Currently, vibration-based unsupervised structural health monitoring (SHM) is attracting increasing attention but remain limited in widespread application by their dependence on artificial excitation, particularly for inservice aircraft where such excitation is impractical. To address this challenge and achieve comprehensive damage detection, localization, and level assessment for aircraft structures, this paper proposes a system-level unsupervised framework that leverages atmospheric turbulence as a natural excitation source. The core advantage lies in the transmissibility function (TF) feature reconstruction technology based on bidirectional long short-term memory autoencoder (BiLSTM-AE), combined with a decision framework integrating the energy integral of reconstruction error (EIRE) and the Jensen-Shannon (J-S) divergence. The BiLSTM-AE is trained exclusively on intact-state data to reconstruct TF from monitored regions, establishing a baseline model of local structural integrity. The EIRE between reconstructed and actual TFs serves as a robust indicator for damage detection and localization. For damage quantification, we introduce a novel assessment metric based on the J-S divergence between generalized extreme value distributions (GEVD) fitted to pre-and post-damage EIRE samples. Simulations on a generic transport aircraft (GTA) model demonstrate that the method could effectively detect, localize, and assess minor damage exceeding 2% stiffness degradation at critical wing root/midspan regions under the simulated flight conditions. Under 15 dB noise conditions, the framework maintains robust localization accuracy. Furthermore, it exhibits exceptional generalization capabilities across varying gust velocities and cruising altitude changes of +/- 230 m. A comparative study against a latest unsupervised approach further assesses the performance and advantages in detecting minor damage in aircraft structures. These results establish the proposed methodology as a promising solution for in-flight SHM and predictive maintenance applications, offering practical potential for next-generation aerospace SHM systems.

The wing of a butterfly consists of partially overlapping forewing and hindwing, and forewing sweeping can dynamically change the shape of the whole wing. In this work, the effect of forewing sweeping on aerodynamic performance of a butterfly like model is studied using a solver based on immersed boundary method and adaptive mesh. For aerodynamic performance, adding a “forward-backward-forward” sweeping motion to the forewing makes it more suitable for fast cruising flight, and compared to the situation without forewing sweeping, the drag is reduced by 46% and the lift to drag ratio is increased by 45%. On the contrary, adding a “backward-forward-backward” sweeping motion to the forewing increases lift and makes it more suitable for climb flight. For downstroke and middle to late upstroke, the forewing sweeping affects the Leading-Edge Vortex (LEV) through two factors: sweeping velocity and forward sweeping angle, and their effects are coupled. A large forward sweeping velocity can enhance the strength of LEV, while a large forward sweeping angle can weaken it. For early upstroke, the forewing sweeping can affect the wake capture mechanism, sweeping backward can enhance it while sweeping forward can weaken it. The findings in this work provide insight into the design of butterfly like Micro Air Vehicles (MAVs).

This work proposed an interpretable neural network specifically designed for structural inverse dynamic modeling, termed the Physical Embedded Neural Network (PENN). Distinguished from the Physics-Informed Neural Network (PINN), the PENN does not use the traditional perceptron as its basic neuron unit; instead, it employs newly designed neurons; instead, it uses newly designed neurons embedded with physical parameters as its fundamental units, enabling a convex optimization-based training through a dual-driven technique combining knowledge and data. Through simulation case studies, we apply the PENN to the inverse dynamics modeling of an 8-degree-of-freedom discrete system and a fixed-supported beam structure, both achieving high modeling accuracy. The dynamic parameter errors of the trained PENN are all less than 1 %. Furthermore, we apply the PENN to the inverse dynamics modeling of a fixed-supported beam structure, and a hybrid glass-carbon laminate experimentally. The results show that the errors of dynamic parameters embedded in the trained PENN for the fixed-supported beam are all less than 1 %, and for the laminate are all less than 10 %. The current study indicates that the proposed PENN can combine the rigor and interpretability of physical models with the flexibility of data-driven modeling methodology to establish analytical mapping relationships, creating a new paradigm for structural inverse dynamics modeling.

