基于融合神经网络的发动机排气温度裕度预测

Exhaust Gas Temperature Margin Prediction of Aeroengine Based on Fusion Neural Network

民用航空发动机排气温度裕度(EGTM)的变化趋势反映了其性能衰退情况,为使用EGTM的变化趋势来映射其性能衰退情况,提出了一种经验模态分解(EMD)算法与卷积长短期记忆网络(CNN-LSTM)相融合的EGTM组合预测模型。采用EMD将原始EGTM序列分解为多个固有模态分量和残差分量,并将所有分量作为模型输入,利用卷积长短期记忆网络捕获各分量非线性相关性并提取长时依赖关系,构建深度学习模型框架。为验证所提方法的有效性,采用某航空公司10 a的EGTM的实际测试数据进行实验分析,并设计了5种神经网络作为竞争模型进行对比研究。实验结果表明:相比于竞争模型,所提出的EMD-CNN-LSTM组合预测模型可使平均绝对误差和均方根误差降低37.82%和33.01%,拟合优度提高1.02%,此外,当EGTM处于敏感区域时,所提出模型的EGTM单点预测精度显著高于其他竞争模型。结果表明,该融合神经网络模型在民航发动机EGTM预测中具有较好的准确性和稳定性。

The change trend of exhaust gas temperature margin(EGTM)of civil aviation engine reflects its performance degradation.In order to employ the change trend of EGTM to map the performance degradation of the engine,an deep integrated neural network prediction method based on empirical mode decomposition(EMD)algorithm and convolution short-term memory network(CNN-LSTM)is proposed in this study.EMD was used to decompose the original EGTM sequence into multiple natural mode components and residual components,and all the resulting components were used as model input.The convolution long-term and short-term memory network was used to capture the nonlinear correlation of each component and extract the long-term dependence to construct the framework of deep learning model.In order to verify the effectiveness of the proposed method,the actual test data of EGTM of an airline for 10 years were used for experimental analysis,and 5 kinds of neural networks were designed as competitive models for comparative study.The experimental results show that the proposed EMD-CNN-LSTM fusion neural network model can reduce the mean absolute error and the root mean square error by 37.82%and 33.01%,and increase the goodness for fitting by 1.02%,compared with competitive models.Furthermore,when EGTM is in the sensitive area,EGTM single-point prediction accuracy of the proposed model is significantly higher than that of other competitive models.Therefore,the proposed fusion neural network model has good accuracy and stability in EGTM prediction of civil aviation engine.