About

Professor Yumeng Li grew up in a small town in the Hunan province of China. She earned her bachelor’s and master’s degrees in Naval Architecture and Ocean Engineering from Huazong University of Science and Technology in Wuhan, China. For her PhD, she studied Aerospace Engineering at Virginia Tech, focusing on nanocomposites, which are microstructures made of multiple types of materials with significant potential to improve material properties. Her PhD work involved developing and analyzing nanocomposites with high stiffness-to-weight ratios through multiscale simulations emphasizing interface effects, with broad applications in aerospace, food packaging, and the medical industry. During her postdoctoral research at Vanderbilt University, she developed a multifunctional simulation framework to study fatigue creep phenomena in alloys, aiming to better predict alloy lifetimes under service conditions. She subsequently worked in the Department of Mechanical Engineering at Wichita State University before joining the University of Illinois Urbana-Champaign's Industrial & Enterprise Systems Engineering department. Her research field is in computational material science, specifically multiscale simulation of multifunctional materials and design. She plans to develop courses related to multiscale simulation and continues her research in this area.

Research topics

• Chemistry
• Nanotechnology
• Materials science
• Photochemistry
• Medicine
• Biomedical engineering
• Inorganic chemistry
• Cancer research
• Biochemistry
• Chemical engineering
• Physical chemistry

Selected publications

• Nonlinear FSI effects of a 3D composite layup model of ultra-long flexible wind turbine blade under typical typhoon conditions

  Engineering Structures · 2026-05-05

  article
  PublisherDOI

• Defect‐Functionalization‐Mediated Tunneling Drives Nonlinear Photoemission in Perovskites

  Advanced Science · 2026-04-09

  articleOpen access

  films act as functional mediators to govern carrier transport, interfacial electron escape, and quantum efficiency multiplication via a defect-mediated tunneling process. First, in solid-state photonics, it overcomes strong exciton binding in an Au/CsPbBr3/Au transistor, yielding a >70-fold photoconductive gain at sub-nA dark current and a spectral response extension of >0.88 eV. Second, in vacuum electronics, electron beam bombardment creates localized electropositive centers, forming a defect-functionalized activation layer that induces pronounced band bending and shifts interface dominance from luminescence to photoemission, offering a robust alternative to chemical methods. Thirdly, we demonstrate transient quantum efficiency multiplication that reaches 16.8% under 266 nm pulsed excitation and achieves a 34 dB gain at 355 nm, functionally establishing a photoemissive comparator. By transforming the conventional role of defects from detrimental to functional, this work represents "defect-functionalized optoelectronics" as an alternative design route for advanced nonlinear device function.

  PublisherDOI

• An Asymmetric Zn Membrane with Degradability, Antimicrobial, and Bone Immunomodulation for Guided Bone Regeneration

  ACS Applied Materials & Interfaces · 2025-05-16 · 4 citations

  article

  As a biodegradable metal, zinc (Zn) offers a promising material option for barrier membranes in the field of bone regeneration due to its suitable degradation rate and mechanical properties. An ideal barrier membrane not only blocks the growth of epithelial fibers but also promotes bone regeneration. Therefore, we prepared an asymmetric Zn membrane with micrometer-sized pores on one side by laser etching, with the pore side facilitating cell adhesion and proliferation, and the smooth side facilitating the blocking of epithelial cell growth entry. And the antimicrobial peptide GL13K (P1) was loaded onto the smooth surface of Zn by Zn-specific binding peptide (NCS) to resist postoperative bacterial infections, and the small intestinal submucosal hydrogel complex (SIS-P2), which was specifically loaded with Substance P (SP), was placed in the pores on the pore side to modulate the immunity and promote osteogenesis. This innovative asymmetric zinc barrier membrane (Zn-P1-SIS-P2 membrane) exhibited excellent mechanical, antimicrobial, and biocompatibility properties. More importantly, the Zn-P1-SIS-P2 membrane promotes macrophage polarization toward the M2 type, thereby promoting osteogenic differentiation and providing a good immune microenvironment for bone regeneration. In addition, the Zn-P1-SIS-P2 membrane inhibited RANKL-induced osteoclast formation and suppressed bone resorption at the site of bone defects. In conclusion, the Zn-P1-SIS-P2 membrane demonstrated all the desirable qualities of a GBR therapeutic barrier membrane.

