基于序列数据异常趋势识别的故障诊断方法

Fault Diagnosis Method Based on Abnormal Trend Identification of Sequential Data

在专家诊断经验中经常会采用序列数据的变化趋势作为诊断依据。由于重大设备诊断缺乏故障样本,多以专家经验为依据,使一般的定性趋势分析方法不易在智能诊断系统中直接应用。因此,提出了一种融合专家经验的序列数据趋势识别方法。该方法基于专家对趋势特征的描述,以模糊矢量形式描述序列数据的变化趋势。然后通过趋势识别决策树,实时判断数据趋势类型。将该方法应用于汽轮机故障案例中,验证了该方法提取的趋势特征可有效提高汽轮机故障诊断模型的准确度。

The variation trend of sequential data is often used as diagnosis basis in experts’ experience.Because of lacking fault samples,fault diagnosis of key equipment usually relays on experts’ experience.Common qualitative trend analysis methods are difficult to be used in intelligent diagnosis.Therefore,a trend feature identification method based on fusing experts’ knowledge is proposed.Based on experts’ description of trend feature,the trend of sequential is described by fuzzy vectors.Then,though decision tree of trend feature identification,the type of trend is identified timely.The method is used in the diagnosis of a real turbine fault case.The result verifies that the identified trend feature can effectively improves the accuracy of diagnosis model for steam turbine.