水电机组早期振动健康状态评估新方法研究

Aiming at the problem of difficulty in finding the early fault on the vibration state monitoring of the hydro-generator unit, the project studies new method for the early vibration state health assessment of the hydro-generator unit. Firstly, in order to reveal the mapping relationship between vibration feature and fault dynamic evolution process, the project uses the finite element to establish the dynamic model of rotor system, and analyze the dynamic behavior and response characteristics for different early fault types. Secondly, to make full use of the vibration state information carried by the encoder angular domain signal, an analysis method based on the encoder angular domain signal is proposed, and the instantaneous angular acceleration signal which reflects the early vibration characteristic is obtained by spectral quadratic weighting of the encoder angular position signal. Then, the vibration health state evaluation method based on the state Gini index is proposed,and the mapping relationship between the different vibration states and the Gini index is studied to determine the normal and abnormal state boundaries of the unit. Finally,when the unit vibration is detected abnormally , the dimensionless parameters and information entropy parameters of the monitoring signals are extracted by the adaptive modified ensemble empirical mode decomposition method, and the intelligent diagnosis of early faults is realized by the optimized extreme learning machine algorithm. This project uses the rotating machinery fault test-bed to verify and modify the proposed methods, and its research production, which possesses important theoretical significance and engineering application value, will not only promote the establishment of health state evaluation system of rotating machinery, but also enhance the intelligent diagnosis level for early fault.

针对水电机组振动状态监测中难于发现早期故障的难题，拟开展水电机组早期振动健康状态评估新方法研究。首先，为揭示故障特征与故障演化过程的映射关系，建立转子系统动力学模型，对动力学行为和响应特性进行分析。其次，为充分利用编码器角域信号携带的振动状态信息，提出基于编码器角域信号的分析方法，对编码器的角位移信号进行谱二次加权处理，获得反映早期振动特征的瞬时角加速度信号。然后，提出基于状态基尼指数的机组振动健康状态评估方法，研究机组不同振动状态与基尼指数的映射关系，确定机组健康状态和异常状态的分界。最后，当监测到机组振动异常时，利用自适应改进集合经验模式分解提取监测信号的无量纲参数和信息熵参数，并利用果蝇优化的极限学习机算法实现早期故障智能诊断。项目利用旋转机械故障实验台验证和修正所提方法，研究成果既能推进旋转机械振动健康状态评估体系建立，又能提高早期故障智能诊断水平，具有重要理论意义和工程应用价值。

研究了水电机组早期振动健康状态评估的新方法，围绕状态基尼指数在机组健康评估中的故障区分性、故障信号的自适应分解算法、故障信号的特征参数提取以及不同故障类型的智能诊断开展研究，利用水电机组和风电齿轮箱数据进行方法性能验证，其主要研究成果包括：根据所采集的水电机组以及风电齿轮箱等设备的故障样本数据，结合实验室所搭建的故障实验台，构建了基于状态基尼指数的机组健康状态评估方法；开展了集合经验模式分解及其改进算法的研究，并引入自适应局部迭代滤波算法，对机组故障信号进行自适应分解；通过提取故障信号的样本熵、模糊熵、排列熵等特征参数，结合极限学习机、支持向量机、灰色关联度等模式识别方法，实现了面向水电机组和风电齿轮箱的不同故障类型的智能诊断。项目资助发表相关科研论文10篇，其中SCI检索论文1篇，EI期刊论文2篇，EI会议论文3篇，中文核心期刊论文2篇；录用论文2篇；申请受理发明专利2项。通过本项目研究，丰富了旋转机械振动健康状态评估和早期故障智能诊断的理论和方法体系，为水电机组的定期维修过渡到基于状态的维修提供理论和实践依据。