Biography: Cheng Yan is a tenured associate professor and doctoral supervisor at Xiamen University’s School of Aerospace Engineering. He earned both his bachelor's and doctoral degrees from Beihang University under Academician Yin Zeyong and specializes in aero-engine component design and optimization. His honors include the China Association for Science and Technology Young Elite Scientists Sponsorship Program, selection as an outstanding young talent in Fujian’s sixth “Chuying Program” cohort, Excellent Young Scientist of Fujian Province, and Class B High-Level Talent of Fujian Province. His achievements include 33 SCI/EI papers as first/corresponding author (23 SCI, 16 Q1), 28 authorized/filed invention patents as first inventor (12 authorized), 6 utility model patents, 3 software copyrights, 4 monographs, 10+ invited talks, and an invitation from the National Natural Science Foundation of China (NSFC) to present a General Program project. Over years, he led 20+ projects, including a National Science and Technology Major Project subproject, a national engineering program project, NSFC General/Youth Programs, a science and technology commission key-project subproject, and China Postdoctoral Science Foundation Special/General Programs. He is a dual-appointed expert and Young Visiting Scholar at Taihang National Laboratory, a young member of the Ninth Committee of the Mechanical Design Branch of the Chinese Mechanical Engineering Society, Guest Editor for Aerospace, and a young editorial board member of Transactions of Nanjing University of Aeronautics & Astronautics, Journal of Nanjing University of Aeronautics & Astronautics (Natural Sciences), Advanced Design Research, and Aeroengine. He reviews for Chinese Journal of Aeronautics, Structural and Multidisciplinary Optimization, Propulsion and Power Research, Energy, and other journals.
Speech Title: Structural Design of Additively Manufactured Disk Components: From Performance-Driven Design to Process Coordination
Abstract：Advanced aero-engine disk components must balance lightweighting, high strength, manufacturability, and long service life, while extreme service environments and complex failure modes impose stringent design requirements. Additive manufacturing expands the geometric design freedom of disk components; however, strength control under complex thermo-mechanical coupling, the manufacturability of internal overhang features, and material anisotropy remain critical challenges. To address these issues, this presentation proposes a topology optimization framework for additively manufactured disk components that spans performance-driven design and process coordination, providing methodological support for the structural design of advanced aero-engine disk components. First, a thermo-mechanically coupled topology optimization framework is established by accounting for design-dependent temperature fields and temperature-dependent material properties. Temperature-dependent thermophysical and elastic parameters are incorporated into nonlinear heat-conduction and thermoelastic analyses. Combined with temperature-dependent yield strength, a hazard-coefficient-based strength constraint is constructed to achieve precise strength control of turbine disks under high-temperature and high-speed operating conditions. Second, an internal self-supporting constraint method is proposed and integrated with stress constraints. This method suppresses difficult-to-print internal overhang features while controlling stress concentration, thereby balancing structural strength and manufacturability. It also preserves the freedom of external overhang evolution, reducing the need for internal supports and lowering post-processing difficulty. Furthermore, for multiaxis additive manufacturing, an anisotropic constitutive model and a coordinate-transformed Tsai–Hill failure criterion are established and embedded in a concurrent optimization framework for structural topology and build direction. The proposed method captures anisotropic failure behavior, avoids stress concentration, and enables the coordinated optimization of structural topology and build direction.
Biography: Chenzhou Xu received his B.E., M.E., and Ph.D. degrees from Northwestern Polytechnical University and obtained his Ph.D. in Fluid Mechanics in 2024. He is currently a postdoctoral researcher and Assistant Researcher at the Department of Precision Instrument, Tsinghua University. His research focuses on aircraft aerodynamic modeling and design, multidisciplinary optimization, and LLM-assisted autonomous decision-making for unmanned systems. He is the principal investigator of the NSFC Young Scientists Fund (C). He has been selected for the Beijing Young Talent Support Program and the Shuimu Tsinghua Scholar Program. He has published 26 papers in aerospace engineering and artificial intelligence, including 1 ESI Highly Cited Paper, with over 500 citations. He has received several honors, including the National Scholarship for Doctoral Students, the championship in an IEEE international optimization competition, and the First Prize in the China Graduate Future Aircraft Innovation Competition.
Speech Title: LLM-Assisted Automated Algorithm Design for Aerospace Engineering Applications
Abstract：The design of high-performance algorithms for complex aerospace problems often relies heavily on domain expertise, manual modeling, and extensive trial-and-error. Recent advances in the reasoning and code generation capabilities of large language models (LLMs) provide new opportunities for automated algorithm design. By integrating LLMs with evolutionary search and task-specific performance evaluation, candidate algorithms can be automatically generated, evaluated, selected, and iteratively improved. This report introduces the basic framework of LLM-assisted automated algorithm design and its applications in aerospace engineering, focusing on unsteady aerodynamic modeling, turbulence modeling, and autonomous decision-making for unmanned aerial vehicles. In these applications, LLMs are used to discover interpretable aerodynamic model structures, generate and refine turbulence model formulations, and evolve guidance, planning, and decision-making rules through simulation feedback. These studies demonstrate the potential of LLM-assisted automated algorithm design to reduce reliance on handcrafted formulations and improve the efficiency and flexibility of algorithm development.
