基于小波包和并行隐马尔科夫的风力机易损部件健康状态评价

HEALTH STATE EVALUATION BASED ON WAVELET PACKET AND PCHMM FOR VULNERABLE COMPONENTS OF WIND TURBINES

考虑到风电机组运行时监测到的轴承、齿轮等易损部件的振动信号具有动态和非平稳特性,提出一种基于小波包分解和并行隐马尔科夫(parallel continuous hidden Markov model,PCHMM)的易损部件健康状态评价方法。该方法对采集的振动信号进行小波包分解,根据信号的采样频率和部件的故障特征频率选取小波包分解层数,提取各节点能量与总能量之比作为健康状态评价的特征向量,并应用并行隐马尔科夫模型建立易损部件的健康状态评价模型,为合理确定评价并行模型中各组成部分的权重,引入信息熵计算各部分权重,以模型输出的对数似然概率值作为状态评价指标。将模型用于轴承退化实验数据和现场数据的研究表明评价模型很好地反映了轴承的运行状态,评价指标在故障早期有很明显的变化,有利于及时发现易损部件的故障,降低维修成本。

Wind turbines have bearings,gears and other vulnerable components and the vibration signals have dynamic and non-stationary characteristics. A health condition evaluation based on wavelet packet decomposition and parallel continuous hidden Markov model is proposed in this paper. The collected vibration signals are decomposed by wavelet packet and the wavelet packet decomposition level is selected according to the sampling frequency of the signal and the fault characteristic frequency of the component. The ratio of energy and total energy of each node is extracted as the feature vector of health state evaluation,then PCHMM is applied for health stage evaluation model of vulnerable components. In order to reasonably determine the weight of each component in the parallel model,we introduce information entropy to calculate the weight of each part. The log likelihood probability of model output is used as the state evaluation index. The model used in bearing degradation simulation of experimental data and field data show that the evaluation model well reflects the bearing running state and evaluation indicators in the early fault has obvious changes.The proposed method can detect the fault of damageable parts and reduce the maintenance cost.