《基于非线性时间序列分析方法的汽轮发电机组流体激励故障辨识研究》由会员分享，可在线阅读，更多相关《基于非线性时间序列分析方法的汽轮发电机组流体激励故障辨识研究（114页珍藏版）》请在金锄头文库上搜索。
1、上海交通大学 博士学位论文 基于非线性时间序列分析方法的汽轮发电机组流体激励故障辨 识研究 姓名：姚宝恒 申请学位级别：博士 专业：机械电子工程 指导教师：陈兆能;佟德纯 20040601 摘 要 i 基于非线性时间序列分析方法的汽轮发电机组 流体激励故障辨识研究 摘 要 汽轮机组的流体激励行为主要发生在高中压缸由轴封汽 封系统和转子偏心引起的轴叶顶不对称力和气流涡动等因素所 且与轴系支承及轴系自身的动力学特性密切有关流体激励行为是 由流体和固体相互耦合并受各种因素综合作用而产生的复杂力学 现象目前的文献中对流体激励行为的认识并不充分随着机组的 容量和初参数的提高流体激励问题变得更加突出已经严
2、重的威 胁发电设备的发展安全和经济运行如上海吴泾第二热电厂 600MW 机组就发生过严重的流体激励故障因此迫切需要对机 组流体激励行为进行深入研究流体激励行为的深入研究具有重要 的理论与工程价值 本文作为上海汽轮机厂国家 863 重点攻关项目大功率汽轮发电 机组防气流激振研究2002AA526013-8及子课题超超临界汽轮机 关键技术研究的前期工作在大量现场实验的基础上从非线性时 间序列分析研究入手解决转子系统流体激励行为中确定性非线性 成分故障特征提取与辨识可预测性以及故障的早期检测等基本 问题为全面的认识机组流体激励故障机理并彻底解决这一问题奠 定基础本文的主要工作如下 1.通过大量的现场
3、实验研究了阀门开度和负荷变化对机组流 体激励行为的影响归纳出密封转子系统发生流体激励故障时的特 征规律及相应的辨识方法转子系统发生流体激励故障时频谱中 除了转频成分外在固有频率分量附近出现分量群频谱成分比较 丰富这些特征对流体激励故障的检测与诊断具有重要的参考价值 和指导意义 2.将非线性非稳态油膜力模型与密封流体 Muszynska 模型结合 起来建立了非线性非稳态油膜支承下的密封转子系统流体激励模 型通过改变密封间隙进行理论分析和数值仿真表明密封转子系 统发生流体激励故障时产生了丰富的振动现象具有明显的非线性 上海交通大学博士学位论文 ii 特征从频谱特征分析流体激励行为数值仿真结果与机组
4、出现流 体激励故障的实测数据相吻合 3.从非线性时间序列分析的基本问题 时间序列非线性检验 出发针对非线性直接检验方法对噪声和数据长度鲁棒性差的缺 点引入基于 VWK 的非线性间接检验方法通过对流体激励故障 信号进行基于 VWK 的非线性间接检验发现流体激励故障信号具 有确定性的非线性行为 4.对 Takens 嵌入理论中的延迟时间参数选择问题进行了深入研 究针对奇异值比率 SVF 方法的不足提出了一种新的延迟时间选 择准则 基于奇异谱熵的延迟时间选择准则通过对周期信号 Rssler 系统和 Lorenz 系统的数值仿真结果证明了该方法的有效性 并分析了奇异谱熵阶次和噪声对延迟时间选择结果的影
5、响 5.为了有效的辨识转子系统流体激励行为并与其它故障进行区 分应用基于混沌理论的辨识方法对流体激励行为进行研究应用 GP 算法研究了不平衡油膜振荡流体激励碰磨四种故障的 关联维数结果发现流体激励故障与其它几种故障明显不同关联 积分局部斜率曲线几乎未能得到收敛平台结合流体激励 VWK 非 线性检验间接方法得出流体激励故障可能产生于高维混沌系统 具有高维混沌的特征的结论为进一步对流体激励故障进行深入研 究奠定了动力学基础 6.引入了基于 p 近邻点的相轨迹非线性预测方法并研究了参 数p对预测误差的影响提出以 p 邻域的方向误差函数作为选择最 佳p值的方法应用基于 p 邻域的相轨迹预测方法对机组流
6、体激励 行为的可预测性进行了研究发现机组流体激励故障具有短期的可 预测性具有确定性的非线性行为与随机信号的不可预测性具有 本质的区别 7.对基于 Duffing 振子的微弱特征信号检测原理中相变的量化 指标问题进行研究提出将 Kullback 信息散度和连续多分辨率熵两 个指标作为振子相变过程的量化测度并对该方法在流体激励故障早 期检测中的可行性问题进行研究发现工程中现场测试得到的流体 激励故障信号具有较强背景噪声非平稳性等特点基于混沌振子 的微弱特征信号检测原理对流体激励故障信号并不敏感 关键词汽轮发电机组流体激励故障非线性非稳态油膜力模型 密封流体 Muszynska 模型Volterra
7、-Wiener-Korenberg 检验相空间 摘 要 iii 重构奇异谱熵关联维数 p 邻域相轨迹预测Duffing 振子 相变量化测度微弱特征信号检测 A Dissertation for Ph.D of Shanhai Jiao Tong University iv RESEARCH ON FAULT RECONGNITION OF FLUID STIMULATION OF STEAM TURBINE GENERATOR BASED ON NONLINEAR TIME SERIES ANALYSIS ABSTRACT Fluid stimulation behavior predomin
8、antly occurs on high and medium pressure antrum of steam turbine generator, which is induced by seal system, unsymmetrical force of shaft and lamina and airflow turbulence and correlates with shafting supported and dynamical character. As complicated mechanical phenomenon, Its influenced by various
