基于残差自注意力机制的航空发动机RUL预测

RUL prediction of aero-engine based on residual self-attention mechanism

针对传统神经网络在多维数据高分辨率特征识别和高精度信号提取方面的缺陷,开展基于残差自注意力机制的剩余使用寿命(RUL)预测算法研究。比较分析卷积神经网络(CNN)和长短期记忆神经网络(LSTM)的结构特性,揭示二者在长序列信息特征关联能力和局部特征提取能力上的局限性。研究自注意力机制,引入双层残差网络抑制误差函数反向传播中扩散性,进而构建了一种卷积记忆残差自注意力机制的深度学习方法。基于上述方法对典型航空涡扇发动机退化实验数据集进行仿真分析,结果表明:所述方法能够有效建立监测数据与发动机健康状态之间的关系,关键评价指标——剩余使用寿命预测的均方误差为225,相比传统自注意力机制均方误差降低了17.9%,验证了所述方法的可行性和有效性。

In this study,the remaining useful life(RUL)prediction algorithm based on the residual self-at⁃tention mechanism is employed to address the shortcomings of traditional neural networks in the context of multi-dimensional data high-resolution feature recognition and high-precision signal extraction.The struc⁃tural characteristics of a convolutional neural network and long short-term memory(LSTM)neural net⁃work are compared and analyzed,whereby their limitations are revealed in terms of their long-sequence in⁃formation feature correlation and local feature extraction abilities.Furthermore,as part of our research on the self-attention mechanism,we introduce a double-layer residual network to suppress the spread of the error function back propagation,following which we construct a deep learning method for the convolution⁃al memory residual self-attention mechanism.Simulation analyses of the typical aviation aero-engine degra⁃dation experiment data set show that the proposed method can effectively establish the relationship between the monitoring data and engine health status.The obtained value of the key evaluation index,name⁃ly,the mean square error(MSE)of the RUL prediction,is 225.When compared with that of the tradi⁃tional self-attention mechanism,the MSE in this case is reduced by 17.9%,which verifies the feasibility and effectiveness of the proposed method.