  PublisherDOI

• AI-driven predictions of electrochemical CO2 reduction catalysts: Insights from in situ spectroscopy

  International Journal of Hydrogen Energy · 2025-06-03 · 1 citations

  article1st author
  PublisherDOI

• Fusing Imbalanced Data via Physical Condition-Aware Surrogate Modeling

  2025-01-03

  articleSenior author

  Predicting material responses under diverse loading and boundary conditions is pivotal for understanding structure-property relationships and guiding material design. Traditional methods, such as physics-based simulations, often involve high computational costs, while experimental exploration across vast design spaces is impractical and resource-intensive. Convolutional Neural Networks (CNNs) have emerged as efficient tools for material response prediction, particularly for materials with intricate microstructures. However, these models face challenges stemming from imbalanced datasets, where critical material attributes such as displacement fields are less accessible and more expensive to obtain than others like strain energy. This imbalance can skew predictions and hinder the generalization of models to unseen material structures. To address these challenges, we propose a novel framework that integrates multi-task learning (MTL) with a Physical Conditions Informed Convolutional (PCIConv) layer to improve prediction accuracy and robustness under imbalanced data conditions. MTL enables the simultaneous prediction of multiple material responses, leveraging shared information across tasks to enhance model generalization. The inclusion of the PCIConv layer embeds physical insights into the learning process, further enhancing the model's adaptability to diverse conditions. We validate our approach using the Material MNIST dataset, demonstrating its ability to predict displacement fields and strain energy across various imbalance scenarios. Results show that the MTL framework, particularly when enhanced with PCIConv, outperforms single-task models by achieving superior accuracy and robustness, even with limited data for underrepresented attributes. This study highlights the potential of combining MTL and PCIConv in advancing material response predictions, offering a pathway to efficient and cost-effective material design while reducing reliance on computationally expensive simulations and experimental methods.

  PublisherDOI

• Physics-informed multitask learning for microstructure-aware material metamodels of heterogeneous materials

  Computational Materials Science · 2025-08-24 · 1 citations

  articleOpen accessSenior authorCorresponding

  Constructing high-fidelity structure–property predictive models is crucial for investigating, developing and designing novel materials to enable technical innovations. Traditional approaches relying on physics-based simulations and experimental testing are often costly and time consuming, particularly for heterogeneous materials with complex microstructures. In this work, we present a multitask learning (MTL) framework to address two major challenges in data-driven modeling of such materials: (1) effective representation of microstructure features and (2) integration of multisource material properties with varying dimensionalities. Our approach leverages easily obtainable strain energy data to improve the prediction of complex displacement fields, thereby improving generalization and efficiency with limited training samples. This method can effectively fuse available limited data and significantly reduce the cost of generating a training dataset. In addition, the selection of input features have been investigated using different machine learning models on two types of heterogeneous material systems, the Material MNIST dataset and the two-phase fiber-reinforced composite material dataset, as the case study. The results of the case study show that the proposed model is effective and efficient in establishing predictive property models of heterogeneous material systems with complex microstructure. • Microstructure-aware material metamodels are developed to accurately predict the properties of complex heterogeneous materials through fusing multi-source data. • Multitask learning has been demonstrated to effectively fuse multi-source data, which significantly improve the accuracy, transferability and efficiency of the developed predictive models. • The selection of the microstructure encoder and property predictor is highly related to the characteristics of microstructure.

  PublisherDOI

• Transportation cost-information inequality for non-linear time-fractional stochastic heat equation driven by space–time white noise

  Statistics & Probability Letters · 2025-08-12

  articleOpen accessSenior authorCorresponding

  We establish transportation cost-information inequalities T 2 ( C ) for solutions of nonlinear stochastic partial differential equation of fractional order in both space and time variables with deterministic and bounded initial conditions: ∂ t β u ( t , x ) + ( − Δ ) α / 2 u ( t , x ) = I t γ σ ( u ( t , x ) ) W ̇ ( t , x ) in ( 0 , ∞ ) × R d , where α > 0 , β ∈ ( 0 , 2 ] , γ ≥ 0 , ∂ t β is the Caputo fractional derivative, − ( − Δ ) α / 2 is the fractional/power of Laplacian, I t γ is the Riemann–Liouville integral operator, W ̇ ( t , x ) is a space–time white noise, and σ : R → R is a bounded and Lipschitz function. Since the space variable is defined on the unbounded domain R d , the inequalities are proved under a weighted L 2 -norm in the spatial domain.

