Owing to the simple geometry and high volumetric energy density, liquid-fueled detonation engines have attracted considerable interest. However, their numerical simulation remains challenging as multiple processes, such as spray atomization, droplet breakup, turbulent mixing, and detonation wave propagation, coexist and interact with each other. In particular, within the Eulerian-Lagrangian framework, existing breakup models fail to accurately predict the droplet survival distance behind detonation waves, which could result in an overestimated propagation velocity of detonation wave. To this end, a theoretical analysis of the droplet breakup process is firstly conducted. It is found that the ratio of the droplet breakup time to the relaxation time can approach unity because of the smaller diameters of the child droplets and the high surrounding flow velocity. This suggests that the breakup behavior of the child droplets may differ from that of the parent droplets. Based on the theoretical analysis, an improved method is proposed in which the breakup process is inhibited when the ratio of breakup time to relaxation time exceeds a critical threshold. The proposed method is then incorporated into the Pilch-Erdman and Reitz-Diwakar models within the OpenFOAM® framework. Numerical simulations of droplet array breakup under detonation waves demonstrate that the improved method has better predictive performances across different conditions. Specifically, for the Reitz-Diwakar model with an initial kerosene droplet diameter of 40.7 μm, the droplet survival distance is improved from 6.11 mm to 12.56 mm (experiment: 12.73 mm), and detonation wave propagation speed from 2153 m/s to 1948 m/s (experiment: 1897 m/s). Similar results are also found for Pilch-Erdman models. Overall, the proposed model is effective, robust, and readily integrable into existing frameworks.
High-performance transitional gas turbine compressor development demands accurate Reynolds-Averaged Navier-Stokes (RANS) transition modeling. Conventional models are developed from physics and high-fidelity data of canonical flows, exhibiting notable deficiencies in transitional flows of compressors. This study addresses this limitation through data-driven recalibration of the widely-used Shear Stress Transport intermittency (SST-γ) model. Three recalibration scenarios are investigated: transition model parameters alone; transition model parameters with inlet turbulence conditions; combined turbulence/transition model parameters with inlet turbulence conditions. Using the Ensemble Kalman Filter method with compressor cascade surface data across 3 distinct operating conditions, the third recalibration scenario is demonstrated to achieve the most significant improvements, reducing prediction errors by 23% in blade surface pressure and 34% in skin friction coefficient for the training conditions. The recalibrated model also shows improved accuracy across 11 unseen operating conditions and multiple flow quantities including velocity profiles, momentum thickness distributions, and turbulence intensity. Sensitivity analysis reveals that prediction accuracy depends more strongly on turbulence model parameters and inlet turbulence conditions than on transition model parameters. The flow mechanism critical to the overall prediction accuracy is the laminar separation reattachment on the suction surface. The findings of this work provide useful tools and understandings for accurate modeling of transitional compressors.
This paper proposes SectionAwareNet, a novel deep learning approach for sonic boom prediction that incorporates aerodynamic area rule. The method addresses the computational inefficiency of traditional Computational Fluid Dynamics based (CFD-based) sonic boom prediction by developing a point cloud neural network that explicitly considers aircraft cross-sectional distributions. SectionAwareNet processes aircraft geometry point clouds through four key modules: point feature extraction, radial fusion, section pooling with attention mechanisms, and global feature aggregation. The model demonstrates superior training efficiency, achieving rapid convergence within a minute and an inference time of milliseconds per sample, while maintaining high accuracy. Experimental results validate the necessity of point cloud input and show that incorporating cross-sectional awareness significantly improves generalization, including reliable interpolation on in-distribution cases and near-extrapolation to unseen geometries under limited distribution shift, as well as robustness compared to conventional point cloud networks. The proposed method provides an efficient alternative for sonic boom prediction in design optimization scenarios requiring repeated modeling and extensive evaluations. Key innovations include the radial fusion module that enhances cross-section awareness and the attention-based section pooling that improves feature representation. The architecture achieves these benefits with relatively few parameters, making it particularly suitable for applications where computational efficiency is critical.