Biography: Dongdong Zhang is an Associate Professor at the College of Aerospace Science and Engineering, National University of Defense Technology. His research focuses on high‑speed flow processes and their control strategies. He has led over ten research projects and published more than 60 academic papers. He is a committee member of the Supersonic Aircraft Branch, CSAA. He also serves on the youth editorial boards of several academic journals. He is a recipient of the CAST Young Talent Support Project and the Hunan Provincial Science and Technology Innovation Talent Program, and was awarded the Third‑Class Natural Science Prize by the CSAA as the first contributor.
Speech Title: Mixing Processes of High-Speed Jets and Relevant Control Approaches
Abstract：Mixing processes of high‑speed jets are widely encountered in propulsion systems of high‑speed vehicles. Rapid mixing is critical for efficient combustion and improved propulsion system performance. Aiming at the challenges in high‑dimensional jet‑mixing research including the difficult identification of flow‑field structures, the poorly understood mixing mechanisms, and the unsatisfactory performance of mixing‑enhancement techniques, Present talk presents recent progress of our research group on the resolving fine‑scale flow structures of high‑speed jet mixing, the machine‑learning‑based flow control, and the practical implementation of mixing‑enhancement methods.
Biography: Jiaao Hao is an Associate Professor in the Department of Aeronautical and Aviation Engineering at The Hong Kong Polytechnic University. His research focuses on aerodynamics, aerothermodynamics, and flow stability. He has led projects funded by the National Natural Science Foundation of China and the Hong Kong Research Grants Council, and has published over 70 papers in leading journals such as Journal of Fluid Mechanics and AIAA Journal. Prof. Hao received his BEng in Aircraft Design and Engineering in 2013 and PhD in Fluid Mechanics in 2018 from Beihang University. After completing his PhD, he joined PolyU as a postdoctoral researcher. He was appointed Assistant Professor in 2020 and promoted to Associate Professor in 2025. In 2025, he also served as a visiting professor at École Nationale Supérieure d'Arts et Métiers in Paris, France.
Speech Title: Secondary Global Instabilities in Shock-induced Separated Flow
Abstract：Two-dimensional shock-induced flow separation can sustain three-dimensional (3-D) global instabilities. Under marginally unstable conditions, only a single stationary global mode is typically observed, implying that the flow undergoes a primary bifurcation to a steady 3-D state. However, direct numerical simulations (DNS) of a Mach 2.15 impinging shock-wave/boundary-layer interaction reveal that, following the primary bifurcation, secondary instabilities emerge and drive the flow into a limit-cycle state. To explain the origin of this unsteadiness, a 3-D global stability analysis is performed, which identifies an oscillatory global mode associated with strong flow distortion due to the primary instability. The effect of spanwise domain width is also discussed.
Biography: 胡吉林，华东师范大学，教授/博士生导师，国家级青年人才。曾担任奥尔堡大学长聘副教授，阿联酋起源人工智能研究院担任研究员，加州伯克利分校访问学者。研究方向主要是时空数据管理与分析，时序数据分析，科学智能。其先后主持丹麦Villum Synergy交叉合作项目、新一代人工智能国家重大专项课题、国家自然科学基金面上项目等项目，相关成果发表于Advanced Materials、ICML、NeurIPS、ICLR、SIGMOD等材料和计算机领域国际顶级期刊/会议上发表论文70余篇，曾获得ICDE 2022最佳论文奖，PVLDB 2024最佳论文奖提名，EMNLP 2025 （工业赛道）最佳论文奖，ACM SIGSPATIAL中国新星奖。
Speech Title: OpenTS-Bench：时间序列预测与异常检测基准平台
Abstract： 时间序列数据广泛存在于航空航天工程的各类关键场景中，例如飞行器状态监测、发动机健康管理、卫星遥测分析、故障预警与能源系统运行等。因此，面向航空航天复杂系统开展可靠的时间序列分析研究具有重要意义。为推动相关方法的研发、验证与工程应用，我们提出了 OpenTS——一个面向时间序列预测与异常检测方法的自动化基准平台。OpenTS 针对现有研究中数据场景覆盖有限、对传统方法关注不足，以及实验流程不统一、扩展性较弱等问题，引入来自交通、健康、能源及工业系统等 10 个领域的数据集，以覆盖航空航天任务中常见的多源异构、强时变和复杂工况特征，并通过时间序列特征化评估数据集的代表性与完整性。此外，OpenTS 支持统计学习、机器学习和深度学习等多类方法的统一集成，并提供多种评估策略与指标，从而实现对不同算法性能、鲁棒性及工程适用性的全面评估，为航空航天系统的状态预测、异常识别和智能运维提供标准化、可复现的实验基础。
Biography: Jingbo Wei, male, born in 1986, is an associate professor and doctoral advisor. He currently serves as Chair of the Department of Aeronautical Engineering at the School of Aeronautics and Astronautics, Sun Yat-sen University, and is a recipient of the university’s “Hundred Talents Program.” His primary research interests include new-concept unmanned aerial vehicle (UAV) technology and high-performance electromagnetic actuation and control technology. He has published more than ten papers in high-impact journals such as "IEEE Robotics and Automation Letters" and "IEEE Transactions on Power Electronics", and has led more than ten national and military research projects.