  PublisherDOI

• Dynamic characteristics of scraper conveyor with direct-driven permanent magnet synchronous motor under impact load

  Proceedings of the Institution of Mechanical Engineers Part E Journal of Process Mechanical Engineering · 2025-03-04

  articleOpen access

  To simulate the dynamic characteristics of a permanent magnet direct-drive scraper conveyor (SC) (PMDD-SC) under impact load, a dynamics model is constructed based on the Kelvin–Voigt theory through AMESim, and the vector control of permanent magnet synchronous motor (PMSM) is carried out with id = 0 through MATLAB/Simulink. Based on the dynamic equations of the SC and the torque equations of the PMSM, the torque expression of the motor was derived and used as the system output. At the same time, the sprocket speed was used as the system's feedback, and the electromechanical coupling model of PMDD-SC was established through the co-simulation of the two software. The dynamic characteristics of PMDD-SC under unsteady coal fall, stuck chain, and broken chain were analyzed. Specifically, the tension and speed of the chain and the key point, as well as the torque, speed, and U-phase current of the motor at the head and tail of the machine were simulated at different positions and times under the above three working conditions. The results show that the maximum tension on the loaded side is 2.79 × 10 5 N and 8.35 × 10 5 N for the unsteady coal fall and the stuck chain, respectively. The peak currents of the unsteady coal fall and the blocking current of the stuck chain were about 340 A and 700 A, respectively. The chain tension will be 0 rapidly after chain breakage, and the head current fluctuates drastically and then tends to stabilize. The characteristics of SC tension and speed are the same. The permanent magnet direct-drive system responds quickly to the load change, and the larger load change has a greater impact on the stability of the drive system.

  PublisherOA PDFDOI

• Life Cycle Assessment of Direct Recycling for Cathode Active Materials

  2025-06-18

  article

  As transportation becomes further electrified, lithium-ion batteries are only increased in demand. Thus, to maintain the growth of transportation electrification, recycling of batteries is of great significance. Cathode active materials are the most valuable components in batteries and direct recycling has arisen as an potentially valuable recycling method for recovering cathode active materials. However, there is no comprehensive comparison of direct recycling methods to guide industrial applications. This research uses life cycle assessment to evaluate the environmental impacts of different direct recycling methods for cathode active materials based on the ReCiPe method and standardized processes. As a result, the most environmentally friendly direct recycling method for each cathode active material was found, which can help the further development of transportation electrification.

  PublisherDOI

• Patterned 3D-printed hydrogel as a novel soilless substrate for plant cultivation

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-08-07

  preprintOpen access

  Plant roots need water, micronutrients, and oxygen to maintain cellular metabolism and tissue growth, yet traditional hydroponic systems often lack sufficient oxygen delivery. While 3D printing artificial substrates has been explored to mimic the physical structure of soil, it remains unclear which design parameters are critical for supporting full plant development. Here, we present a synthetic, soilless substrate based on 3D-printed hydrogels incorporating triply periodic minimal surface (TPMS) patterns to create internal air-filled channels. These channels are open to the atmosphere, enabling passive gas exchange throughout the substrate. We tested five TPMS geometries (Lidinoid, Split-P, Schwarz-D, Schwarz-P, and Schoen), each with an identical hydrogel volume but with different surface-to-volume ratios. Arabidopsis thaliana seeds germinated directly on the substrates and were monitored for vegetative and reproductive growth over five weeks. Among the designs, the Lidinoid substrate consistently led to the highest number and surface area of leaves and the earliest and most complete flowering, outperforming both hydroponics and unpatterned hydrogel controls. Our results indicate that the surface-to-volume ratio is a key parameter influencing substrate performance, likely due to its impact on oxygen availability at the root interface. Plants grown on substrates with higher surface areas transitioned to flowering more reliably and rapidly, with flowering efficiency showing a strong positive correlation with surface area. These findings suggest that interconnected air-channel architectures can overcome the oxygen limitations of traditional hydroponic systems without requiring active aeration. This work supports the use of additive manufacturing as a powerful tool for engineering soil analogues tailored for indoor agriculture. By combining passive aeration with hydration and nutrient delivery, patterned hydrogels offer a promising, scalable solution for sustainable soilless plant cultivation.

  PublisherOA PDFDOI

Frequent coauthors

• Pingfeng Wang

  University of Illinois Urbana-Champaign

  30 shared

• Zheng Liu

  University of Illinois Urbana-Champaign

  19 shared

• Zhuoyuan Zheng

  17 shared

• Renbin Ge

  Shanghai First People's Hospital

  16 shared

• Zhongling Wang

  Shanghai First People's Hospital

  16 shared

• Parth Bansal

  University of Illinois Urbana-Champaign

  16 shared

• Rong Cao

  16 shared

• Akash Singh

  Manipal Hospital

  14 shared

Labs

• Industrial & Enterprise Systems EngineeringPI