Shu Li

Traditional density-based clustering algorithms are sensitive to global parameter settings and struggle with datasets containing clusters of varying densities and complex shapes. In this paper, we propose O-CDBC (Optimized Connection Density-Based Clustering), a novel and efficient density-connection clustering algorithm. The core idea of O-CDBC is to introduce a compute connection density based on shared neighbors to capture complex structures, and design a data-driven mechanism for automatic outlier detection and cluster scale determination. The main optimizations include: dynamically determining the outlier ratio and minimum cluster size threshold without manual tuning. Experimental results demonstrate that O-CDBC significantly improves computational efficiency and parameter robustness while preserving strong capability in identifying clusters with complex shapes.

Zinc-based flow batteries (ZFBs) are promising candidates for large-scale renewable energy storage, yet zinc dendrite growth severely limits their long-cycle stability and practical deployment. To address this critical challenge, this study employs a nonlinear phase-field model to systematically investigate the dynamic evolution of zinc dendrites in the porous electrodes of ZFBs, focusing on the regulatory effects of electrolyte flow velocity, electrochemical reaction constant (Lη, characterizing exchange current density), and carbon fiber arrangement on dendrite morphology, deposition efficiency, and growth kinetics. The simulation results show that dendrite growth in porous electrodes undergoes three stages: initial uniform deposition, middle-stage anisotropic bridging, and late-stage continuous network formation. Lower electrochemical reaction constant yields regular dendrites but slow charging, which can be improved by optimizing electrolyte flow velocity; higher electrochemical reaction constant accelerates charging but induces disordered dendrites, which is mitigated via pulse charging. Disordered carbon fiber arrangement exacerbates flow/concentration inhomogeneity and dendrite abnormal growth, while uniform arrangement suppresses disordered growth. This work fills the gap in phase-field simulation of dendrites in porous electrodes and provides targeted regulatory strategies, offering theoretical support for enhancing the long-cycle safety and stability of ZFBs in large-scale energy storage applications.

This paper reviews the research progress and application prospects of machine learning technologies in the field of polymer materials. Currently, machine learning methods are developing rapidly in polymer material research; although they have significantly accelerated material prediction and design, their complexity has also caused difficulties in understanding and application for researchers in traditional fields. In response to the above issues, this paper first analyzes the inherent challenges in the research and development of polymer materials, including structural complexity and the limitations of traditional trial-and-error methods. To address these problems, it focuses on introducing key basic technologies such as molecular descriptors and feature representation, data standardization and cleaning, and records a number of high-quality polymer databases. Subsequently, it elaborates on the key role of machine learning in polymer property prediction and material design, covering the specific applications of algorithms such as traditional machine learning, deep learning, and transfer learning; further, it deeply expounds on data-driven design strategies, such as reverse design, high-throughput virtual screening, and multi-objective optimization. The paper also systematically introduces the complete process of constructing high-reliability machine learning models and summarizes effective experimental verification, model evaluation, and optimization methods. Finally, it summarizes the current technical challenges in research, such as data quality and model generalization ability, and looks forward to future development trends including multi-scale modeling, physics—informed machine learning, standardized data sharing, and interpretable machine learning.

The exploration of high-performance materials presents a fundamental challenge in materials science, particularly in predicting properties for materials beyond the known range of target property values (extrapolation). This study formally investigated the interpolation-extrapolation trade-off phenomenon in the prediction capabilities of machine learning (ML) models. A new ML scheme was proposed, featuring a newly developed ML model and forward cross-validation-based hyperparameter optimization, which demonstrated superior extrapolation prediction across multiple materials datasets. Based on this ML scheme, multi-objective optimization was performed to systematically identify lightweight Mg-Zn-Al alloys with both high bulk modulus and high Debye temperature. Subsequently, the designed alloys were validated through density functional theory calculations. Furthermore, a three-category classification strategy was summarized through the dual-driven approach combining domain knowledge and data, emphasizing their synergistic potential for materials discovery. The practical framework developed in this study provides a novel research perspective for exploring high-performance materials.

