基于数据驱动的大型转子机械健康状态评估与维护策略研究

The unreasonable maintenance method for the large mechanical rotor (steam turbine,compressor, generator, fan, etc) often causes huge resource waste and great losses in the process of enterprise production. The theory research with simulation and experiment is applied in this project. Firstly, the vibration characteristic law caused by mechanical performance degradation of large rotor machines are studied. The multivariate statistical characteristics of vibration signal are extracted, selected, and optimized. The Hidden semi-Markov model(HSMM) is adopted to describe the relationship between vibration signal, degradation state, and trend of state. Secondly, The HSMM degradation state recognition and prediction models based on data-driven are set up to recognize the performance degradation state and predict the developing trend, respectively. Combined with the historical state data, a new method for quantitative evaluation the health state of lagre rotor machine is developed by using Fuzzy theory and synthetic evaluating method. Thirdly, we indicate the influence law of the health state adopted different maintenance strategies, and define the effect impact factor to reflect the impact degree. Then, the criterias of maintenance threshold, period and numbers are established. The maintenance strategy model based on the health state evaluation value on the condition of multi-constraints (the economy, availability, and risk) is builded to determine the best maintenance time for different maintenance ways. This method may break the traditional planned maintenance routine maintenance mode, implement scientific and reasonable maintenance strategy, and form the active preventative maintenance theory and technology system to ensure the safety, high efficiency, and health operation of machines. This project will be good to enrich the equipment maintenance theory and promote the advanced maintenance technology development of our country.

针对大型转子机械(汽轮机、压缩机、发电机、风机等)不合理维护给企业生产造成巨大浪费和损失的问题。项目拟通过理论与仿真试验相结合研究方式，掌握大型转子机械性能退化导致的振动特性规律，利用和挖掘振动信号多元统计特征信息，采用隐式半马尔科夫模型(HSMM)定量描述振动信号与性能退化状态及发展趋势的关系，构建基于数据驱动的HSMM性能退化状态识别和预测模型，结合历史状态数据，采用模糊理论和综合评价方法研究健康状态量化评估新方法。揭示不同维护策略对健康状态的影响规律，定义效果影响因子刻画其影响程度，研究维护阈值、周期和次数准则，构建经济性、可用度和风险性多目标约束下基于健康状态值的维护策略模型，确定不同维护方式下最佳维护时机，突破传统计划维护定期维修模式，实施科学合理的维护策略，形成保证机械安全、高效、健康运行的主动预防性维护理论和技术体系。项目有利于丰富设备维护理论，促进我国先进维护技术的发展。

项目针对汽轮机、压缩机、电机、风机等转子机械振动特性、健康状态评估与维护策略进行了系统研究，以上汽西门子N+1支撑百万超超临界汽轮机组N+1支撑汽轮机组、涡轮增压器为对象，在转子机械动力学建模、轴系不平衡振动特性、动平衡健康维护技术、健康状态评估及维护策略与技术等方面取得了一些重要进展。(1) 建立了多支承-联轴器-转子系统动力学方程，提出轴系动力学建模基本原则和方法，构建了N+1支撑轴系动力学有限元模型，分析出临界转速、振型和不平衡响应特性，设计了模拟超超临界百万汽轮机组轴系的四跨五支承转子实验台。(2) 开展轴系稳态不平衡响应分析，总结了不同位置、大小和相位差下轴系不平衡振动响应变化规律，采用转子或轴承座处的振动响应信号特征，实现对转子机械不平衡故障状态的描述。(3) 针对转子动平衡故障，结合模态振型平衡法和影响系数平衡法，研究了基于有限元模型分析的转子动平衡影响系数求解法，提出旋转机械无试重虚拟动平衡健康维护方法，开展超超临界汽轮机低压转子高速动平衡、四跨五支承轴系整机动平衡实验，降幅效果达到50%以上，大大减少设备启停机次数，缩短维护时间。(4) 开展含浮环轴承的涡轮增压器动力学有限元建模方法与振动特性研究，分析涡轮增压器转子轴承系统动力学行为，设计了涡轮增压器动力学特性实验台，研究涡轮和叶轮处不平衡振动响应规律，为涡轮增压器动力学设计和运行维护提供参考。(5) 研究了基于Fuzzy-AHP的机械设备多特征参数健康状态综合评价方法，提出用健康值来定量描述机械设备运行状态程度方法。以水泥行业中核心设备回转窑状态评估为例进行验证。采用故障树对大型多支承回转机械健康运行轴线检测与优化系统进行了可靠性与人因影响因素分析，以完善设备运行状态分析与评估方法。本项目研究结论对保证转子机械安全、高效、健康运行，实施科学维护，节约维护资源，创造经济效益具有重要意义。项目组已在国内外期刊发表学术论文11 篇，其中SCI 源刊2 篇，EI 收录4篇次，申请国家发明专利7项、获授权实用新型专利1 项，获湖南省自然科学三等奖等奖励7项，进行国内外学术交流11人次，培养博士研究生1名，硕士研究生6名，完成了项目计划工作和任务。