基于振动--红外异构数据的航天器损伤证据深度学习抗冲突融合诊断

Anti-conflict fusion diagnosis of spacecraft damage based on evidential deep learning vibration-infrared heterogeneous data

摘要

为突破单一监测手段在航天复杂环境下信息感知的局限性，利用异构数据实现优势互补成为必然趋势。针对航天器损伤诊断中振动−红外异构数据面临的强噪声干扰与源间信息冲突问题，传统融合方法因缺乏不确定性度量易引发“高置信度误判”风险。对此，提出一种基于证据深度学习（EDL）的抗冲突融合诊断框架。首先，构建双流视觉 Transformer （ViT）骨干网络，分别提取红外图像的损伤纹理与振动时频图的瞬态能量特征，并将网络输出映射为狄利克雷分布以量化不确定性；其次，设计非线性信度门控机制，根据不确定性动态修正基本概率分配，自适应阻断不可信信源干扰；最后，通过加权证据融合策略实现决策级融合。实验结果表明，该方法在振动−红外异构数据集上的总体诊断准确率达到 98.26%，实现了对单信源性能的全面超越；特别针对单一振动模态易漏检的微小裂纹损伤，融合识别率从 82.1% 提升至 92.9%，有效解决了异构数据融合中的合成悖论难题。

Abstract

The utilization of heterogeneous data is essential for spacecraft damage diagnosis, yet traditional fusion methods suffer from high-confidence misclassification due to strong noise and inter-source conflicts. To address these issues, an anti-conflict fusion framework based on evidential deep learning (EDL) is proposed. A two-stream vision transformer (ViT) backbone is employed to extract damage textures from infrared images and transient energy features from vibration time-frequency diagrams, whose outputs are mapped to a Dirichlet distribution for uncertainty quantification. Furthermore, a nonlinear belief gating mechanism is designed to dynamically adjust probability assignments based on uncertainty, adaptively suppressing interference from unreliable sources. Experimental results demonstrate that the proposed method achieves an overall diagnosis accuracy of 98.26% on the heterogeneous dataset, surpassing the performance of single-source methods. Specifically, the recognition rate for micro-cracks, which tend to be missed with a single vibration modality, is improved from 82.1% to 92.9%, effectively resolving the synthesis paradox in heterogeneous data fusion.

关键词

航天器损伤 / 异构数据融合 / 证据深度学习 / 不确定性量化 / 故障诊断

Key words

spacecraft damage / heterogeneous data fusion / evidential deep learning / uncertainty quantification / vibration-infrared diagnosis

参考文献

[1] Huang Xuegang, Yin Chun, Ru Hongqiang, et al. Hypervelocity impact damage behavior of B₄C/Al composite for MMOD shielding application[J].Materials & Design, 2020, 186: 108323.

[2] Huang Xuegang, Yin Chun, Huang Jie, et al. Hypervelocity impact of TiB₂-based composites as front bumpers for space shield applications[J].Materials & Design, 2016, 97: 473-482.

[3] 范志涵, 张宇, 芮小博 . 航天器舱壁结构碎片撞击声发射定位技术研究[J].仪器仪表学报, 2020, 41(1): 178-184.

[4] Fan Zhihan, Zhang Yu, Rui Xiaobo . Research on space debris impact location of spacecraft bulkhead structure based on the acoustic emission technique[J].Chinese Journal of Scientific Instrument, 2020, 41(1): 178-184. (in Chinese)

[5] 李正宇, 贾光辉, 张品亮, 等. 超高速斜撞击下航天器 Whipple 结构碎片特性研究[J].空间碎片研究, 2023, 23(2): 21-33.

[6] Li Zhengyu, Jia Guanghui, Zhang Pinliang, et al. Study on the fragment characteristics of spacecraft Whipple structure under hypervelocity oblique impact[J].Space Debris Research, 2023, 23(2): 21-33. (in Chinese)

[7] 任思远, 武强, 张品亮, 等. 活性弹丸超高速撞击蜂窝夹芯板双层结构的损伤特性[J].爆炸与冲击, 2024, 44(7): 073302.

[8] Ren Siyuan, Wu Qiang, Zhang Pinliang, et al. A study of damage characteristics caused by hypervelocity impact of reactive projectile on the honeycomb sandwich panel double-layer structure[J].Explosion and Shock Waves, 2024, 44(7): 073302. (in Chinese)

[9] 武江凯, 迟润强, 韩增尧, 等. 载人航天器密封舱结构超高速撞击易损性[J].哈尔滨工业大学学报, 2023, 55(8): 25-31.

[10] Wu Jiangkai, Chi Runqiang, Han Zengyao, et al. Vulnerability of pressurized cabin for manned spacecraft under hypervelocity impact[J].Journal of Harbin Institute of Technology, 2023, 55(8): 25-31. (in Chinese)

[11] 陈韩青, 徐志远, 屈仲毅, 等. 蜂窝夹层结构无损检测方法研究综述[J].材料导报, 2024, 38(10): 296-310.

[12] Chen Hanqing, Xu Zhiyuan, Qu Zhongyi, et al. Review of nondestructive testing methods for honeycomb sandwich structures[J].Materials Reports, 2024, 38(10): 296-310.

[13] 王磊, 孙志成, 王磊, 等. 基于 DRSN-CW 和 LSTM 的轴承故障诊断[J].电子科技大学学报, 2022, 51(6): 921-927.