Speech Title: Research on Overdriven Active and Passive Tiltrotor UAV Technology
Abstract: Tiltrotor technology transforms traditional multirotor UAVs from underdriven to overdriven by expanding thrust vectoring from a fixed direction to a controllable one, thereby enabling decoupled control of position and attitude. In terms of active tilting, the team developed a two-degree-of-freedom tiltable trirotor UAV, whose thrust vector can be directed toward any orientation in three-dimensional space, enabling omnidirectional maneuverable flight; the tilting quadcopter, meanwhile, achieved a major breakthrough in wind-resistant control under wind conditions of force 8, with each rotor tilting independently to instantly compensate for external disturbances. In terms of passive tilting, the team proposed a tiltrotor design based on passive articulation, which enables thrust vector deflection without the use of tilt servos, resulting in a simpler structure and higher reliability; by further integrating a lift-generating fuselage design, the system utilizes lift generated by the fuselage itself to effectively enhance range and efficiency. The research has established distinctive strengths in modal transition control, overdrive control allocation, and wind-resistant robustness, demonstrating clear application potential in scenarios such as emergency rescue, ship-based takeoff and landing in rough seas, and high-wind inspections.
Biography: Jin-Hua Hu is a Lecturer at the School of Construction Machinery, Chang’an University. Her research focuses on network modeling, assembly schedule monitoring, and intelligent workshop scheduling methods for aircraft final assembly in the context of intelligent manufacturing. As a key technical contributor, she has participated in multiple projects on aircraft final assembly scheduling optimization. She has published five papers in leading journals, including the Journal of Manufacturing Systems (JCR Q1, IF:14.9), Advanced Engineering Informatics (JCR Q1, IF:11.5), and Computers & Industrial Engineering (JCR Q1, IF:7.3).
Speech Title: AI-Native Scheduling Cognition and Optimization for Aircraft Final Assembly with Resource Unavailability Time Windows
Abstract: Aircraft final assembly is a critical stage in aircraft manufacturing, characterized by complex precedence relationships, strong resource coupling, scarce skilled workers, and frequent resource unavailability caused by maintenance, leave, dispatching conflicts, or job interruptions. These factors make it difficult to generate stable and efficient assembly schedules using conventional resource-constrained project scheduling methods. To improve the practicality of scheduling schemes, this study proposes an AI-native scheduling cognition and optimization approach for aircraft final assembly by integrating complex network-based job influence evaluation, resource-unavailability-aware scheduling modeling, and job importance-guided intelligent optimization. First, a network-based multi-attribute job influence evaluation method is developed. The proposed method incorporates job duration, resource requirements, local topological characteristics, and global topological structure to evaluate job importance. The method enables a more comprehensive understanding of critical assembly jobs and provides knowledge-guided priority information for scheduling optimization. Second, a resource-constrained project scheduling model with resource unavailability time windows is established to minimize the project makespan. Furthermore, a job importance-guided hybrid genetic algorithm is designed to generate high-quality scheduling solutions. Numerical experiments on benchmark project scheduling instances and real aircraft final assembly data demonstrate that the proposed method achieves better scheduling quality. The proposed approach provides a practical pathway for advancing AI-native scheduling cognition, resource-aware decision-making, and intelligent optimization in aircraft manufacturing.
Biography: Lifang Zeng, born in January 1991 in Shaoyang, Hunan, holds a Doctor of Philosophy in Fluid Mechanics. She is currently a Master’s Supervisor at the School of Aeronautics and Astronautics, Zhejiang University. In the past five years, Dr. Zeng has published more than 20 SCI/EI indexed journal papers as the first or corresponding author, and obtained 10 invention patents as the primary inventor. She presides over one research project supported by the National Natural Science Foundation of China (NSFC), together with three provincial and ministerial-level research projects. Dr. Zeng acts as a judge for the China University Aircraft Design Competition (CUADC), and serves on the Youth Editorial Board of Journal of Rocket Propulsion and Aerospace Technology. Her main research interests cover intelligent design of unmanned aerial vehicles, bio-inspired aerodynamics, rotorcraft aerodynamics, and the development of unmanned aerial systems.
Speech Title: Generative Concept and Aerodynamic Design of UAVs Driven by Artificial Intelligence
Abstract: Conventional aircraft design relies on expert experience, extensive numerical simulations and physical tests. Such trial-and-error design processes is time-consuming and always fails to obtain the optimal configuration. Combined with considerable number of UAV design databases and aircraft design knowledge, a theoretical framework and methodology for generative concept and aerodynamic design of UAVs are proposed. Large language models (LLMs) and generative artificial intelligence are applied in this framework. Prompt engineering and retrieval-augmented generation (RAG) techniques are integrated to realize end-to-end mapping from natural-language design requirements to UAV conceptual configurations. Furthermore, a generative innovation framework based on diffusion models is constructed to rapidly generate three-dimensional aerodynamic shapes. This report finally discusses the development trends and prospects of UAV intelligent design.
Biography: Dr. Ming Zhao currently serves as a researcher/doctoral supervisor at Tianjin University, focusing on research on "high-precision and high-efficiency multi field simulation - active and passive load control". Published over 60 SCI papers and 37 first author/corresponding papers in JCP, AIAAJ, PRF, CAF, etc. Hosted 10 national level projects, including 2 National Natural Science Foundation/Youth Project, 1 National Numerical Wind Tunnel Project, 2 National Key Research and Development Program sub projects, and 2 National Defense Basic Research Projects. Hosted 8 provincial and ministerial level projects.
