Scholarlink

<p>Engineering design evaluation of complex spatial products—such as vehicle interiors, building environments, and aircraft cabins—frequently demands high-fidelity digital representations that couple geometric accuracy with semantic interpretability. Yet a persistent methodological gap exists between raw 3D data acquisition and computable, performance-evaluable parametric models: conventional approaches either produce geometrically precise but semantically opaque point clouds, or rely on idealized CAD models that deviate from as-built reality. This paper proposes a general-purpose semantic-geometric digital twin framework that bridges this gap, enabling automated conversion of physical environments into structured, parameterized virtual models that directly support multi-dimensional design performance evaluation and optimization. The framework comprises four methodological layers applicable across engineering domains: (1) a multi-sensor fusion acquisition layer with microsecond-level time synchronization for efficient spatial data capture; (2) a globally consistent 3D reconstruction layer combining error-state Kalman filtering with factor graph optimization to suppress cumulative drift in elongated or repetitive environments; (3) a semantic-geometric hybrid modeling layer integrating deep learning segmentation with parametric geometric reconstruction to automatically identify components and extract design-critical parameters; and (4) a model-driven performance evaluation layer that quantifies multi-dimensional design metrics (comfort, safety, economics) and supports Pareto-optimal design decision-making. The framework is validated through a full-scale civil aircraft cabin case study (C919 simulator), where it achieves a 6× improvement in modeling efficiency over stationary scanning, sub-centimeter geometric accuracy (RMSE &lt; 1.5 cm), and measurement precision better than 8 mm for key human factors dimensions. The case study demonstrates the framework’s capacity for airworthiness compliance verification, ergonomic heatmap analysis, and layout optimization. Beyond aviation, the proposed methodology generalizes to any engineering design domain where physical-to-digital conversion, semantic decomposition, and performance-driven evaluation of spatial layouts are required—such as automotive interiors, hospital operating rooms, factory floor planning, and architectural space design.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A semantic-geometric digital twin framework for performance-driven design evaluation: methodology and application to civil aircraft cabins

  • 期刊论文

  • 发表日期：

    2026年3月17日

查看全文
  • Tao Shen,
  • Yanyi Li,
  • Man Zhang,
  • Qianqian Li,
  • Akram Akbar,
  • Jiao Mo,
  • Yongbin Gao,
  • Yuanyuan Li,
  • Cheng Li,
  • Chun Liu,
  • Yongqi Lou

摘要

Engineering design evaluation of complex spatial products—such as vehicle interiors, building environments, and aircraft cabins—frequently demands high-fidelity digital representations that couple geometric accuracy with semantic interpretability. Yet a persistent methodological gap exists between raw 3D data acquisition and computable, performance-evaluable parametric models: conventional approaches either produce geometrically precise but semantically opaque point clouds, or rely on idealized CAD models that deviate from as-built reality. This paper proposes a general-purpose semantic-geometric digital twin framework that bridges this gap, enabling automated conversion of physical environments into structured, parameterized virtual models that directly support multi-dimensional design performance evaluation and optimization. The framework comprises four methodological layers applicable across engineering domains: (1) a multi-sensor fusion acquisition layer with microsecond-level time synchronization for efficient spatial data capture; (2) a globally consistent 3D reconstruction layer combining error-state Kalman filtering with factor graph optimization to suppress cumulative drift in elongated or repetitive environments; (3) a semantic-geometric hybrid modeling layer integrating deep learning segmentation with parametric geometric reconstruction to automatically identify components and extract design-critical parameters; and (4) a model-driven performance evaluation layer that quantifies multi-dimensional design metrics (comfort, safety, economics) and supports Pareto-optimal design decision-making. The framework is validated through a full-scale civil aircraft cabin case study (C919 simulator), where it achieves a 6× improvement in modeling efficiency over stationary scanning, sub-centimeter geometric accuracy (RMSE < 1.5 cm), and measurement precision better than 8 mm for key human factors dimensions. The case study demonstrates the framework’s capacity for airworthiness compliance verification, ergonomic heatmap analysis, and layout optimization. Beyond aviation, the proposed methodology generalizes to any engineering design domain where physical-to-digital conversion, semantic decomposition, and performance-driven evaluation of spatial layouts are required—such as automotive interiors, hospital operating rooms, factory floor planning, and architectural space design.