build thermodynamic knowledge graph

This work addresses the limitation of current large language models (LLMs) in multi-zone HVAC control, which typically lack explicit modeling of building physics and thermodynamic processes. To bridge this gap, the authors propose a knowledge graph that integrates thermodynamic principles with spatial semantics, constructed upon the Brick ontology and enriched with historical environment-controller interaction data to provide LLMs with structured contextual information. This approach represents the first integration of physics-informed spatial semantic graphs into an LLM-based control framework, explicitly capturing inter-zone thermal couplings and building dynamic responses. Evaluated in a five-zone building simulation, the method significantly improves the trade-off between energy efficiency and occupant comfort compared to both conventional and existing LLM-based strategies, achieving the lowest PMV violation rate while maintaining high energy performance.

This work proposes the first fully automated framework to transform unstructured B-rep geometry from early architectural design into a knowledge graph–based building information model (BIM) and a corresponding executable building energy model (BEM). Addressing the limitations of conventional B-rep representations—which lack explicit spatial, semantic, and performance-aware structures—the method integrates automatic geometry cleaning, multi-strategy space generation, graph-based topological extraction, ontology-aligned modeling, and bidirectional BIM–EnergyPlus conversion. This enables, for the first time, a robust end-to-end mapping from freeform design geometries to semantically rich BIMs and reversible BEMs. Experimental validation on parametric, sketch-based, and real-world building datasets demonstrates high topological consistency and reliable energy performance modeling, establishing a critical infrastructure for AI-driven, performance-informed design in the early stages of architectural development.

Semantic representation of time-continuous dynamic models—particularly differential equations—in knowledge graphs remains challenging throughout the cyber-physical system (CPS) lifecycle, leading to high manual instantiation costs. Method: This paper proposes a standards-based, modular semantic modeling approach that integrates ontology engineering (OWL), Semantic Web technologies, and CPS modeling standards (SysML/ISO 10303) to construct a knowledge graph generation framework. The framework enables unified contextual linkage of dynamic behavior and heterogeneous data sources (e.g., design, maintenance). Contribution/Results: It achieves, for the first time, direct, reusable, and inference-ready embedding of differential equations in knowledge graphs. Evaluated in aerospace maintenance, the method fully captures the complex differential equations governing an electro-hydraulic servo actuator and significantly reduces manual modeling and instantiation effort.

Existing building ontologies exhibit semantic discrepancies that hinder data interoperability and reuse. This study presents the first systematic evaluation of four prominent building ontologies—Brick Schema, RealEstateCore, Project Haystack, and Google Digital Buildings—by jointly considering both TBox (axiomatic definitions) and ABox (instance assertions) dimensions, leveraging the OQuaRE quality assessment framework and empirical data from the building domain. The findings reveal that Project Haystack and Brick Schema feature more compact axiomatizations, while Brick Schema and RealEstateCore demonstrate superior expressiveness and completeness. These results indicate that no single ontology currently offers universal applicability across all building-related use cases. This work provides both methodological guidance and empirical evidence to support ontology alignment, integration, and selection in practice.

To address poor scalability in cross-system data integration, high manual modeling costs, and data privacy risks induced by cloud-based transmission in building automation, this paper proposes a localized, semantics-driven building information modeling framework. Centered on the Brick ontology, the framework integrates hierarchical tree structures with graph data models and employs a lightweight transformation algorithm to achieve automatic semantic alignment and offline structured organization of heterogeneous building data—including sensor readings, actuator states, and spatial configurations. Unlike conventional cloud-dependent approaches, our method enables rapid configuration adaptation across multiple buildings, reduces manual modeling effort by approximately 65% (empirically measured), and eliminates external transmission of sensitive data, ensuring privacy compliance. Experimental results demonstrate superior performance over existing ontology-driven methods in both historical data retrieval efficiency and cross-platform interoperability.

This work addresses the limitation in current building energy modeling caused by the absence of large-scale datasets that explicitly link geometric, topological, and physical properties. The authors introduce ArchEGraph, the first large-scale graph-based building energy dataset that aligns geometry, topology, and physics by representing buildings as heterogeneous graphs incorporating spaces, surfaces, weather conditions, and thermal loads. The dataset enables two benchmark tasks: graph reconstruction and topology-aware load prediction. ArchEGraph encompasses 5,481 buildings, 49,326 simulation cases, 133,000 space nodes, and 1.44 million surface nodes, supporting generalization evaluations across buildings and climates. It facilitates research on scalable surrogate models, and its standardized protocols validate both the effectiveness of the proposed tasks and the robustness of evaluated models.

This study addresses key challenges in reliability modeling of cyber-physical systems (CPS), including heavy reliance on expert knowledge, incomplete failure documentation, and inadequate representation of subsystem interactions. To overcome these limitations, the authors propose a Capability Interaction Graph (CIG) grounded in the Unified Foundational Ontology (UFO) to construct a semantic knowledge graph for CPS. This framework automatically derives fault trees to identify failure propagation paths and minimal cut sets. By integrating an ontology-driven CIG with knowledge graph technology, the approach enables unified semantic modeling across engineering domains and supports automated fault tree generation. Experimental results demonstrate that the proposed method explicitly captures functional dependencies and system semantics, significantly reducing modeling complexity while enhancing the accuracy of fault analysis.

This work addresses the challenges in Brick ontology classification within building management systems (BMS), where vendor heterogeneity and inconsistent metadata lead to a proliferation of classes, insufficient domain knowledge in large language models (LLMs), and high costs of manual validation. To tackle these issues, we propose Brick-DICL, a two-stage dynamic in-context learning framework that introduces dynamic context learning to Brick classification for the first time. Our approach leverages metadata-RAG to enrich domain knowledge and class-RAG to narrow the candidate space, while integrating multi-LLM collaborative prediction with a confidence-based filtering mechanism to automatically flag low-confidence results for human review. Experiments on multiple real-world building datasets demonstrate that Brick-DICL significantly outperforms existing methods, achieving high-accuracy, vendor-agnostic, and format-agnostic automated classification, thereby advancing standardization and interoperability in building systems.

This study addresses the challenge of automating geometric compliance checking in Building Information Models (BIM), where existing methods struggle with multi-hop reasoning and cross-component spatial dependencies due to a semantic gap. To overcome this, the authors propose SGR-BIM—the first graph-driven semantic reasoning system that dynamically constructs a cross-modal knowledge graph integrating user intent, regulatory semantics, and BIM geometry. This approach enables interpretable, hard-coding-free compliance reasoning, transcending the limitations of static rule templates and supporting flexible, transparent geometric validation. Evaluated on 679 expert-verified fire safety regulation queries, SGR-BIM achieves an accuracy of 84.3%, outperforming an enhanced single-agent baseline by 8.6%.

This study addresses the limitations of existing assessment tools—particularly their lack of interactivity and computational efficiency—in tackling urban heat island effects and high building energy consumption in tropical cities. The authors propose an intelligent agent framework that integrates a large language model (LLM) with lightweight physical models. Through prompt engineering, the LLM is guided to comprehend design tasks, retrieve relevant policy knowledge, and orchestrate microclimate and energy simulation models. This enables rapid, interpretable, and computationally frugal joint evaluation of thermal comfort (based on Physiological Equivalent Temperature, PET) and building energy use. By uniquely combining the LLM’s autonomous reasoning capabilities with physics-based simulations, the approach efficiently validates mitigation strategies such as green walls and cool coatings, demonstrating significant improvements in both thermal comfort and energy efficiency.