【摘要】 以神经网络为代表的人工智能技术,为通过状态监测信息全面表征航空发动机状态提供了可能。如何获取足够的监测样本成为神经网络成功应用的关键。生成式对抗网络(Generative adversarial networks,GAN)的应用可实现在已有状态监测信息的基础上扩大样本量。结合经典的误差反向传播BP(back propagation)神经网络预测方法,设计一种新的具有扩展训练样本能力的GANBP预测模型。以航空发动机为例,利用生成式对抗网络生成航空发动机状态监测样本,通过算例来说明本方法的可行性。实验结果表明在大量的网络迭代训练后,GAN能够提取监测样本的特征信息,利用BP算法对航空发动机性能退化预测并与其它预测方法相比较,证明本文构建的GANBP模型能够有效解决因航空发动机状态监测样本量过小而导致性能衰退预测不准确的问题。更多还原

【Abstract】 Artificial intelligence technology represented by neural networks,provides the possibility to fully characterize the condition of aero-engines through condition monitoring information.How to obtain enough monitoring samples becomes the key to the successful application of neural networks.The application of Generative adversarial networks(GAN) can expand the sample size based on existing condition monitoring information.Combined with the classical back propagation(BP) nets prediction method,a new GANBP prediction method with extended training samples is designed.Taking the aero-engine as the research object,the monitoring samples are generated by GAN,and the feasibility of the method is illustrated by an example.The experimental results show that after a large number of network iterative training,GAN can extract the characteristic information of the monitoring samples,and use BP algorithm to predict the performance degradation of aero-engine.We compare GANBP method with other prediction methods.It proves that the proposed GANBP method can effectively solve the problem that the sample size is too small and the performance degradation prediction is inaccurate.更多还原