A dual-driven ensemble framework for battery state of health (SOH) prediction is proposed to address the challenges of weak interpretability and insufficient physical mechanism correlation of existing methods. First, two key health features (HFs) containing physical information are extracted based on the first-order resistance–capacitance (RC) equivalent circuit model and the incremental capacity (IC) curve. Secondly, combined with traditional statistical features, a multivariate degradation representation collection is constructed to enhance the ability to characterize HFs. Then, by means of ensemble strategies, three basic models - kernel ridge regression (KRR), random forest (RF) and explainable boosting machine (EBM) are integrated, thereby significantly improving prediction efficiency, robustness and applicability capabilities. Finally, the feature ablation experiment is employed to verify the effectiveness of the proposed multivariate HFs system, and the interpretability of the framework is enhanced with the assistance of SHAP analysis. Experimental results on two public data sets reveal that, in comparison with several conventional deep learning models, the proposed framework maintains or even improves the prediction accuracy at the model inference level while increasing the processing speed by at least 60%.

Lithium-ion batteries, as a critical pillar in the field of new energy, face challenges in further development because of the lithium dendrite issue at the anode. This study established a phase-field model to investigate the effects of charging protocols and external pressure on dendrite growth under static electrolyte conditions. Additionally, the study emphasizes the influence of electrolyte flow, with a specific focus on investigating the interplay among charging protocols, external pressure, and electrolyte flow rates in modulating dendrite growth within a dynamic electrolyte environment. Results demonstrate that pulsed-voltage charging significantly suppresses dendrite formation, with enhanced inhibition observed at lower duty cycles. External pressure exhibits a positive correlation with dendrite suppression up to a critical threshold. Notably, the synergistic combination of pulsed-voltage charging, optimized external pressure, and directional electrolyte flow achieves the most effective suppression of dendritic growth.

Lithium metal batteries are considered promising candidates for next-generation high-energy-density storage systems, owing to their high theoretical capacity and low electrochemical potential. However, the formation of lithium dendrites during cycling significantly compromises battery efficiency and safety. To address this challenge, structural design and optimization of lithium metal anodes have become a research hotspot. In this study, a nonlinear phase-field model was utilized to examine the dendrite growth behavior in anodes with various patterned structures. Particular emphasis was placed on both analyzing the deposition characteristics of the sawtooth anode and conducting a comparative evaluation with the conventional bare lithium anode. Results reveal that the sawtooth anode significantly enhances lithium deposition uniformity and effectively suppresses disordered dendrite growth, owing to its large specific surface area and uniformly distributed nucleation sites. Moreover, the sawtooth anode maintains excellent deposition uniformity even under high noise and exchange current density conditions, showcasing superior electrochemical performance. The study further examined the impact of surface pits on dendrite growth behavior in both types of anodes. These findings provide new theoretical insights and design guidelines for the structural optimization of lithium metal battery anodes.

The growth velocity of the crystal–melt interface during solidification is one of the important parameters that determine the crystal growth morphology. However, both experimental investigations and theoretical calculations are time-consuming and labor-intensive. Moreover, machine learning (ML)-based methods are severely limited by the limited amount of available experimental data. In this work, the crystal–melt interface velocity of four alloy systems under different values of undercooling was calculated by molecular dynamics simulation. The results showed a similar trend to the experimental data. A framework including molecular dynamics (MD) calculation and a transfer learning (TL) model was proposed to predict the interface velocity of binary alloys during free solidification. In order to verify the effectiveness of the model, eight ML models were constructed based on pure experimental data for model comparison. The prediction ability of the different models was assessed from two perspectives: interpolation and extrapolation. The results show that, regardless of whether it is interpolation or extrapolation, the TL model driven by both physical information and experimental data is superior to ML models driven solely by experimental data. The interpretability analysis method reveals the specific role of feature values in the interface velocity prediction of binary alloys.