TL;DR
This paper introduces a proactive "Design for Safety" (DFS) framework that automatically identifies and quantifies construction risks directly within Building Information Modeling (BIM) software. By integrating a specialized knowledge base with an improved risk quantification model, the researchers enable engineers to catch potential hazards—like foundation pit collapses—during the design phase before a single shovel hits the ground.
What Problem Does This Paper Address?
In the construction industry, safety management is traditionally reactive. We analyze what went wrong after an accident occurs or try to manage risks on-site during the high-pressure construction phase. However, many disasters, such as the 2008 Hangzhou Metro collapse, are rooted in "hidden defects" created during the planning and design stages.
The challenge is that identifying these risks manually is grueling. Designers must cross-reference complex 2D drawings against hundreds of safety regulations and historical accident data. This process is prone to human error and often ignores "constructability"—how the design choices impact the safety of the workers who have to build it. This paper seeks to bridge the gap between design intent and construction reality through automation.
Methods
The authors developed a four-step "BIM-enhanced" approach:
- Knowledge Base Construction: They compiled a database of safety rules by analyzing 36 Chinese technical regulations and 52 real-world subway construction safety accidents (SCSA).
- Mechanism Analysis: Using accident causality and trajectory crossing theories, they mapped how specific design parameters (like anchor length or reinforcement ratios) directly lead to "unsafe states" on-site.
- Improved FEC Quantification: They refined the traditional FEC (Frequency, Exposure, Criticality) method:
- Frequency (F): Derived from historical accident rates per hour.
- Exposure (E): Calculated automatically based on the construction schedule (how many hours workers are exposed to a specific hazard).
- Criticality (C): Scored based on potential economic losses and casualty severity.
- BIM Integration: Using C# and the Revit API, the team developed a custom plug-in. This tool traverses the BIM model, extracts geometric data, compares it against the knowledge base, and automatically calculates and visualizes risk scores.
Datasets
- Historical Accident Data: A collection of 52 detailed subway construction safety accidents in China (2002–2016) used to determine risk frequencies and common failure modes.
- Normative Regulations: 36 sets of design and technical standards covering subway survey, design, and construction.
- Case Study: A real-world subway station project featuring a 170m long, 16.5m deep foundation pit.
Key Insights and Ideas
- Design as a Safety Lever: The paper emphasizes that a design can be "structurally safe" (the building won't fall down) but "constructively dangerous" (it's impossible to build safely). DFS forces the designer to consider the worker's environment.
- Quantifying the Subjective: Traditional risk assessment relies heavily on "expert intuition," which is arbitrary. By tying Exposure to the actual construction schedule and Frequency to historical data, the authors move toward a more objective, data-driven metric.
- The "Aha" Moment in Automation: The BIM tool doesn't just flag a rule violation; it provides a "Risk Score." For example, it can tell a designer that an insufficient anchor depth represents a specific monetary risk (e.g., 46,000 yuan), making the "business case" for safety improvements immediate and clear.
Key Results
The framework was validated through a case study of a deep foundation pit:
- Automatic Detection: The BIM plug-in successfully identified that the embedded depth of anchor rods was insufficient according to technical regulations ($< 0.6h$).
- Risk Quantification: The system calculated a safety risk score of 3.8.
- Economic Translation: By converting this score, the researchers estimated a potential direct economic loss of 46,000 yuan if the design wasn't corrected, providing a tangible metric for decision-makers.
- Efficiency: The automated check replaced what would have been a manual, multi-day review of drawings and spreadsheets with a near-instantaneous software check.
Limitations and Future Work
While promising, the authors acknowledge several hurdles:
- Scope: The current knowledge base is specifically tailored to subway deep foundation pits. Expanding this to other complex structures (tunnels, bridges) requires significant data entry.
- Perspective: The model primarily reflects the designer's viewpoint. Future iterations should incorporate data from construction agents, equipment suppliers, and on-site workers for a "multi-party" risk model.
- Data Limitations: The quantification relies on existing historical data, which may be incomplete or vary significantly across different regions or countries.
- Logic Complexity: Currently, the BIM tool handles relatively simple, parameter-based rules. Capturing complex, interdependent risks (where three "safe" parameters combine to create an "unsafe" condition) remains a challenge for future software development.
Summary updated