The wing of a butterfly consists of partially overlapping forewing and hindwing, and forewing sweeping can dynamically change the shape of the whole wing. In this work, the effect of forewing sweeping on aerodynamic performance of a butterfly like model is studied using a solver based on immersed boundary method and adaptive mesh. For aerodynamic performance, adding a "forward-backward-forward" sweeping motion to the forewing makes it more suitable for fast cruising flight, and compared to the situation without forewing sweeping, the drag is reduced by 46% and the lift to drag ratio is increased by 45%. On the contrary, adding a "backward-for ward-backward" sweeping motion to the forewing increases lift and makes it more suitable for climb flight. For downstroke and middle to late upstroke, the forewing sweeping affects the Leading-Edge Vortex (LEV) through two factors: sweeping velocity and forward sweeping angle, and their effects are coupled. A large forward sweeping velocity can enhance the strength of LEV, while a large forward sweeping angle can weaken it. For early upstroke, the forewing sweeping can affect the wake capture mechanism, sweeping backward can enhance it while sweeping forward can weaken it. The findings in this work provide insight into the design of butterfly like (c) 2024 The Author(s). Published by Elsevier Ltd on behalf of Chinese Society of Aeronautics and Astronautics. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/ licenses/by-nc-nd/4.0/).

The inner product matrix constructed by time domain vibration response under ambient excitation is a good structural characteristic parameter in structural health monitoring. In order to improve the identification accuracy of the structural health monitoring method using inner product matrix, more vibration response measurement points are often needed, which will directly affect the engineering practicability of the method.Based on the correlation analysis theory of time domain vibration responses, the inner product matrix is extended to the correlation function matrix to obtain more structural health characteristic information from a small number of vibration response measurement points, and the requirement of the number of measurement points for the structural health monitoring method will be reduced. Furthermore, combining with the excellent data feature extraction capability of convolutional neural network, a structural health monitoring method based on correlation function matrix and convolutional neural network is proposed with correlation function matrix as input and structural health status as output. The experimental results of the bolt loosening monitoring of a typical aeronautical stiffened panel show that the identification accuracy of the proposed method for bolt loose position can reach more than 99% by using only the time domain vibration responses of any two measurement points.

螺栓连接结构中的螺栓松动容易导致结构失效,如何对结构中的螺栓松动状态进行监测是当前研究的一个热点.该文利用环境激励下结构振动响应的相关性分析,结合深度学习技术,研究了一种联合使用内积矩阵(inner product matrix,IPM)和卷积自编码器(convolutional autoencoder,CAE)的神经网络模型,即基于内积矩阵及卷积自编码器(inner product matrix and convolutional autoencoder,IPM-CAE)的深度学习模型.通过对螺栓连接搭接板的螺栓松动状态监测的试验研究,验证了该方法的可行性及有效性,并与使用IPM的卷积神经网络(convolutional neural network,CNN)、堆栈自动编码器(stack autoencoder,SAE)及胶囊网络(capsule network,CapsNet)相比,IPM-CAE方法具有较快的网络训练收敛速度和较高的识别精度.

环境激励下仅利用振动响应的结构健康监测方法,因其便于实现在线监测受到了越来越多的关注.该文回顾了以振动时域响应相关性分析为基础的结构特征参数(即内积向量)的基本概念及特征.为了从已有测试数据中提取更多的结构特征参数,分别以各个响应测点为参考点来构建多个内积向量并组成矩阵,将内积向量扩展到了内积矩阵.进而以内积矩阵为结构特征参数,结合深度卷积神经网络的特征提取能力,提出了基于内积矩阵及深度学习的结构健康监测方法.典型航空加筋壁板螺栓松动监测的实验研究结果表明,仅利用结构在环境激励下部分测点的振动时域响应,该文方法可以准确地识别螺栓松动位置.

为了研究双脉冲发动机燃烧室内复杂热环境下三元乙丙(EPDM)绝热层的烧蚀性能,开展了工作时间为15s和两次点火工作时间为7.5s+7.5s的发动机实验.采用SEM电镜扫描、微米CT测试分析获得了烧蚀试件的表面宏观形貌、炭化层表面和断面微观形貌以及炭化层三维构型;利用测厚仪测量结果计算了试件的烧蚀率.结果表明,在总工作时间相等的情况下,双脉冲发动机中EPDM绝热层的烧蚀率比传统发动机大.与传统发动机中单次热冲击下烧蚀后试件相比,双脉冲发动机二次热冲击下烧蚀后试件的炭化层厚度减小约50％,总体孔隙率增大约13％;烧蚀表面致密层的致密程度也有所减小.双脉冲发动机工作时,EPDM绝热层的烧蚀性能在二次热冲击下发生较大变化,需在燃烧室内绝热层的设计过程中予以重视.