The classical Circumferential Mode decomposition method (CM) has been widely adopted in industrial applications. AECC observed the breakdown of the Nyquist criterion when circumferential modes are decomposed using a single non-uniform circumferential array. To identify the causes of CM failure, this study focuses on the relationship among array uniformity, the condition number of the decomposition matrix, and mode decomposition error. A randomized perturbation strategy was employed to investigate the dependence between array non-uniformity and the condition number. A metric C1 was introduced to quantitatively characterize the overall degree of array non-uniformity. Furthermore, CM influenced by the condition number is explored in detail for AECC typical practical application using Monte-Carlo simulations. These Monte-Carlo simulations allow the comprehensive evaluation of the accuracy of the dominant and weak mode decomposition error trend under investigation. The analysis reveals that single-ring circumferential mode decomposition using the CM is highly sensitive to the condition number, which is governed by both the mode decomposition range and the array uniformity. Consequently, uniformly spaced arrays are preferred when applying CM with a single circumferential microphone ring. The Basis Inner-product Residual (BIR) is introduced as a quantitative indicator to measure the deviation from discrete orthogonality in circumferential sampling and to reveal the intrinsic mechanism of error amplification in mode decomposition, and a detailed analysis was conducted for both dominant and weak mode errors.
This paper presents a comprehensive Multidisciplinary Design Optimization (MDO) framework for the aeroelastic tailoring of a Strut-Braced Wing (SBW) using tow-steered composites manufactured via the Discrete Stiffness Tailoring (DST) technique. The primary objective is to minimize structural weight and eliminate material failure while maximizing favorable aeroelastic twist under strict strength and stability constraints. A Genetic Algorithm (GA) optimizer is implemented in MATLAB and coupled with MSC Nastran to perform high-fidelity sequential analyses that consist of static aeroelasticity, flutter, composite failure, linear buckling, and discrete gust response. A representative rectangular wing with an aspect ratio of 19.4 and a strut located at 50% of the span is considered. The wing semispan is discretized into 14 spanwise partitions. Design variables include ply orientations and stacking sequences for the lower and upper skin within each spanwise partition. Optimization results demonstrate that the DST configuration achieves an 8.1% weight reduction under Single Objective Optimization (SOO) and a 5.1% reduction under Multi Objective Optimization (MOO), compared to an optimized Uniform Laminate Configuration (ULC) baseline. The tailored stiffness distribution reduces the static wingtip vertical displacement and provides significant passive gust load alleviation, attenuating the peak gust response by approximately 50%. Furthermore, a critical trade-off is identified: while a negative unit twist effectively alleviates static aeroelastic loads, it introduces stiffness couplings that reduce flutter margin, necessitating a balanced multi-objective approach. The proposed framework highlights the substantial structural and aeroelastic performance gains offered by DST for next-generation sustainable aircraft.
As the core power source of aero‑engine systems, the manufacturing of aero‑engine components requires the joining of dissimilar superalloys to achieve complementary advantages. Brazing features low influence on base metals and high compatibility for dissimilar materials, which enables it to achieve reliable joining of high‑precision aviation components. In this study, BNi‑2 filler was adopted to achieve the joining between GH4169 and GH2132 superalloys. The effects of brazing temperature and holding time on the microstructural evolution and mechanical behavior of the joints were systematically examined in this research, alongside an analysis of the brazed interface formation mechanism. According to the characteristics of the interfacial microstructure and elemental distribution, the brazed joint can be divided into three distinct zones: ISZ, DZGH4169, and DZGH2132. As the brazing temperature rises, the widths of both ISZ and DZ increase gradually. With extending holding time, the width of ISZ decreases gradually, while the width of DZ increases continuously. The typical interfacial microstructure of the joint can be described as follows: GH2132/(Cr,Mo)-rich borides+MC/ (Ni,Si,Nb)-rich phases+(Cr,Nb,Mo)-rich borides+γ-Ni(s,s)+(Cr,Mo)-rich borides/(Cr,Nb,Mo)-rich borides+(Nb,Ti)-rich phases/GH4169. The maximum average tensile strength is 622.1 MPa, corresponding to a brazing temperature of 1100 ℃ and a holding time of 20 min. The maximum average fracture strain is 9.49 %, corresponding to a brazing temperature of 1100 ℃ and a holding time of 10 min. Further elevated-temperature tensile tests at 650 ℃ reveal significant deterioration in tensile strength and fracture strain for both groups of brazed joints. Overall, this study provides an experimental basis for joining dissimilar superalloy components using BNi‑2 brazing filler.