Speech Title： Efficient Discontinuous Finite Element Method and Its Application in Flow Field Simulation
Abstract: This work presents two complementary enhancement strategies to address the key bottlenecks of implicit discontinuous Galerkin (DG) methods—namely, inefficient linear system iteration and degraded time-marching due to increased degrees of freedom. First, a Morton curve reordering based on bitwise operations is introduced to improve cache performance by ensuring contiguous memory storage of neighboring cells. Second, an information compressed Jacobian difference (ICJD) scheme is developed, which achieves linear time complexity and ease of implementation through graph coloring estimation adapted to cell adjacency graphs. Both algorithms are integrated into an implicit DG framework and validated via viscous and inviscid flow simulations under subsonic and transonic conditions. The Morton reordering not only enhances cache performance by up to 39.4% but also accelerates GMRES convergence in certain cases. Quantitatively, ICJD executes in approximately 0.1% of the runtime of the original finite-difference method, while Morton reordering reduces iteration counts by 88.9% in a viscous case compared to the reverse Cuthill–McKee (RCM) ordering. Performance gains are more pronounced for viscous flows, indicating strong potential for efficient turbulence simulations.
In parallel, to further accelerate DG solvers, we investigate the discontinuous Galerkin spectral element method (DGSEM) with a p-multigrid approach. Recognizing that fixed level-skipping coarsening strategies are often problem-dependent, we propose an adaptive p-coarsening multigrid method based on quasi-a priori truncation error estimation (τ-estimation), which dynamically determines the hierarchy of polynomial orders. Exploiting the tensor-product property of DGSEM, both global and locally anisotropic coarsening strategies are developed. Two-dimensional numerical experiments for inviscid and viscous flows demonstrate significant efficiency improvements, with the locally anisotropic strategy achieving over 45% CPU time acceleration in viscous cases. Notably, the non-isolated truncation error formulation consistently outperforms its isolated counterpart, and the inclusion of a correction term proves necessary when using non-converged solutions in the error estimation. The proposed method is further extended to a 3D viscous flow around a sphere, yielding 44% acceleration in multigrid cycles, and applied to a large-eddy simulation (LES) of flow past an SD7003 airfoil, confirming its feasibility and promise for complex turbulent flow simulations.
Together, these two enhancement frameworks provide robust and efficient tools for implicit DG and DGSEM solvers, with particular efficacy for viscous and turbulent flow problems.
Biography: Dr. Ping Zhou is an Assistant Researcher at Harbin Institute of Technology, Master's Supervisor (Mechanics). She has achieved a series of accomplishments in multibody dynamics numerical methods and explicit simulation, spacecraft multibody system dynamics modeling, multi-solver and multiphysics simulation. Published over 30 SCI journal papers, with more than 20 as first and corresponding author in authoritative journals such as IJMS, AST, JSV, ND, and MSD, including 2 highly cited papers. She holds 3 software copyrights and has filed 3 invention patents. Led and participated in 7 national, provincial, and enterprise-level projects. She have served as chair or invited speaker at numerous domestic and international conferences, and currently is an expert committee member for the Anhui Key Laboratory of Automotive Industry Software and the guest editor for the Multibody Dynamics Special Issue of Mathematics (SCI JCR Q1 journal).
Speech Title: Dynamic Modeling and Model Order-Reduction of On-Orbit Assembled Multi-Plate Structures
Abstract: On-orbit assembly is a pivotal enabling technology for advancing spacecraft modularity and autonomy. However, the accurate modeling and efficient simulation of on-orbit assembled structures face significant challenges, due to their evolving configurations, large degrees of freedom, and high interconnection complexity. This work proposes an efficient model order-reduction (MOR) method to enable fast and accurate simulation for space on-orbit assembled multi-plate structures. The dynamic model is developed using a reference nodal coordinate formulation, which captures the high-fidelity dynamics, albeit resulting in a high-dimensional system. A nonlinear modal reduction approach is further applied to generate a compact reduced-order model, incorporating global vibration modes and modal derivatives to preserve good accuracy. The key issue in MOR-optimal mode selection-is resolved through a novel kinetic-energy-based modal interaction method, which identifies dominant modes without prior full-order simulations, making it particularly suited for varying topologies. The proposed method is validated with octagonal and rectangular assemblies. Numerical results demonstrate that the method exhibits high accuracy, strong adaptability, and excellent efficiency. Notably, the computational cost is decreased for nearly 99%. The proposed method offers a promising technique for the dynamic analysis and control in space missions.
Biography: Tao Liu, Ph.D. from Tsinghua University, is an associate professor at the School of Civil Aviation, Northwestern Polytechnical University. He focuses on health monitoring and intelligent fault diagnosis of rotating machinery, intelligent agent interaction and trustworthiness evaluation, as well as large language model-based fault root-cause reasoning and explainable diagnosis. More than 30 papers have been published in leading domestic and international journals, including 19 first-authored papers and three corresponding-authored papers in internationally recognized SCI-indexed journals such as Advanced Engineering Informatics, Mechanical Systems and Signal Processing, Aerospace Science and Technology, and Science China Technological Sciences. More than 10 invention patents have also been obtained. Currently, he serves as a member of the Mechanical Dynamics Committee of the Chinese Society for Vibration Engineering and as a reviewer for renowned academic journals, including Advanced Engineering Informatics, Mechanical Systems and Signal Processing, IEEE Transactions on Automation Science and Engineering, IEEE Transactions on Industrial Electronics, Measurement, ISA Transactions, and Nondestructive Testing and Evaluation.
