Institute of Automation

In this study, supervised machine learning (ML) regression models are employed to predict water uptake during the sorption process in a sorption reactor for thermal energy storage applications. Two main methods are used to study sorption storage systems: experimental studies and numerical simulations. Experimental studies involve physical testing and measurements but are often costly and time-consuming. Numerical simulations are more flexible and cost-effective, though they can require significant computational resources for large or complex systems. To address these challenges, researchers are increasingly employing various machine learning techniques, which offer strong potential for data analysis and predictive modeling. In this study, CFD-based sorption simulations are integrated with machine learning models to predict the spatiotemporal evolution of water uptake. Several ML techniques including support vector regression (SVR), Random Forest, XGBoost, CatBoost (gradient boosting decision trees), and multilayer perceptron neural networks (MLPs) are evaluated and compared. A fixed-bed reactor equipped with fins and tubes is considered within a closed adsorption thermal storage system. Numerical simulations are conducted for three different fin lengths (10 mm, 25 mm, and 35 mm) to generate a comprehensive dataset for training the ML models and capturing the complex temporal evolution of water uptake, thereby enabling predictions for unseen fin geometries. The results indicate that neural network-based models achieve superior predictive performance compared to the other methods. For water uptake training, the mean absolute error (MAE), root mean squared error (RMSE), and coefficient of determination R2 are approximately 2.83, 4.37, and 0.91, respectively. The predicted water uptake shows close agreement with the numerical simulation results. For the prediction cases, the MAE, MSE, and R2 values are approximately 1.13, 1.2, and 0.8, respectively. Overall, the study demonstrates that machine learning models can accurately predict water uptake beyond the training dataset, indicating strong generalization capability and significant potential for improving thermal management system design. Additionally, the proposed approach reduces simulation time and computational cost while providing an efficient and reliable framework for modeling complex sorption processes in thermal energy storage systems.

ABSTRACT Carbon fiber reinforced polymer (CFRP) composite has undoubtedly revolutionized advanced engineering applications due to its lightweight and high strength. In this study, self‐made zirconia fiber (ZF) via electrospinning was introduced into the interlayer of laminated CFRP composites to construct the fiber bridging to prevent crack generation and propagation, which could improve weak regions of resin‐rich region (RRR) and interfacial transition region (ITR). Low‐velocity impact results showed ZF‐implanted CFRP composites exhibited better impact resistance than unreinforced CFRP composites. Three‐point bending testing results demonstrate that the flexural strength of CFRP composites with 0.25 wt.% ZF before and after impact yielded the greatest 26.3% and 133.4% increments respectively compared with unreinforced ones, and the former had 55% and 121.5% higher energy absorption than that of the control group in pre impact and post‐impact damaged energy absorption respectively. XRM‐CT and SEM scanning results indicated the trace of fiber bridging of ZF on the damaged surface and its effects on shifting delamination failure of unreinforced CFRP composite to shear failure of ZF‐implanted CFRP composites. Overall, introducing ZF into the interlayer might be an easy and effective method for manufacturing high‐performance CFRP composites in industries.

Introduction: The chemical analysis of QizhenShengbai Compound (QZSBC) liquid was conducted using ultra-high pressure liquid chromatography tandem quadrupole time-offlight mass spectrometry (UPLC-Q-TOF-MS/MS) to investigate the material basis of QZSBC. Additionally, network pharmacology and molecular docking were employed to predict the potential mechanisms underlying its anti-radiation effects on the hematopoietic system. Methods: The UPLC-Q-TOF-MS/MS method was employed to analyze the chemical components of QZSBC. Additionally, a network pharmacology approach was used to construct a 'drugactive ingredient-target-disease' topological network. A protein-protein interaction (PPI) network was constructed for the disease targets to identify key proteins. Furthermore, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) functional enrichment analyses were performed on these key targets, and the network was validated through molecular docking. Results: GO enrichment analysis revealed significant enrichment in biological processes, such as DNA damage repair, cell cycle regulation, and inflammatory response; KEGG analysis indicated high association with pathways including JAK-STAT, PI3K-Akt, and cell cycle. Molecular docking demonstrated that the active components in QZSBC exhibited strong binding affinity with the core targets. Discussion: Among the 10 core proteins screened, the HSP90 family and SRC were identified as critical nodes in radiation-induced DNA damage repair; the JAK-STAT pathway, mediated by STAT3, plays a central role in the survival and differentiation of hematopoietic stem/progenitor cells, which is highly consistent with the enrichment results. Among the active components, quercetin, luteolin, and others were found to alleviate radiation damage through antioxidant effects and the promotion of DNA repair, while amide compounds, such as Aurantiamide acetate, were reported to bind to HSP90, suggesting a novel mechanism of action. Conclusion: QZSBC may exert its radioprotective effects on the hematopoietic system by acting on key targets, such as HSP90AA1, SRC, and STAT3, through active ingredients, such as Aurantiamide acetate, Licarin A, and Quercetin, thereby interfering with signaling pathways, including DNA damage repair, JAK-STAT, and PI3K-Akt.