Three-dimensional Method of Curved-shock Characteristics (3D-MOCC) is derived based on the three-dimensional Curved Shock Theory based on orthogonal frame of shock surface (3D-CST-boos). Explicit equations are derived in the form of gradients. The proposed method is applied to three-dimensional supersonic flow field inverse design and flow analysis. The main idea is to determine the gradients of the pressure and flow deflection angle in the streamline-characteristic coordinate system. With the acquired derivatives, the flow parameters of the post-shock flow field can be found. Compared with the Local-Turning Osculating Cones (LTOCs), the gradient information obtained through 3D-MOCC are the important parameters for analyzing supersonic fluid dynamics. Axisymmetric/non-axisymmetric supersonic internal flow fields are inversely solved using the 3D-MOCC, and the results show that the proposed method obtains higher efficiency than the LTOCs while achieving superior accuracy. The fundamental mechanics of flow differences between axisymmetric and non-axisymmetric internal flow fields are analyzed using gradient information. The accuracy of 3D-MOCC is verified by numerical values and wind tunnel experiments. This method is of significant value for analyzing and designing supersonic flow fields.
Surrogate-assisted transfer optimization methods facilitate data reuse and improve efficiency in high-dimensional, high-fidelity problems. However, their performance is often limited in practice by inadequate domain adaptation and poor interpretability. To address these limitations, this study proposes an explicit knowledge transfer method with fitness-based domain adaptation. First, Radial Basis Function (RBF) surrogate models are constructed for source tasks, and domain transformation coefficients are derived using target-task samples. A Similarity Metric based on Fitness (SMF) is introduced to quantify adaptation effectiveness. Second, the transformed source model with the minimum SMF value is used to generate an interpretable set of transferable solutions. By incorporating these solutions into the initial target-task sample set, an RBF surrogate model is constructed for the target task, enabling optimization to start from a near-optimal region. Finally, the target task is iteratively optimized using a balanced sampling criterion. The proposed method is evaluated on high-dimensional benchmark examples and two-dimensional airfoil optimization cases in Computational Fluid Dynamics (CFD). Compared with conventional optimization methods, it significantly reduces target-task model evaluations while improving the interpretability and effectiveness of knowledge transfer.
The anti-/de-icing method employing Surface Dielectric Barrier Discharge (SDBD) plasma actuators have gained considerable attention for aircraft applications, owing to their compact structure, fast response, minimal weight, and ease of integration. Despite extensive research over the past decade, the physical mechanisms governing anti-/de-icing based on SDBD plasma actuators remain inadequately understood. In particular, the mutual interactions between impinging water droplets and the flow fields created by SDBD plasma actuators have not been fully explored. This work experimentally investigates the dynamic coupling between a single impacting droplet and a SDBD plasma actuator using infrared thermography and high-speed Particle Image Velocimetry (PIV). The evolution of droplet morphology and the induced flow field are discussed throughout the droplet impact process, including free fall, spreading, and recoiling stages. Based on the measured data, a predictive model is developed to estimate the maximum spreading ratio of droplets under plasma actuation within limited ranges of Reynolds and Weber numbers.
