基于EMD和KNN的发动机辐射噪声预测研究

Prediction of engine radiated noise based on EMD and KNN

针对基于发动机表面结构单通道振动的辐射噪声预测问题,提出了一种结合经验模态分解(Empirical Mode Decomposition,EMD)和KNN(K-Nearest neighbor)的预测算法,通过EMD将单一振动时域信号分解为多个本征模态函数(Intrinic Mode Function,IMF)信号,并将每个IMF信号作为振动数据集的特征,最后以新的振动数据集为输入建立辐射噪声预测模型。试验结果表明,基于该算法建立的预测模型可解释方差分数为0.97,有着较小的预测误差。

Aiming at the problem of radiated noise prediction based on single-channel vibration of engine surface structure,this paper proposes a prediction algorithm combining Empirical Mode Decomposition(EMD)and K-Nearest neighbor(KNN).EMD decomposes a single vibration time domain signal into multiple Intrinic Mode Function(IMF)signals,takes each IMF signal as a feature of the vibration data set,and finally establishes a radiated noise prediction model with the new vibration data set as input.The experimental results show that the explainable variance score of the prediction model based on the algorithm is 0.97,and the small prediction error is small.