Speech Title: A Semantic Memory Compression and Data-driven Prototype-Consistency Reasoning Framework for Fault Diagnosis Based on Edge-Side Small Models
Abstract: Edge-side small language models face a key challenge in industrial fault diagnosis: heterogeneous sensor evidence must be fused without losing traceability, while long numerical streams can cause hallucinated evidence and unstable outputs. This paper proposes SMARF, a semantic memory compression and data-driven prototype-consistency reasoning framework that distributes information fusion across feature, temporal, and decision levels. At the feature level, an agent converts multimodal windows into schema-constrained Semantic Logs encoding operating conditions, normalized residuals, trends, candidate fault-response patterns, and evidence references. At the temporal level, a memory manager compresses the logs into window-, episode-, trajectory-, and prototype-level memories under a limited context budget. At the decision level, a reasoning agent re-ranks candidate root causes using semantic confidence, memory-retrieval relevance, and training-fold prototype consistency; feedback coordinates cross-time updates without online fine-tuning. On aviation piston pump and UCI Hydraulic Systems datasets, SMARF achieves 0.936 unseen-condition Macro-F1, improving over four baselines by 3.0–7.4 percentage points and reducing the seen-to-unseen drop from 0.111 to 0.047. It reduces the hallucinated-evidence rate from 0.216 to 0.018 and, with a 1.7B-parameter model, approaches 32B-level accuracy using 980 prompt tokens and 1.34 s latency.
Biography: Assistant Researcher at Shanghai Jiao Tong University, focus on computer vision-based vibration measurement and intelligent condition monitoring of key aerospace components. He has led several national-level research projects, including the Youth Program of the National Natural Science Foundation of China, the Special Grant and General Program of the China Postdoctoral Science Foundation, and sub-projects of the National Key R&D Program of China. He pioneered an adaptive phase-based full-field vibration measurement and successfully conducted full-field vibration measurement and dynamic analysis on the Mengtian Resource Module of the China Space Station, providing a new pathway for condition monitoring of large-scale flexible complex equipment. To date, he has published over 20 academic papers, obtained more than 10 invention patents and software copyrights, and received the Gold Award at the Innovation and Entrepreneurship Competition hosted by the National Innovation Centre.
Speech Title: Research on Adaptive Phase–Motion Estimation Method for Structural Full-Field Vibration Sensing
Abstract:Conventional vibration measurement techniques face inherent limitations in extreme-environment adaptability, non-contact operation, and full-field coverage. Existing vision-based methods, meanwhile, are constrained by target dependence, image noise, and illumination variations, making it difficult to simultaneously achieve high spatiotemporal resolution and measurement accuracy. This report systematically presents our breakthroughs in target-free, high-spatiotemporal-resolution visual sensing modal measurement technology—enabling us to "see" the invisible vibrations. The proposed method overcomes the bottlenecks of traditional contact-based measurements, including the added-mass effect, reliance on artificial targets, and degraded accuracy caused by complex noise and phase instability. We have developed a multi-scale fused phase vibration measurement theory and constructed an adaptive fusion strategy based on confidence measurement factors and amplitude responses, achieving high-fidelity phase extraction and modal parameter identification from heavily noisy images. The developed measurement instrument provides a technical solution for dynamic modal analysis of high-end equipment and major engineering projects.
Biography: Xinxing Ma, Ph.D., is an Associate Professor and Master Supervisor at the School of Mechatronic Engineering and Automation, Shanghai University, China. His research interests focus on nonlinear dynamics of rotating machinery, intelligent vibration control, advanced vibration mitigation technologies, and dynamic design of mechanical systems. Dr. Ma has been engaged in the investigation of nonlinear vibration mechanisms and control strategies for high-speed rotor systems, particularly spline-coupled transmission shaft systems, friction-induced self-excited vibrations, and active vibration suppression using intelligent supports. He has developed dynamic modeling and vibration control methods integrating passive structural optimization and active electromagnetic control technologies for advanced rotating machinery. He has led several research projects, including the National Natural Science Foundation of China, China Postdoctoral Science Foundation projects, and other national and provincial research programs. He has published more than 20 peer-reviewed papers in international journals, and holds several granted invention patents.
Speech Title: Vibration Control of Spline-Coupled Transmission Shaft Systems: Passive Suppression and Active Regulation
Abstract: Spline-coupled transmission shaft systems are widely used in aerospace propulsion systems and advanced mechanical equipment due to their high load-carrying capacity and compact structure. However, under high-speed and supercritical operating conditions, these systems are prone to complex vibration problems, including critical-speed resonance, nonlinear vibration, and friction-induced self-excited vibrations, which may significantly affect operational stability and reliability.