The folding fin, mainly designed for space saving of aircraft, exhibits structural nonlinearities that significantly influence its aeroelastic behavior. This study focuses on the nonsmooth stiffness induced by freeplay among the connecting components of the folding mechanism. Based on experimental observations, these nonlinearities are captured using a hyperbolic tangent function for low-amplitude softening combined with a linear function for high-amplitude hardening, corresponding to the distinct responses at different resonances. An improved substructure synthesis technique is proposed to reduce the finite element model with nonsmooth stiffness, enhancing computational efficiency while improving accuracy in dynamic analysis. Following calibration using vibration test data, the resulting reduced-order model with these parameters matches the experimentally measured jump frequencies well. Aeroelastic analysis conducted with this model reveals the complete bifurcation behavior of limit cycle oscillations through numerical continuation, while aperiodic responses are obtained by direct numerical integration. Subcritical Hopf bifurcation, fold bifurcation, and Neimark-Sacker bifurcation are characterized using Floquet multipliers. Multiple branches of stable and unstable limit cycle oscillations are identified within specific flow velocity ranges, along with other complex aeroelastic phenomena. Finally, the effects of nonlinear stiffness parameters on aeroelastic bifurcation characteristics are investigated, revealing the influence patterns of parameters on flow velocities at bifurcation points and the variation of velocity ranges for aperiodic motion.
Accurately modeling the complex nonlinear drag map remains a central challenge in preliminary nacelle design. However, traditional surrogate models often suffer from substantial data demands, pronounced deviations in key performance metrics, and limited physical consistency during the design process. This paper evaluates the modeling performance of classical Response Surface Methods (RSM) and neural-network methods, and further proposes a Knowledge-Constrained Neural Network (KCNN) framework. The core idea is to incorporate the monotonic relationship between drag coefficients and mass flow capture ratios in nacelle design, along with empirical drag spillage formulas, as prior knowledge into the modeling process. The proposed approach is validated using a high-resolution numerical example based on representative nacelle design parameters. The results show that the KCNN reduces sampling density by 84.7% compared to the conventional approach, while preserving the accuracy of key performance metrics, including cruise-condition drag and the drag-rise boundary. Under small-sample conditions, the prediction errors of the KCNN for these metrics are reduced to approximately one quarter and one third of those obtained by traditional RSM and conventional neural network methods, respectively. This approach provides an effective perspective for rapid performance evaluation in preliminary nacelle design.
This paper provides a concise review of riblet-based strategies for skin-friction drag reduction, with a particular focus on emerging and unconventional designs. We first summarize the performance of conventional straight riblets, aligned with the mean flow direction, which are known to achieve drag reductions of up to about 10% under optimal conditions. We then explore alternative riblet geometries that have been proposed to address key limitations of straight configurations, notably their sensitivity to the design point and the rapid degradation of performance under off-design conditions. Concepts such as wavy, converging–diverging, and herringbone riblets suggest new pathways for broadening the operational envelope and potentially achieving enhanced drag-reduction levels. In this context, the integration of advanced optimization frameworks and machine-learning techniques offers a promising avenue for systematically exploring high-dimensional design spaces and uncovering novel riblet configurations with improved robustness and performance.
This study introduces a novel approach for generating smooth three-dimensional paths for Unmanned Aerial Vehicles (UAVs) using a Surface-based Continuous Frame (SCF) methodology. The proposed technique addresses a fundamental limitation in traditional path planning by ensuring continuity of position, tangent, and normal vectors at waypoints and across regions of zero curvature. By mapping a planar Pythagorean Hodograph curve onto a minimal surface with Pythagorean normal, the resulting spatial curve provides a mathematically precise representation that facilitates exact arc length computation and efficient offset curve generation. The SCF approach overcomes discontinuities inherent in conventional Frenet frames, enabling consistent frame transitions essential for stable UAV guidance and control. Building upon this foundation, a real-time collision avoidance algorithm is developed that generates smooth detour paths around unexpected obstacles while minimizing deviation from the reference trajectory. Comparative analyses demonstrate that the proposed approach achieves superior performance over conventional methods, maintaining continuous frames throughout the path and producing more efficient obstacle avoidance trajectories. The algorithm’s effectiveness is validated through extensive Monte Carlo simulations and 6-DOF tracking simulations, confirming its practical utility for UAV applications requiring precise path planning and adaptive navigation in dynamic environments.
