| 摘要: |
| 针对综合射频系统中的故障预测与健康管理技术模型泛化能力弱、数据成本高及均衡性差的问题，提出了一种空地协同的系统设计方法。为优化泛化能力，降低数据成本并增强数据均衡性，该方法对机载射频系统、数字孪生系统及人工智能（Artificial Intelligence,AI)控制中心进行整合，实现数据共享和模型同步。进一步，通过采集不同模块通用单元的多种传感器特征以扩充数据集，并集合K-means聚类算法与生成对抗网络生成极端数据，改善了数据平衡性。最终，基于卷积神经网络实现对机载射频系统可靠性的评估，预测值与实际值之间高度拟合，均方差为0.000 2，平均绝对误差为0.008 9，决定系数达到0.945 2。本研究为综合射频系统中的故障预测与健康管理技术的发展提供了新的思路和方法。 |
| 关键词: 机线综合射频系统 故障预测与健康管理(PHM) 卷积神经网络(CNN) 反向传播神经网络(BPNN) 生成对抗网络(GAN) |
| DOI：10.20079/j.issn.1001-893x.250216001 |
|
| 基金项目: |
|
| A Health Status Assessment Method for Airborne Integrated RF System Based on CNN |
| DING Chencong |
| (Naval Research Institute,Shanghai 200436,China) |
| Abstract: |
| For the problems of insufficient generalization ability of prognostics and health management(PHM) models in the integrated radio frequency system(IRFS) field,high data acquisition costs,and poor data balance,a design method for an air-ground cooperative system for PHM is proposed. This system achieves data sharing and model synchronization through the airborne IRFS in the air segment,the digital twin system and artificial intelligence(AI) control center on the ground segment,thereby enhancing the generalization ability of PHM models,reducing the cost of data acquisition,and improving data balance. Furthermore,the dataset of PHM is augmented by collecting various sensor characteristics from common units of different modules,and extreme data is generated using the K-means clustering algorithm and the generative adversarial network(GAN) to improve data balance. Finally,the reliability of the airborne IRFS is evaluated based on a convolutional neural network(CNN),resulting in a high fit between predicted and actual values,with mean squared error(MSE) of 0.000 2,mean absolute error of 0.008 9,and R2 Score of 0.945 2. This study provides new ideas and methods for the development of PHM technology in IRFS. |
| Key words: airborne integrated RF system prognostics and health management(PHM) convolutional neural network(CNN) back propagation neural network(BPNN) generative adversarial network(GAN) |