该文基于科研实验平台和成果,构建了基于固有频率向量的结构健康监测实验教学系统,旨在使学生掌握结构健康监测基本理论知识的同时,进一步接触科技前沿,提高科研实践能力和学习热情.实践结果表明,该教学系统力学模型简单明确,实验方法具有先进性,实验结果演示度高,有助于学生快速理解和掌握结构健康监测基本原理和实现方式.

间隙的存在给航空结构带来了很大的安全隐患.间隙会影响舵面的旋转频率,此时对舵面施加预载可以有效地克服间隙非线性的影响,从而获取舵面的旋转频率.本文根据静力学原理,设计加工了克服舵面间隙的预加载系统,并以此建立了带预载间隙舵面旋转频率测试方法.针对某飞机舵面旋转频率的试验结果表明,本文所提方法可以有效实现对舵面旋转频率的测试.

为了加深学生对静定薄壁结构力学分析方法的理解,让学生做到"知识-能力"相贯通,研制了静定薄壁结构教学实验装置,该装置结构紧凑,操作方便,并具有明确的工程背景.由学生根据电测法测量结构的应变,并与理论解进行比较,分析实验误差原因,来提高学生分析问题、解决问题和独立研究的能力.实践表明:静定薄壁结构实验与理论教学有机融合,有助于学生对综合分析和知识应用能力的培养.

为了保障固体火箭发动机C/C喷管的可靠性,建立了一套正确反映发动机喷管烧蚀过程的流固耦合计算模型,以实现对喷管烧蚀率的高精度预估.依据热化学烧蚀理论以及喷管内燃气与喷管结构体界面的质量平衡和能量平衡关系,建立并验证了考虑壁面退移的C/C喷管流固耦合方法,实现了燃气流动、异相化学反应、结构体传热三者间的耦合.通过实验发动机喷管的烧蚀计算,论证了模型的正确性,并分析了不同金属铝含量对烧蚀率的影响,计算所得的烧蚀率与实验值最大相对误差为4.3％,与不考虑壁面退移的耦合算法计算结果对比,计算精度最高可提升46％.计算结果表明:C/C喷管在喉部附近烧蚀最为严重;推进剂中Al含量的增加导致燃气中氧化组分浓度降低,进而减少了烧蚀速率,这些结论与C/C喷管烧蚀相关研究结果一致.

提出了一种基于灵敏度分析的平稳随机动载荷分段时域识别分析技术.将平稳随机动载荷样本分为若干小段,把每一小段内的平稳随机动载荷表示为正弦级数叠加的形式,通过灵敏度迭代分析来确定相应正弦级数的幅值,从而确定该时间段内结构所受的平稳随机动载荷.最后将各个时间段组合,即得到平稳随机动载荷样本的识别结果.试验结果表明,灵敏度分析识别方法能够很好地识别出作用在结构上的随机动载荷,并具有良好的抗噪性.

随着计算机与通信技术的发展,个人信息泄露的问题日益严重,人们对个人信息安全也越发重视,身份鉴别则是信息安全的重要一环,寻求更加方便快捷且可靠性高的身份验证方式成为了当今许多研究人员的研究重点.语言作为人类交流最重要的工具,每个人的声音都是独一无二的,因此将声音作为身份鉴别的技术引起了研究人员的兴趣.使用声音鉴别身份的原理是提取说话人语音的特征参数,为其建立数学模型,与待测语音进行比对,从而判断出说话人身份.同时随着深度学习技术的不断发展,解决了传统数学模型的过拟合问题,并且可以更好地对说话人特征进行学习.