The arresting process during carrier-based aircraft landing involves highly transient and nonlinear dynamics. To enhance the recovery safety of lightweight Unmanned Aerial Vehicles (UAVs), this study establishes an integrated numerical–experimental framework for analyzing wheel–cable impact dynamics. A refined finite element model couples a visco-hyperelastic tire, an equivalent beam cable, and a nonlinear shock absorber. The implicit–explicit coupling strategy efficiently captures high-frequency impact responses while improving computational efficiency. A scaled experiment validated the model, with errors in touchdown time, vibration period, and rebound amplitude within 6%, and the theoretical bending-wave velocity matching simulations within 1.5%. Parametric analyses show that cable pretension is the dominant factor, with an optimal value of 2.5 kN producing the maximum amplitude of vertical vibration and the most effective energy storage and release. Vertical load and approach velocity have minor effects, whereas support configuration (height, spacing, stiffness) strongly influences both vertical and longitudinal responses, with amplitude differences exceeding a factor of two. The framework quantitatively elucidates wheel–cable dynamics and provides guidance for UAV arresting-system optimization.
The energy balance of a satellite is centered around its battery subsystem. The importance to closely monitor the health conditions of the batteries on-board a satellite is evident. To provide such monitoring, traditional methods demand significant amounts of on-board sensor data to be transmitted back to ground, then be processed with sufficient ground-based processing power. In recent years, the rapid proliferation of Low-Earth Orbit (LEO) constellations with tens of thousands of LEO satellites planned to be launched into orbit is out-pacing the development of ground-based operation management capacity, while simultaneously saturating the satellite-ground telemetry link. This, in turn, introduces unprecedented challenges for constellation-level health management. The industry has shown an urgent shift to on-board autonomies, leveraging the abundance of in-situ data while off-loading the workload on the ground system, in order to enable satellites to partially govern their own operational status. In this paper, we introduce the Physics-Informed Transformer for Battery Health (PITBH), a novel deep learning architecture designed with on-board integration constrains in mind. The model’s innovations are threefold: A hybrid Linformer-GRU architecture that effectively utilizes the intrinsic advantages of both models while maintaining a small cost footprint; A delta-prediction paradigm that targets cycle-to-cycle capacity changes rather than absolute SOH, fitting the practical requirements of on-board systems; The integration of Differential Capacity Analysis (DCA) into the loss function, providing physical constraints that the model enforces. The resulting framework is computationally efficient, making it a promising candidate for engineering implementation. Evaluation demonstrates that despite its fundamentally different cumulative integral approach, PITBH achieves highly competitive accuracy compared to established models, validating it as a powerful and practical solution for intelligent battery management in future on-board missions.
A fundamental inverse problem, extracting the wavespace of metastructures from structural response spectra, is crucial to understanding the dynamic properties of metastructures and the optimal design of their vibroacoustic performance. The validity examination of inverse problem algorithms is still an open question. The present study systematically investigates the robustness and accuracy of two main wavespace identification algorithms for determining complex wavenumber fields: the linear Algebraic Wavenumber Identification (AWI) method and the nonlinear Green’s Function Correlation (GFC) approach based on an inhomogeneous wave correlation model. Direct numerical approaches are employed to evaluate both algorithms by various metastructural configurations. The robustness analysis demonstrates that GFC fails for sandwich plates with thick soft cores due to its inherent elastic assumptions, whereas AWI maintains reliable performance. To assess accuracy, sandwich structures with frequency- and temperature-dependent viscoelastic cores are examined. The AWI framework accurately reconstructs the omnidirectional dispersion relation and the corresponding damping performance. Finally, a metastructure incorporating tuned mass dampers is designed and validated through a closed-loop framework of the inverse AWI and a forward wave-based surrogate model.