9、factors including coupled fluid and solid and disagreement occurs in present references. Fluid stimulation comes into prominence with the increase of power capacity and initial parameters of steam turbine generator, which seriously threatens the development, safety and economical operation of electr
10、ical equipmente.g. the one happens in 600MW steam turbine generator set in Shanghai Wujing Second Power Plant . Therefore, further research on fluid stimulation behavior is urgent, which is valuable for theoretical research and engineering application. As prophase of National 863 pivotal project of
11、Research on Fluid Stimulation Behavior prevention and the Key Technique of Super Supercritical of Steam Turbine Generator (2002AA526013-8), nonlinear time series analysis methods are applied to solve deterministic nonlinear components, character extraction and recognition, predictability and incepti
12、ve detection of fluid stimulation based on substantive experiments of fieldwork. The research helps to further identify fault mechanism and settle the hidden trouble. The contents are outlined as follows: 1. The characters and identification methods are summarized when fluid stimulation fault occurs
13、 through plentiful experimental datum of fieldwork where the influences of valve opening and load variation are involved, whose components of frequency spectrum are abundant including the clusters around inherent frequency besides rotary frequency. The result is valuab1e and significant to detect an
14、d diagnose fluid stimulation fault. ABSTRACT v 2. Muszynska model of seal fluid is combined with the one of nonlinear and non-stationary oil film force and fluid stimulation mode is established with nonlinear and non-stationary oil film supported. Abundant phenomena with characteristic nonlinear beh
15、avior happen via numerical simulation and theoretical analysis with the variation of seal eccentricity, whose character of frequency spectrum acts in accordance with experimental datum in qualitative analysis. 3. The basic problem of nonlinear time series analysis, detecting nonlinearity in time ser
16、ies, is studied. Indirect method based on Volterra- Wiener-Korenberg of nonlinearity test is introduced, which overcomes the disadvantage of direct method due to ill robust to noise and data length in time series. The VWK test result of fluid stimulation fault demonstrates deterministic nonlinear behavior involved in time series. 4. The choice of time delay in Takens embedding theorem is deeply studied. A new criterion for the choice of time delay based on singular v
