。计算算法可以集成到涡轮APCS中并在线执行（例如控制系统算法和蒸汽引入系统算法），也可以用于处理术后任务（例如辅助设备的诊断或解决技术和经济问题）。当计算模型的开发涉及很大的困难时（例如，诊断涡轮机热膨胀监测系统或振动监测系统），将采用专家算法。数字算法与任务类型，特定设备或其功能或操作模式无关。检测到的缺陷的确定程度取决于算法的类型。使诊断系统适应其他设备的需求不仅涉及开发新的诊断算法，还涉及论述的形式化。