With the sufficient maneuverability, the larger flight range is an important requirement for future Air Superiority Fighters (ASFs) propulsion systems. It is determined by the fuel consumption rate and energy conversion efficiency in subsonic cruise state. However, the current Low-Bypass Ratio Turbofan Engine (LBRTE) is limited by low-pressure rotor shaft power during subsonic cruising, making it difficult to significantly reduce fuel consumption. To address these issues, this paper proposes a novel Inter-mixed Two-stage Low-Pressure Turbine Engine Architecture (ITLPTEA), establishes a component-level model for the engine, and analyzes its variable cycle characteristics and performance. This engine is based on a traditional LBRTE and common variable cycle components. It innovatively incorporates an inter-mixed two-stage low-pressure turbine and adjustable cooling bleed air. After a series of simulation tests, it is found that the ITLPTEA engine can provide the same specific thrust as a traditional LBRTE, ensuring ASF’s maneuverability requirements for supersonic mission. At the (13 km, Ma = 0.9) subsonic cruising, the ITLPTEA engine reduces the fuel consumption from 0.981kg/(kgf∙h) to 0.855kg/(kgf∙h), down by 12.8%. For the same mission profile, it can save 359.4 kg of fuel, which is equivalent to increasing the flight range by 161.7 km.
Buzz-saw noise radiated by transonic fan rotors is highly sensitive to blade geometric deviations. The additional effect of in-service blade vibrations, another type of geometric deviation, remains unclear. To bypass the computational costs of high-fidelity methods, an approximate method is proposed for rapidly predicting buzz-saw noise from vibrating fan blades. The key idea is to establish an analogous relationship between the dominant vibration modes (first bending and/or first torsion) and periodic azimuthal, axial, and stagger deviations of the blade. This analogy enables the integration of a Mode-based Fast Reconstruction (MFR) technique for source generation and an Eikonal-based Solver (EBS) for nonlinear shock propagation. The method is validated through fluid-structure interaction simulations and applied to modified NASA rotor 67 configurations with prescribed initial deviations and vibrational motions. Comparisons among non-vibrating, same-phase, and random-phase vibrating rotors reveal that: (A) the Overall Sound Pressure Level (OASPL) contributions from initial deviation, same-phase vibration, and random-phase vibration are approximately 5.4 dB, 5.5 dB, and 7.3 dB, respectively; (B) among geometric factors, stagger deviation amplitude exerts the strongest influence, followed by inter-blade phase angle, and then azimuthal and axial deviation amplitudes. The proposed approach offers a computationally efficient tool for assessing vibration-induced buzz-saw noise at the design stage.
Resource allocation and detection threshold adaption are critical to unlocking the full potential of Phased Array Radar (PAR) network for maneuvering target tracking under clutter environment. A Dwell Time Allocation (DTA) incorporating Interacting Multiple Model-Bayesian Detector (IMM-BD) is proposed in this background. The IMM-BD is derived to tune detection thresholds. The information reduction factor influenced by IMM-BD is calculated, which is then embedded in the derived posterior Cramér-Rao lower bound in IMM framework. The DTA optimization model is formulated as minimizing the sum of weighted predicted PCRLBs under dwell time budget of each PAR. It is shown that the optimization model is nonconvex. The modified gradient projection method is proposed for solution. Simulation results confirm the effectiveness and efficiency of proposed strategy in terms of lower tracking errors and lower track loss ratios, compared with state-of-the-art algorithms.
