基于WPT和t-SNE的直升机桨叶损伤特征提取

Fault Feature Extraction Method for Helicopter Rotor Blades Based on WPT and t-SNE

旋翼桨叶的损坏可能会导致直升机坠落损毁,开展桨叶健康状态的在线监测评估对保障飞行安全至关重要。提出一种将小波包变换(WPT)与t-分布随机近邻嵌入(t-SNE)相结合的桨叶损伤识别方法。首先利用振动台模拟直升机服役时的真实振动,用传感器获取不同故障桨叶模型在振动环境下的输出响应。然后对信号进行小波包分解,提取小波包能量作为原始特征向量,接着用流形学习对特征向量进行维数约简,最后输入到K近邻分类器进行故障识别。实验结果表明:首先,在原始特征选取方面,小波包能量特征优于时域特征与小波包能量组合成的混合特征;其次,t-SNE的降维效果优于PCA、Sammon映射、LTSA、HLLE、SNE这5种方法,且不受嵌入维数的制约。研究结果证明了所提出的方法能提高桨叶损伤评估的准确性。

The damage o f rotor blades may cause air crash. It's critical to monitor and evaluate the blade health to ensure flight safety. A damage identification method is presented in this paper, which combines wavelet packet transform (WPT) with t-distributed stochastic neighbor embedding (t-SNE). First of all, the real vibration of a helicopter in service is simulated with vibration table in laboratory tests, and the output response of different fault blade models in vibration environment are obtain by sensors. Then, the signals are decomposed into wavelet packets, and the wavelet packet energy is extracted as the original feature vector. After that, the feature vector is reduced in dimension by manifold learning, and input to k-nearest neighbor classifier (KNN) finally for fault identification. The experimental results show that, firstly, in terms of the original feature selection, the wavelet packet energy features are superior to the mixed features composed of time domain features and wavelet packet energy. Secondly, the dimensionality reduction effect of t-SNE is better than that of PCA, Sammon mapping, LTSA, HLLE and SNE, and is not restricted by embedding dimension. The results prove that the proposed method can improve recognition accuracy of rotor blade damage identification.