弹底压力残缺信号的时频特征融合填充方法  

Fusion Filling Method of Time-Frequency Characteristics for Pressure-Deficient Signal at Bottom of Shell

作　　者：胡晋刚 原玥 赵永壮 王宇[2,3] 孙传猛 武耀艳[2,3] HU Jingang;YUAN Yue;ZHAO Yongzhuang;WANG Yu;SUN Chuanmeng;WU Yaoyan(Department of Industry and Information Technology of Shanxi Province,Taiyuan 030032,China;State Key Laboratory of Dynamic Measurement Technology,North University of China,Taiyuan 030051,China;School of Electrical and Control Engineering,North University of China,Taiyuan 030051,China)

机构地区：[1]山西省工业和信息化厅,山西太原030032 [2]中北大学省部共建动态测试技术国家重点实验室,山西太原030051 [3]中北大学电气与控制工程学院,山西太原030051

出　　处：《测试技术学报》2025年第2期180-189,共10页Journal of Test and Measurement Technology

基　　金：省部共建动态测试技术国家重点实验室基金资助项目(2023-SYSJJ-08);山西省基础研究计划资助项目(202203021212129,202203021221106)。

摘　　要：针对火炮测试中因极端环境导致的弹底压力信号残缺问题,提出基于长短时记忆网络(LSTM)与生成对抗插补网络(GAIN)的时频特征融合填充方法以提高信号填充的准确性。运用GAIN网络的对抗训练原理,深入学习信号内部的复杂规律和潜在分布特征,确保填充过程中保持信号全局结构与局部特征的一致性;采用时频特征融合策略,通过串并行双分支结构提取并融合弹底压力信号的时域与频域特征,从而全面捕捉信号的关键特征信息;引入具有时序处理能力的LSTM网络,学习并捕捉信号中的时序模式和长期依赖关系,确保填充信号在时序上的完整性和连贯性。试验结果表明:重构后的信号与完整信号高度相似,15 dB和30 dB信噪比情况下拟合优度达到0.9736和0.9968,实现了对弹底压力信号的精准填充。Missing projectile base pressure signals often occur due to extreme environments during artillery testing.To address this issue,a time-frequency feature fusion imputation method is proposes based on LSTM and GAIN to enhances the accuracy of signal reconstruction.The adversarial training principle of the GAIN network is utilized to learn the complex internal patterns and potential distributions of the signal and to ensure consistency between the global structure and local features during the imputation process.A time-frequency feature fusion strategy and a dual-branch parallel and serial structure are adopted to extract and integrate both time-domain and frequency-domain features of the base pressure signal.As a result,the critical signal information of the signal is comprehensively captured.LSTM networks with sequential processing capability is incorporated to learn and capture temporal patterns and long-term dependencies within the signal,as well as ensure the temporal completeness and coherence of the reconstructed signal.Experimental results show that the reconstructed signals are highly similar to the complete signals.The goodness-of-fit reach 0.9736 under a 15 dB signal-to-noise ratio(SNR)and 0.9968 under a 30 dB SNR,respectively.

关 键 词：弹底压力 残缺信号填充 时频特征融合 长短时记忆网络 生成对抗插补网络 

分 类 号：TP206.3[自动化与计算机技术—检测技术与自动化装置]