Abstract

Advanced manufacturing processes involve material microstructural evolution, complex interfacial behavior, and multiphysics coupling, resulting in significant spatial and temporal cross-scale relationships between microscopic manufacturing mechanisms and macroscopic in-service performance. Although conventional high-fidelity numerical simulations offer strong physical interpretability, they are often constrained by high computational costs and limited real-time predictive capabilities. This presentation focuses on an integrated framework combining multiphysics simulations, data-driven modelling, and digital twins. It briefly introduces multiscale, multiphysics, and multiphase numerical modelling methods for practical engineering systems, with particular emphasis on integrating artificial intelligence to enable the rapid prediction and real-time reconstruction of complex physical fields.  Representative applications in additive manufacturing, energy devices, and high-power-density systems will also be presented, demonstrating the engineering value and future potential of these methods for the optimal design, state prediction, and intelligent control of complex engineering systems.

Abstract

Thin-walled structures, characterized by weak stiffness and low damping, are prone to significant vibrations during service and are typical in aerospace equipment. Embedding smart materials—such as piezoelectric, magnetoelectric, and shape memory alloys—into passive matrices enables smart structures with integrated sensing and actuation capabilities, offering great potential for shape control, vibration suppression, and health monitoring. However, these structures involve complex stress distributions from multi-material lamination, large deformations, and strong multi-physics couplings, posing severe challenges to high-fidelity modeling and efficient computation. To address these issues, this report presents a series of research efforts: (1) plate/shell theories, evolving from classical assumptions to higher-order shear deformation and further to a dual-deflection higher-order theory; (2) nonlinear modeling, unifying von Kármán to large-rotation nonlinearities and incorporating material nonlinearity for combined geometric-material solutions; (3) multi-field coupling mechanisms, establishing a generalized constitutive framework from electro-elastic to fully coupled magneto-electro-thermo-elastic formulations. Additionally, the report explores AI-assisted mechanical modeling, with preliminary applications in accelerating flexible spacecraft dynamics analysis and topology optimization of machine tool structures, aiming to provide new computational strategies for efficient smart structure design and analysis.

Abstract

Plasma electrolytic oxidation (PEO) coatings have been widely developed to regulate the degradation behaviour of biomedical magnesium alloy implants in physiological environments. However, accurately quantifying the stress corrosion cracking (SCC) behaviour of PEO-coated magnesium alloys under both in vitro and in vivo conditions remains a significant challenge. To better understand and optimize implant performance under complex chemo-mechanical loading conditions, this presentation introduces a multiphysics phase-field framework for modelling the interactions among corrosion, fracture, and hydrogen diffusion during SCC. The proposed computational model employs the displacement field, fracture field, corrosion field, magnesium concentration field, and hydrogen concentration field as coupled primary variables. Numerical examples are presented to investigate hydrogen embrittlement, validate the model against experimental observations of PEO-coated WE43MEO magnesium alloys, and evaluate the effectiveness of post-sealing treatments. The framework provides an effective computational tool for predicting SCC susceptibility and supporting the design of next-generation biodegradable implants.

Abstract Title: To meet the multi-objective design requirements of complex-structure preforms and to enhance the manufacturing efficiency and quality, this study develops a digital-twin-based intelligent manufacturing management system for specialized complex-structure preform equipment, enabling visual preform design, virtual manufacturing simulation, and multi-system integrated control. A multi-medium yarn tension transmission simulation module was developed: the Capstan method was used to measure the tension variations of carbon fiber under different spatial angles, wrap angles, guide roller radii, and friction coefficients. A unified theoretical model for tension prediction, which incorporates spatial angle effects and a power-law friction relationship, was proposed for the first time, systematically revealing the “tension–path–angle–medium” correlation among yarns. Furthermore, a physics-informed neural network-based surrogate model for yarn tension prediction was constructed, achieving high-precision prediction of yarn tension under complex operating conditions. This work provides a theoretical foundation for the design of large-capacity yarn tension control systems for three-dimensional woven preforms.

Abstract

Industrial air-knife systems involve complex three-dimensional turbulent flows, where local upward airflow, recirculation, and vortex structures may affect manufacturing quality. However, high-fidelity CFD simulations are computationally expensive and difficult to use for rapid design exploration. This presentation introduces a data-efficient surrogate modelling framework for full-field reconstruction and optimization of an industrial air-knife system. A fixed-grid Fluent model is combined with a 150-case mixed-variable design of experiments covering two air-knife inlet mass flow rates and different side-suction configurations. Reduced-order modelling is then employed to compress the three-dimensional velocity and pressure fields, while surrogate models learn the mapping from operating conditions to the reduced flow representation. Preliminary CFD results indicate that appropriately configured side suction can suppress local upward airflow and recirculation. The framework provides a computational basis for rapid flow-field prediction, engineering optimization, and future digital-twin applications.