This invited talk presents recent developments in vibration control of spline-coupled transmission shaft systems, focusing on integrated passive and active vibration suppression strategies. A series of dynamic models are established to investigate the nonlinear vibration characteristics and control mechanisms of spline-coupled shaft systems. Passive vibration reduction approaches based on optimized spline structures and auxiliary supports are introduced to improve system stability and suppress vibration transmission. Furthermore, an active vibration control method using auxiliary electromagnetic bearings is presented. The electromagnetic force characteristics and equivalent stiffness/damping regulation mechanisms of magnetic bearings are analyzed, enabling adaptive modification of the rotor support characteristics under different operating conditions. The proposed control strategies are investigated for vibration suppression during critical-speed crossing and supercritical operation. Numerical analyses demonstrate that the combination of passive structural design and active electromagnetic control can effectively reduce resonance response and suppress friction-induced self-excited vibrations, providing a promising approach for improving the stability and reliability of high-speed transmission shaft systems.
Biography: Yaping Ju is currently a professor and doctoral advisor at Xi’an Jiaotong University, as well as the director of the Shaanxi International Joint Research Center for Fluid Machinery. He has long been engaged in research on the aerodynamic design of compressors and the quantification of uncertainty. He has led nine national-level research projects, including grants from the National Natural Science Foundation of China and basic research topics under the national “Two Engines” Special Project. As the first or corresponding author, he has published more than 60 academic papers, and his research findings have received sustained attention and positive citations from leading international research institutions in the fields of internal combustion engines and fluid machinery, such as Pratt & Whitney (U.S.), Rolls-Royce (U.K.), the German Aerospace Center (DLR), and ABB (Switzerland).
Speech Title: A Study on the Intelligent Optimization of Centrifugal Compressor Series Based on a Zero-Dimensional Prediction Model
Abstract: The serialization and standardization of centrifugal compressor stages are critical during the preliminary design phase, in which zero-dimensional design is responsible for planning the dimensionless design parameters of each master stage. However, existing research and engineering practices primarily rely on designers’ experience or traditional geometric progression rules to plan the distribution of master stages, resulting in either insufficient coverage of the aerodynamic performance range or excessive overlap. This, in turn, leads to repeated iterative adjustments between subsequent three-dimensional aerodynamic design and zero-dimensional design, significantly increasing the design cycle and computational costs. An optimal planning model based on an improved zero-dimensional aerodynamic prediction model is proposed. For the zero-dimensional aerodynamic prediction model, the prediction accuracy and generalization capability were significantly improved by modifying the peak efficiency correlation and introducing a radial basis function neural network to correlate the dimensionless design parameters. The optimization model aims to maximize the overall performance coverage and minimize the overlap between the performance spectra of adjacent stages; by combining a multi-objective optimization algorithm with the zero-dimensional prediction model, it achieves rapid optimization of the dimensionless design parameters for the master stage. This method has been validated in the zero-dimensional design of a family of centrifugal compressors with flow coefficients ranging from 0.03 to 0.11.
Biography: Yi Li, a lecturer at Ningxia University and the Technical Director of the Intelligent Process Equipment Center at Liupanshan Laboratory, has been selected for the Young Science and Technology Talent Cultivation Program of the Chinese Instrumentation Society. He mainly engages in research on robot force interaction and intelligent assembly, as well as assembly testing. Currently leading projects funded by the National Natural Science Foundation, sub projects of National Science and Technology Major Project, and provincial and municipal science and technology plans.
Speech Title: Research and Practice on Intelligent Assembly and Testing Technology for High Sealing Pressure Components
Abstract: High-integrity pressure-bearing components (such as valve components and tube fittings) are critical parts of aerospace systems such as liquid rocket engines and aircraft hydraulic and fuel systems, and their assembly quality directly affects the sealing reliability and in-service safety of these systems. The assembly of such components is characterized by small fitting clearances, strong force-position coupling, and stringent leak-tightness requirements, while the process remains highly dependent on manual expertise and suffers from low efficiency, poor consistency, and latent quality risks that are difficult to predict. This talk presents our research on intelligent assembly and inspection technologies for high-integrity pressure-bearing components, including multi-modal information fusion methods that integrate vision, force/position for assembly state perception, as well as compliant assembly control strategies based on contact state modeling. Engineering practices and lessons learned are further shared based on an in-house developed flexible intelligent assembly and inspection platform for small-and medium-sized ball valves, supporting mixed-model assembly and online inspection. Finally, future research directions in intelligent assembly process generation and domain-specific model collaborative regulation are discussed.
Biography: Dr Yiqian Mao is an Assistant Professor in the Department of Aerospace Propulsion, School of Astronautics, Beihang University, which he joined in 2025. He received
his BEng and MSc degrees from Central South University and his PhD in Aerospace Engineering from the University of Manchester in the UK in 2024. During his doctoral studies, he received First Place in the 2023 Osborne Reynolds Doctoral Prize for advances in active flow control using deep reinforcement learning. He was selected for a national postdoctoral research programme in China in 2025. His research interests encompass active flow control, reinforcement learning, generative methods and numerical modelling, with applications to bluff-body flows, combustion dynamics and aerospace propulsion.