将时延神经网络引入动载荷识别研究中,结合时延神经网络的"记忆"特性、因果有限长冲激响应(FIR)系统理论与振动响应的求解原理,提出一种利用时延神经网络的时域动载荷倒序识别方法.对一个受两点随机动载荷作用的舵面模型结构进行载荷识别验证实验,结果表明,用本文方法识别的两个激励点上识别载荷样本的时间序列与真实载荷样本的时间序列之间的均方根误差分别为0.635 4和2.543 7,识别载荷样本时间序列与真实载荷样本时间序列的相关系数分别为0.965 7和0.826 2,功率谱密度曲线也能够较好吻合.本文提出的方法具有不需要结构动力学模型、识别精度高的优点.

SOQPSK-TG(Telemetry Group version of Shaped Offset Quadrature Phase Shift Key)具有良好的频率利用率和功率利用率,广泛应用于无线通信系统当中.在连续通信模式下,SOQP SK-TG信号的同步主要采用直接判决算法.为进一步降低算法复杂度,推导了基于线性相位近似的最大似然估计误差鉴别器,理论上分析了算法估计性能,并搭建了简化的接收模型.通过仿真证明了算法在估计性能上优于脉冲幅度调制方法,算法误码率接近理论性能.

文章针对新工科背景下航空航天卓越工程师培养的需求,分析了"飞行器结构力学"课程在教学过程中的不足,提出了教学改革目标,并结合教学实践,开展引入概念结构力学知识、融入工程背景和典型案例分析、借助有限元仿真技术以及增加实验实践教学等具体的教学改革,从优化教学内容和构建实践平台等方面提出了进一步改革的构想.力求让学生在掌握结构力学基本概念的同时,不断了解新知识,锻炼工程能力,从而培养学生的工程创新能力.

为保证发动机能在恶劣的环境中运行,在绝热层的设计中,绝热层的厚度将直接影响着发动机结构的稳定性,而绝热层的烧蚀预估对于绝热层厚度的合理设计非常重要.为解决固体火箭发动机三元乙丙橡胶(EPDM)绝热层烧蚀性能工程预估问题,结合固体火箭发动机内两相流动的环境特点,以热化学烧蚀三方程模型和扩散化学动力学双控制机制为基本数学模型,以炭化层表面孔隙率为耦合参数,并综合考虑气流和粒子的侵蚀效应,建立了绝热层多因素耦合烧蚀模型的控制方程.通过对控制方程的隐式求解和对绝热层温度分布以及烧蚀线、炭化线、热解线位置的综合分析,获得了两相环境下EPDM绝热层的理论炭化烧蚀率.所得烧蚀率与实验结果对比,误差小于10％,表明给出的烧蚀预估方法可用于固体火箭发动机两相环境下EPDM绝热层烧蚀工程分析.

建立准确的结构动力学模型是结构响应分析的基础,由于模型简化的不确切等因素,必然会带来一定的误差,为了获得高精度的动力学分析模型,需要结合试验数据对模型进行修正.模态试验结果中包含了试件不同状态不同阶次的频率和振型信息,模型修正时需要建立多个目标函数,提出了一种基于动态加权系数的多目标模型修正方法.通过对解的群体实施进化,在每一代非劣解中,挑选各个子目标函数的局部最优解,计算各个局部最优解与子目标期望值的差距,并根据差距对加权系数动态调整,从而在进化过程中对加权系数进行优化,避免维数灾难问题,实现各个子目标函数的快速收敛.采用该方法对导弹全弹动力学模型进行了修正,子目标函数个数达到16个,与基于Pareto最优的模型修正方法相比,用较少的代数实现了各个子目标函数的收敛,提高了群体搜索的效率,取得了较好的修正效果.

基于时域振动响应的结构损伤检测方法因其便于实现在线监测受到了越来越多的关注.该文回顾了两种利用时域响应相关函数建立的结构特征向量(即内积向量及互相关函数幅值向量)及其对应的损伤检测方法.为了从时域响应相关函数中提取更多的结构健康信息,通过利用不同的结构响应组合,将上述两种结构特征向量扩展到了多种结构特征向量,并进一步采用数据融合理论,提出了检测精度更高的结构损伤检测方法.针对8层框架结构损伤检测的试验研究结果表明,该文方法可以对框架结构上的微小损伤进行定位.