Answering complex medical questions requires not only domain expertise and patient-specific information, but also structured and multi-perspective reasoning. Existing multi-agent approaches often rely on fixed roles or shallow interaction prompts, limiting their ability to detect and resolve fine-grained logical inconsistencies. To address this, we propose MedLA, a logic-driven multi-agent framework built on large language models. Each agent organizes its reasoning process into an explicit logical tree based on syllogistic triads (major premise, minor premise, and conclusion), enabling transparent inference and premise-level alignment. Agents engage in a multi-round, graph-guided discussion to compare and iteratively refine their logic trees, achieving consensus through error correction and contradiction resolution. We demonstrate that MedLA consistently outperforms both static role-based systems and single-agent baselines on challenging benchmarks such as MedDDx and standard medical QA tasks. Furthermore, MedLA scales effectively across both open-source and commercial LLM backbones, achieving state-of-the-art performance and offering a generalizable paradigm for trustworthy medical reasoning.

The WAVE seismic network is a dense, multi-instrument monitoring system deployed on a scientific campus in Hamburg, Germany. It combines seismometers, geophones, and a 19 km distributed acoustic sensing fiber loop installed in existing telecommunication infrastructure. The network covers large-scale research facilities including the European X-ray Free-Electron Laser (EuXFEL) and particle accelerators at DESY. Its primary goal is to characterise natural and anthropogenic ground vibrations and to quantify how these signals couple into ultra-precise measurement infrastructures that are limited by environmental noise. Beyond local applications, WAVE serves as a testbed for fibre-optic sensing concepts relevant to fundamental physics, including seismic and strain monitoring for gravitational wave detection.The EuXFEL is a femtosecond X-ray light source designed for ultrafast imaging and spectroscopy. Its performance depends critically on precise timing and synchronisation of the electron bunches along the linear accelerator. Measurements of bunch arrival times reveal significant noise contributions in the 0.05–0.5 Hz frequency band, with peak-to-peak timing jitter of up to 25 femtoseconds. Using distributed acoustic sensing data, we demonstrate that this jitter is largely explained by secondary ocean-generated microseism, which is identified as a significant limiting factor for stable, high-precision XFEL operation in the sub-Hz regime. To assess the potential for prediction and mitigation, we investigate whether ocean wave activity in the North Atlantic can be used to anticipate microseismic signals observed at the EuXFEL site. Output from the WAVEWATCH III ocean wave model is used to generate synthetic Rayleigh wave spectrograms with the WMSAN framework. These are compared to seismic observations at the EuXFEL injector. By subdividing the North Atlantic into source regions, we evaluate their relative contributions to the observed seismic wavefield. While absolute amplitude prediction remains challenging, the modelling reproduces key spectral characteristics and temporal variability.Our results demonstrate that combining dense fibre-optic sensing with physics-based ocean wave modelling provides a framework to characterise microseismic noise and assess its limiting impact on high-precision experiments. This approach supports noise mitigation efforts at high-precision accelerator facilities and is directly relevant to future ground-based gravitational wave detectors.

Verifying the complex and multi-step reasoning of Large Language Models (LLMs) is a critical challenge, as holistic methods often overlook localized flaws. Step-by-step validation is a promising alternative, yet existing methods are often rigid. They struggle to adapt to diverse reasoning structures, from formal proofs to informal natural language narratives. To address this adaptability gap, we propose the Graph of Verification (GoV), a novel framework for adaptable and multi-granular verification. GoV's core innovation is its flexible node block architecture. This mechanism allows GoV to adaptively adjust its verification granularity—from atomic steps for formal tasks to entire paragraphs for natural language—to match the native structure of the reasoning process. This flexibility allows GoV to resolve the fundamental trade-off between verification precision and robustness. Experiments on both well-structured and loosely-structured benchmarks demonstrate GoV's versatility. The results show that GoV's adaptive approach significantly outperforms both holistic baselines and other state-of-the-art decomposition-based methods, establishing a new standard for training-free reasoning verification.

The rapid progress of multi-modal large language models (MLLMs) has boosted the task of image quality assessment (IQA). However, a key challenge arises from the inherent mismatch between the discrete token outputs of MLLMs and the continuous nature of quality scores required by IQA tasks. This discrepancy significantly hinders the performance of MLLM-based IQA methods. Previous approaches that convert discrete token predictions into continuous scores often suffer from conversion errors. Moreover, the semantic confusion introduced by level tokens (e.g., “good”) further constrains the performance of MLLMs on IQA tasks and degrades their original capabilities to related tasks. To tackle these problems, we provide a theoretical analysis of the errors inherent in previous approaches and, motivated by this analysis, propose a simple yet effective framework, Q-Scorer. This framework incorporates a lightweight regression module and IQA-specific score tokens into the MLLM pipeline. Extensive experiments demonstrate that Q-Scorer achieves state-of-the-art performance across multiple IQA benchmarks, generalizes well to mixed datasets, and further improves combined with other methods.