Speech Title: Extending Deep Reinforcement Learning from Active Flow Control towards Flashback Control: Challenges and Prospects
Abstract: Deep reinforcement learning (DRL) has shown promise in active flow control, yet its application to combustion control is constrained by high computational cost, spatiotemporal partial observability, and the limited expressiveness of unimodal policies. This talk addresses these challenges through three canonical DRL-based circular-cylinder wake-control studies and examines their implications for combustion control. First, an action-informed, episode-based neural ordinary differential equation surrogate learns controlled low-dimensional dynamics for model-based DRL, attaining about 70% of the model-free DRL return with around 10% of the interaction samples and one third of the wall-clock time. Second, a delay-encoded autoregressive proximal policy optimisation (PPO) algorithm restores temporal context under post-action delay; at a Reynolds number of 400, it reduces fluctuations in the drag and lift forces by approximately 90% while maintaining comparable mean drag reduction. Third, a progressive Gaussian-diffusion policy couples a probability-flow diffusion actor with Laplacian maximum-a-posteriori (L-MAP) state reconstruction to represent complex action distributions under spatially incomplete observations. At a Reynolds number of 100, the diffusion actor achieves 8.5% drag reduction, 78% suppression of lift fluctuations and 66% lower mean control cost than PPO. With only 25% of the probes retained and the same L-MAP reconstruction used in both cases, the probability-flow diffusion actor improves the average reward by about 32% over Gaussian-based PPO. Furthermore, DRL is coupled with three-dimensional reacting-flow simulations to investigate flashback prediction and preventive control in a bluff-body swirl combustor operating with a premixed methane-air flame, thereby exploring the potential of DRL to address combustion-safety challenges in propulsion systems.
Biography: Yong Wang is an associate professor at the School of Aeronautics and Astronautics, Sun Yat-sen University. He earned his Ph.D. from the School of Aeronautics and Astronautics at Shanghai Jiao Tong University. He was a visiting scholar at the University of Central Florida and a postdoctoral fellow at the University of Ottawa. He previously worked at the Research Department of Huawei HiSilicon. His primary research interests include multi-source perception and navigation for unmanned aerial vehicles (UAVs) and counter-UAV technologies. His work has been published in over 60 papers in renowned journals and conferences, including IEEE Transactions on Intelligent Transportation Systems, IEEE Robotics and Automation Letters, Acta Automatica Sinica, Robotics, ICME, and ICASSP. He served as Area Chair for the 2023 PRCV conference. He is a member of the Pattern Recognition and Machine Intelligence Technical Committee of the Chinese Association of Automation, a member of the Popularization Committee of the Chinese Association of Automation, a member of the Embodied Intelligence Technical Committee of the Control and Command Society, and a member of the Low-Altitude Aviation Industry Working Committee of the Control and Command Society.
Speech Title: UAV-to-UAV Object Tracking
Abstract: This report systematically presents the latest advances in this field. First, we have successively constructed two datasets for tracking fixed-wing UAVs, providing crucial support for algorithm evaluation and deep learning training. Second, to provide more training and testing data, the RSTrack dataset—an RGBD tracking dataset comprising 209 sequences—is generated based on the AirSim simulation platform. Finally, to meet the requirements of fixed-wing UAV pose estimation, a PCA-enhanced SiamMask method is proposed based on the pixel-level segmentation and rotated bounding box annotations extended from the UAV2UAV dataset, significantly improving the rotated bounding box localization accuracy and segmentation quality. In summary, these works—spanning from dataset construction, multimodal fusion, and training strategy optimization to rotation-aware segmentation—have collectively advanced the development of UAV-to-UAV object tracking technology, laying a solid foundation and pointing the way for future research.
Biography: Yuxuan Shi is a Ph.D. candidate in Mechanical Engineering at Shanghai University. His research interests include large language models, knowledge engineering, and flexible assembly systems. He has participated in several national and provincial-level research projects, including a General Program of the National Natural Science Foundation of China, the “AI + High-End Equipment” Program of the Shanghai Municipal Commission of Economy and Informatization, and the Outstanding Academic/Technical Leader Program of the Shanghai Municipal Science and Technology Commission. He has published six SCI-indexed papers in journals including the International Journal of Production Research (IJPR) and Quality and Reliability Engineering International (QREI), and holds one authorized invention patent. He also participated in the development of a digital and intelligent DT-PLM system for civil aircraft assembly, which has been deployed on the C919 panel production line at COMAC Shanghai Aircraft Manufacturing Co., Ltd. He has received several honors, including the Fangzhou Self-Improvement Scholarship of Shanghai University and the “Best Performer” recognition at MEAE 2025.
Speech Title: Large Language Model-Based Multi-Agent Scheduling Decision-Making for Large Aircraft Assembly Islands
Abstract: To meet the requirements of mixed-model flexible manufacturing for large aircraft, the assembly-island paradigm enables parallel assembly of complex components and flexible resource allocation through the coordination of multiple assembly units. With dynamic task changes and the increasing integration of embodied robots, intelligent AGVs, and other heterogeneous resources, assembly-island scheduling has evolved from conventional optimization into a complex decision-making process involving process understanding, resource matching, task planning, and scheduling optimization. Existing methods remain limited in integrating cross-domain knowledge and adapting to dynamic manufacturing environments. To address these challenges, this study proposes a large language model-based multi-agent scheduling decision-making method for large aircraft assembly islands. A knowledge graph integrating process, resource, and skill information is first constructed to provide shared knowledge for the agents. Building on this knowledge base, a knowledge-guided LLM-based multi-agent architecture is developed, in which specialized agents collaborate on task understanding, resource matching, task planning, and scheduling optimization. The proposed method is validated in a civil aircraft large-section transfer and docking assembly island. The results demonstrate its potential to improve the autonomy and adaptability of scheduling decisions in complex and flexible assembly environments, providing a new technical approach to intelligent large aircraft assembly.