[14] Wang Lei, Sun Zhicheng, Wang Lei, et al. Bearing fault diagnosis based on DRSN-CW and LSTM[J].Journal of University of Electronic Science and Technology of China, 2022, 51(6): 921-927. (in Chinese)

[15] Zhang Kuo, Huo Jianliang, Wang Shengzhe, et al. Damage quantitative assessment of spacecraft in a large-size inspection[J].Frontiers of Information Technology & Electronic Engineering, 2022, 23(4): 542-554.

[16] Yin Chun, Huang Xuegang, Tan Xutong, et al. Infrared thermographic NDT-based damage detection and analysis method for spacecraft[M]. Singapore:Springer, 2024.

[17] Yin Chun, Huang Xuegang, Liu Junyang, et al. Dynamic defect feature extraction via infrared thermography for monitoring hypervelocity impact damage from space debris[J].Thermal Science and Engineering Progress, 2025, 66: 104003.

[18] Zhang Peng, Tang Keke, Chen Guangxu, et al. Multimodal data fusion enhanced deep learning prediction of crack path segmentation in CFRP composites[J].Composites Science and Technology, 2024, 257: 110812.

[19] 寇梓良, 张西宁, 李兵, 等. 基于多模态数据融合与改进 CNN 的轴承故障诊断[J].振动、测试与诊断, 2025, 45(6): 1082-1089.

[20] Kou Ziliang, Zhang Xining, Li Bing, et al. Bearing fault diagnosis based on multi-modal data fusion and improved CNN[J].Journal of Vibration, Measurement & Diagnosis, 2025, 45(6): 1082-1089. (in Chinese)

[21] 熊鹏文, 胡慕烨, 黄雨轩, 等. 面向小样本纹理分类的多模态证据融合框架[J].仪器仪表学报, 2025, 46(6): 154-165.

[22] Xiong Pengwen, Hu Muye, Huang Yuxuan, et al. Small-sample multi-modal evidence fusion framework for texture classification[J].Chinese Journal of Scientific Instrument, 2025, 46(6): 154-165. (in Chinese)

[23] 尹刚, 朱淼, 颜玥涵, 等. DMIFD: 一种基于深度学习的多模态工业故障诊断方法[J].仪器仪表学报, 2025, 46(6): 215-227.

[24] Yin Gang, Zhu Miao, Yan Yuehan, et al. DMIFD: A deep learning-based method for multimodal industrial fault diagnosis[J].Chinese Journal of Scientific Instrument, 2025, 46(6): 215-227. (in Chinese)

[25] 吴海滨, 左云逸, 王爱丽, 等. CNN 结合 Transformer 的高光谱图像和 LiDAR 数据协同地物分类方法[J].仪器仪表学报, 2025, 46(8): 286-301.

[26] Wu Haibin, Zuo Yunyi, Wang Aili, et al. Collaborative land classification method using CNN combined with Transformer for hyperspectral images and LiDAR data[J].Chinese Journal of Scientific Instrument, 2025, 46(8): 286-301. (in Chinese)

[27] Han Kai, Wang Yunhe, Chen Hanting, et al. A survey on vision transformer[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023, 45(1): 87-110.

[28] 张桃红, 郭学强, 郑瀚, 等. Dual-path vision transformer 用于急性缺血性脑卒中辅助诊断[J].电子科技大学学报, 2024, 53(2): 307-314.

[29] Zhang Taohong, Guo Xueqiang, Zheng Han, et al. Dual-path vision transformer for auxiliary diagnosis of acute ischemic stroke[J].Journal of University of Electronic Science and Technology of China, 2024, 53(2): 307-314. (in Chinese)

[30] 乌达巴拉, 万鑫鑫 . 基于视觉 Transformer 的工业图像异常检测方法研究[J].皖西学院学报, 2025, 41(5): 62-70.

[31] Odbal, Wan Xinxin . A vision transformer-based method for image anomaly detection[J].Journal of West Anhui University, 2025, 41(5): 62-70. (in Chinese)

[32] 孙露露, 刘建平, 王健, 等. 细粒度图像分类上 Vision Transformer 的发展综述[J].计算机工程与应用, 2024, 60(10): 30-46.

[33] Sun Lulu, Liu Jianping, Wang Jian, et al. Survey of vision transformer in fine-grained image classification[J].Computer Engineering and Applications, 2024, 60(10): 30-46. (in Chinese)

[34] Zakariah M, Al-Razgan M, Alfakih T. Dual vision transformer-DSUNET with feature fusion for brain tumor segmentation[J].Heliyon, 2024, 10(18): e37804.

[35] Hou Qibin, Zhou Daquan, Feng Jiashi. Coordinate attention for efficient mobile network design[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Nashville: IEEE, 2021: 13708-13717.

[36] Yue Xiaoyu, Sun Shuyang, Kuang Zhanghui, et al. Vision Transformer with progressive sampling[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV). Montreal: IEEE, 2021: 377-386.

[37] Soleimany A P, Amini A, Goldman S, et al. Evidential deep learning for guided molecular property prediction and discovery[J].ACS Central Science, 2021, 7(8): 1356-1367.

[38] Smets P, Kennes R. The transferable belief model[J].Artificial Intelligence, 1994, 66(2): 191-234.

基金资助

国家自然科学基金(92571201)

国家自然科学基金(62273074)

四川省科技厅杰出青年基金(2024NSFJQ0015)