复杂远洋环境下船载天线半实物精确建模与指向精度非线性退化趋势预测

Aiming at the high reliability, high security, and easy maintenance requirements of shipborne antenna under the whole process of the task execution, this project carries out research on accurate semi-physical modeling and nonlinear degradation trend prediction of pointing accuracy for shipborne antenna under complex ocean environment. However, the evaluation and prediction of pointing accuracy for shipborne antenna under complex ocean environment is still a challenge, because of the nonlinearity of degradation process, the difficulty of acquiring the multi-sample and life-cycle degradation data, the multisource unlabeled non-stationary monitoring data. According to the above mentioned problems, this research project aims to seek improvements in the following aspects. Firstly, the dual coordination model of abrasion and the multifactor competition model of crack propagation under the complex ocean environment are established to reveal the effect of wear gap and changes in crack respiration stiffness on antenna pointing accuracy. Next, a retaining dimension reduction method of non-matching sub-structure for antenna run-time high-precision finite element model is proposed to establish a semi-physical high-precision model of antenna, which can generate multi-sample and life-cycle degradation data with reflecting the nonlinear multi-state degradation process of pointing accuracy. Then, an unsupervised feature extraction method based on sparse stacked denoising autoencoder method is proposed to realize the unsupervised extraction of degradation features from multisource unlabeled non-stationary monitoring data that represent the monotonous change of pointing accuracy under ocean environment. Finally, a nonlinear multi-state degradation model of pointing precision based on hyper-parameter optimized nonhomogeneous hidden semi-markov model is established using multi-sample and life-cycle degradation data in order to achieve the quantitative assessment and accurate prediction of the pointing accuracy degradation, and the result can provide a scientific foundation for the short-term parameter adjustment and mid-term mission planning of the shipborne antenna. The research will greatly enhance the ability of warship to carry out their missions, which has important theoretical and practical value.

本项目瞄准船载天线在任务执行全过程中高可靠、高安全的迫切需求，开展复杂远洋环境下天线半实物精确建模与指向精度非线性退化趋势预测方法研究。针对远洋环境下天线指向精度退化过程非线性且多样本全寿命退化数据获取难，实测数据具有多源无标签非平稳等特点，结合理论模型基于试验研究建立远洋环境下二元协同磨损模型和多因素裂纹扩展竞争模型，揭示磨损间隙与裂纹呼吸刚度变化对指向精度的影响规律；提出运行态高精度模型的非匹配子结构保真降维方法，实现半实物高精度模型的高效求解并产生反映指向精度变化规律的多样本全寿命退化数据；提出基于稀疏栈式降噪自编码器的特征非监督提取方法，实现多源无标签非平稳数据下的退化特征提取；利用全寿命数据建立基于超参数优化NHSMM的非线性多状态退化模型，并利用指向精度退化特征提取结果实现定量评估与精确预测，为天线任务规划提供科学依据。研究将有力提高舰船任务执行能力，具有重要理论意义与工程应用

船载天线作为舰船的关键通信装置，在国防安全保障中发挥着不可替代的作用。本项目瞄准船载天线在任务执行全过程中高可靠、高安全的迫切需求，开展了复杂远洋环境下船载天线指向精度非线性退化趋势预测方法研究。针对远洋环境下天线指向精度退化过程非线性且多样本全寿命退化数据获取难，实测数据具有多源无标签非平稳等特点，开展了以下研究内容：1）基于半实物高精度模型的船载天线全寿命退化数据获取方法。结合理论模型基于试验研究并建立了考虑远洋环境的二元协同锈蚀磨损模型；建立了齿轮副非线性啮合动力学模型，通过动力学仿真及实验揭示了齿根裂纹故障的振动响应机理；完成了船载天线全尺寸多功能多故障模拟试验台改造并开展了多种故障类型及不同程度故障模拟实验，建立了多种故障实验数据集；2）多源无标签非平稳数据下指向精度退化特征非监督提取方法。提出了基于数据增强的少标签样本故障特征提取方法，实现了小样本和无故障样本下天线设备故障特征的自适应提取；提出了基于深度降噪自编码器的天线运行退化特征提取方法，实现了多源无标签数据下退化特征的非监督提取；3）基于超参数优化模型的非线性退化状态评估与预测方法。提出了剪枝和多目标增强下基于可微神经网络结构搜索的模型超参数优化方法，实现了模型超参数的自适应搜索并构建最优诊断模型；提出了机械设备非线性退化状态评估与预测方法，实现了机械设备非线性退化状态的定量评估与剩余寿命预测。基于项目研究成果，共发表论文29篇，其中SCI检索论文25篇；申请专利9项，已授权3项；本项目研究对提升船载天线的运行安全保障能力、提高舰船的任务执行能力，具有重要的学术研究意义和工程应用价值。

内容获取失败，请点击重试