Biography: Zhenbo Lu is a professor and assistant dean at the School of Aeronautics and Astronautics, Sun Yat-sen University, and deputy director of the Ministry of Education Engineering Research Center for Low-Altitude Intelligent Flight Systems. In 2025, he was elected a Fellow of the Royal Aeronautical Society (RAeS). He earned his Ph.D. in Mechanical Engineering from the Hong Kong Polytechnic University and spent several years conducting research at the Temasek Laboratory at the National University of Singapore, where he held positions as a researcher and senior researcher; he currently serves as a part-time senior researcher there. He has long been engaged in research in the fields of low-altitude aircraft technology, aerodynamic design, aerodynamic noise, bionic aircraft, and smart materials and structures, and has led more than ten projects of various types. He has published more than 80 high-impact academic papers, filed more than 30 invention patent applications, and received the Second Prize of the Guangdong Provincial Science and Technology Progress Award. He currently serves as an editorial board member or young editorial board member for the journals "Journal of Aeronautics", "Journal of Mechanics", and "Aerospace Technology"; as a member of the Aerodynamic Acoustics Technical Committee of the Chinese Society of Aerodynamics; as a member of the Shenzhen Low-Altitude Economy Standardization Technical Committee; and as a technical advisor to the Beijing Civil Aircraft Technology Research Center of COMAC.
Speech Title: Research on Aerodynamic Design and Low-Noise Control Technologies for Low-Altitude Aircraft
Abstract: This study conducts a systematic investigation into the aerodynamic design and low-noise control technologies for low-altitude aircraft, with a particular focus on electric vertical takeoff and landing (eVTOL) aircraft. In terms of aerodynamic configuration, the research team drew inspiration from biomimicry—specifically from flying creatures such as dragonflies, swifts, and owls—to develop high-efficiency, high-lift airfoils, tandem-wing configurations, and distributed tilting ducted fan systems, thereby significantly enhancing the aircraft’s aerodynamic efficiency and flight stability. To address the issue of unsteady aerodynamic disturbances during tilt transition, the team established parametric modeling and high-precision numerical simulation methods, systematically revealing the strong coupling mechanism among propulsion, aerodynamics, and noise. In terms of low-noise control, drawing on the silent flight mechanisms of owls, we innovatively designed leading-edge serrations, trailing-edge fringes, and biomimetic propeller structures; experiments demonstrate that these can achieve noise reduction of up to 18 dB under typical operating conditions. The research further integrated artificial intelligence with multi-objective optimization algorithms to construct a forward-design closed-loop system comprising “parametric modeling—numerical evaluation—intelligent optimization,” effectively balancing the trade-off between lift-to-drag ratio and acoustic performance.
Biography: Dr. Zhijie Zhao is a lecturer and master’s supervisor at national university of defense technology, and a provincial‑level young talent. His research focuses on intelligent flow control for flight vehicles. He has led 10 projects, including general program and joint fund sub‑projects of national natural science foundation of China. His honors include national aerospace doctoral competition grand prize, the first prize in technical innovation from Chinese society of aerodynamics (5/15), major scientific and technological progress award in Chinese aviation (3/15), AVIC golden idea grand prize, and three consecutive first prizes of China future aircraft design competition (ranked first each time). He has published over 20 SCI papers as first/corresponding author, including a best cover and highly cited paper in the top journal, and has authored one monograph. He holds 10 authorized invention patents. He currently serves as a youth committee member of the flow control and thermal management branch of the Chinese society of aeronautics and astronautics.
Speech Title: Dual-synthetic-jet-based Intelligent Flow Control Technology for Aircraft
Abstract: Active flow control using dual synthetic jets (DSJ) is characterized by high dimensionality, nonlinearity, and time delays in actuation and sensing, which make it difficult to establish an accurate dynamic model. Even if such a model is established, deriving a high‑performance nonlinear feedback control policy via Lyapunov stability theory and optimal control allocation remains challenging. Artificial intelligence offers a new opportunity to address these issues and is expected to overcome difficulties in modeling and controlling DSJ-based active flow control. Following a research roadmap from open‑loop parameter optimization, through closed‑loop PID control, to explicit/implicit policy optimization, authors have progressively developed a series of integrated control methods for DSJ-based active flow control, including particle swarm optimization‑based parameter control, convolutional neural network‑based flow field recognition control, PID self‑tuning control, linear genetic programming for explicit policy optimization, and deep reinforcement learning for implicit policy optimization. These methods have facilitated the emergence of new high‑efficiency flow control strategies and mechanisms. In addition, authors have built the first cloud platform for parallel intelligent flow control with hundreds of environments, which achieves the highest computational training efficiency reported in the literature. A typical application is the use of deep reinforcement learning for synergistic lift enhancement on a control surface with DSJ. The learning algorithm discovered a novel collaborative pulsed control strategy for distributed DSJ, revealing a new flow evolution mechanism in which flow separation control and the virtual Gurney flap effect produce a “1+1>2” lift augmentation. Compared with open‑loop control, the operational envelope is expanded by 16.8%, energy consumption is reduced by 52%, and the efficiency‑to‑cost ratio is increased by 146%.
© MEAE 2018-2026. All Rights Reserved | Contact Us meae_conf@163.com
