基于深度迁移学习的航天器故障诊断

随着航天科技的发展，智能故障诊断技术是确保航天器控制系统安全、自主运行的关键技术之一.由于在轨航天器遥测数据样本少、噪声高、未标记，因此缺乏自适应能力、学习能力的传统故障诊断方法难以准确诊断在轨航天器故障.本文针对上述问题提出一种基于深度迁移学习的航天器故障诊断方法，为在轨航天器实时故障诊断提供了可行方法.首先，对航天器运行数据进行预处理，将多维时域信号转换为二维图像信号；其次，搭建基于残差网络的故障诊断深度学习框架，并利用地面测试数据与其他航天器在轨运行数据对网络进行预训练；进而，为了实现当前在轨航天器实时故障诊断，本文采用迁移学习自适应方法，设计网络联合分布自适应代价函数，对故障诊断模型进行参数重调，使模型适应当前在轨航天器故障诊断任务.仿真结果表明，所提出的基于深度迁移学习的故障诊断方法可以快速准确的诊断出航天器故障.

Deep Transfer Learning-Based Fault Diagnosis of Spacecraft Attitude System

With the development of aerospace science and technology, intelligent fault diagnosis technology is one of the key technologies to ensure safe and autonomous operation of spacecraft control system. Due to the small number of unlabeled telemetry data samples with high noise, it is difficult to diagnose fault signals accurately of the spacecraft in orbit by traditional fault diagnosis methods. A deep transfer learning based fault diagnosis method is proposed to realize real time fault diagnosis of spacecraft in orbit. First, one dimensional time domain signals are converted into two dimensional image signals to realize the preprocessing of spacecraft operation dataset. Secondly, a residual network based deep learning fault diagnosis framework is built and pre trained via ground trained dataset and on orbit operation data of other spacecraft. Then, in order to realize real time fault diagnosis of current spacecraft in orbit, parameters of the fault diagnosis model are readjusted to adapt the model to the current spacecraft fault diagnosis. Simulation results show that the proposed deep transfer learning based fault diagnosis method can diagnose spacecraft fault signal quickly and accurately.