为解决传统故障诊断方法效率低的问题，以某步兵战车柴油机为研究对象，提出了一种补充的集合经验模态分解与奇异值分解相结合提取信号的特征，使用模拟退火算法优化循环神经网络对曲轴轴承磨损程度进行预测的方法。采用补充的集合经验模态分解方法对振动信号进行分解，用奇异值分解方法进行特征提取，利用特征对模拟退火算法优化的循环神经网络进行训练及预测。对所提出的算法进行试验分析，结果显示预测准确率达到97.48%，比普通的循环神经网络系统预测的准确率提高了5%以上。

In order to solve the low efficiency of traditional fault diagnosis method, the singular value decomposition and complementary ensemble empirical mode decomposition were combined to extract the characteristics of signal and the crankshaft bearing wear was predicted by optimizing the recurrent neural network with the simulated annealing algorithm based on diesel engine of infantry fighting vehicle. The complementary ensemble empirical mode decomposition was used to decompose the vibration signal, the singular value decomposition was used to extract the features, the simulated annealing algorithm was used to optimize the recurrent neural network, and the training and prediction were further conduced. Finally, the experimental analysis of the proposed algorithm shows that the prediction accuracy is 97.48%, which is more than 5% higher than that of the ordinary recurrent neural network system.