Safety code compliance checking before construction is a key step in risk control. However, the conventional safety compliance checking methods are static model-oriented, which can lead to both the low adaptability of the model to the dynamic construction process, and low checking efficiency. This paper develops a dynamic safety checking method based on BIM and topology for enhancing construction safety management, by incorporating actual construction processes. Firstly, based on the four stages of automatic safety checking, a comprehensive dynamic safety checking framework is proposed. Secondly, the object attributes and spatial location in the BIM model are extracted to form a dynamic topological relationship database. Following this, the dynamic safety checking method is designed, and the checking results are intuitively reported to users based on BIM software. An actual construction scenery is taken as an example to verify the feasibility of the method in the final stage. The results showed that the dynamic safety checking method, based on topology and rules, can help to accurately identify safety risks in the pre-construction stage and reduce the safety risks due to poor design considerations or construction process modification.

Modeling accident investigation reports is crucial for elucidating accident causation mechanisms, analyzing risk evolution processes, and formulating effective accident prevention strategies. However, such reports are typically long, hierarchically structured, and information-dense, posing unique challenges for existing language models. To address these domain-specific characteristics, this study proposes SAFE-Transformer, a Structure-Aware and Format-Enhanced Transformer designed for long-document modeling in the emergency safety context. SAFE-Transformer adopts a dual-stream encoding architecture to separately model symbolic section features and heading text, integrates hierarchical depth and format types into positional encodings, and introduces a dynamic gating unit to adaptively fuse headings with paragraph semantics. We evaluate the model on a multi-label accident intelligence classification task using a real-world corpus of 1632 official reports from high-risk industries. Results demonstrate that SAFE-Transformer effectively captures hierarchical semantic structure and outperforms strong long-text baselines. Further analysis reveals an inverted U-shaped performance trend across varying report lengths and highlights the role of attention sparsity and label distribution in long-text modeling. This work offers a practical solution for structurally complex safety documents and provides methodological insights for downstream applications in safety supervision and risk analysis.