摘要
针对风机齿轮箱实际工况复杂多变及含有强噪声,传统故障诊断方法对风机齿轮箱故障诊断识别准确率较低的问题,文章提出了MTF-Swin Transformer风机齿轮箱故障诊断模型。首先,采用马尔科夫变迁场(MTF)图形编码方法将原始一维振动时序信号转化为具有关联时间信息的二维特征图谱;然后,将特征图谱作为Swin Transformer模型的输入,基于自注意力机制进行自动特征提取;最后,实现对不同故障类型的分类。仿真结果表明,该方法对齿轮箱故障诊断准确率达到了99.48%,证明了该方法的有效性和优越性。
In response to the challenge posed by the limited accuracy of traditional fault diagnosis methods in wind turbine gearbox applications due to the complex and variable operational conditions and the presence of significant noise,the MTF-Swin Transformer wind turbine gearbox fault diagnosis model is proposed.Initially,the one-dimensional vibration time series signal is transformed into a two-dimensional feature map with correlated temporal information using the Markov Transition Field(MTF)graph encoding method.Subsequently,this feature map is employed as the input for the Swin Transformer model,which utilizes a self-attention mechanism for automatic feature extraction.This process culminates in the classification of various fault types.The results demonstrate a fault diagnosis accuracy of 99.48%,affirming the effectiveness and superiority of the proposed method.
作者
张彬桥
雷钧
万刚
Zhang Binqiao;Lei Jun;Wan Gang(Electric and New Energy Faculty,China Three Gorges University,Yichang 443002,China;Hubei Key Laboratory of Operation and Control of Cascade Hydropower Station,Yichang 443002,China;China Yangtze Power Co.,Ltd.,Yichang 443002,China)
出处
《可再生能源》
CAS
CSCD
北大核心
2024年第5期627-633,共7页
Renewable Energy Resources
基金
国家自然科学基金面上项目(52077120)。
