Research on Risk Assessment Method of Long-Tube Trailer Road Transportation

Caiyan Dai, Wenkun Wang, Ming Xu, Chenglong Ma, Lianqing Yang, Hong Zhao, Yuan HeJournal of Software Engineering and Applications 2023 · DOI

TL;DR

This paper proposes a dynamic risk assessment model for long-tube trailers transporting high-pressure industrial gases. By combining Cloud Model theory with the Analytic Hierarchy Process (AHP), researchers distilled 56 potential risk factors into 12 core indicators across three dimensions—likelihood, severity, and spatiotemporal sensitivity—enabling real-time safety monitoring for these "mobile pressure vessels."

What Problem Does This Paper Address?

Long-tube trailers are the backbone of industrial gas distribution (like hydrogen and natural gas) in China, with numbers growing from 2,000 in 2005 to over 15,000 by 2020. These vehicles carry gases at pressures exceeding 20 MPa, making them significant safety hazards if involved in road accidents.

While the technical manufacturing standards for these trailers are mature, the transportation phase remains a "black box" regarding risk. Current safety management often relies on static checklists that fail to account for dynamic variables like changing weather, road types, or the proximity of sensitive areas (e.g., schools or water reserves). The authors aim to bridge this gap by creating a quantitative, real-time risk assessment framework.

Methods

The researchers employed a multi-stage hybrid approach to move from qualitative expert opinion to a quantitative model:

  1. Cloud Model Screening: To handle the inherent "fuzziness" and randomness of human judgment, the authors used Cloud Models to analyze the opinions of 30 experts. This method converts linguistic scales (e.g., "very important") into three numerical characteristics: Expectation ($Ex$), Entropy ($En$), and Hyper-entropy ($He$). This allowed them to filter 56 initial parameters down to 29 based on expert consensus and "cloud map" cohesion.
  2. SMART Filtering: The 29 parameters were further refined using SMART principles (Specific, Measurable, Attainable, Relevant, Trackable) to ensure they could be automatically collected via sensors or GPS. This resulted in 12 core risk indicators.
  3. Analytic Hierarchy Process (AHP): AHP was used to determine the weights of these 12 indicators. The authors constructed judgment matrices based on a survey of 66 experts from specialized inspection institutes and industry groups.
  4. The R-PLS Model: The final risk ($R$) is calculated as a product of three dimensions:
  5. Likelihood ($P$): Probability of an event (e.g., equipment age, speed, weather).
  6. Severity ($L$): Potential consequences (e.g., gas type, filling capacity).
  7. Sensitivity ($S$): Environmental vulnerability (e.g., passing through residential areas or near rivers).

Datasets

The study relied on several data sources rather than a single pre-existing machine learning dataset:

  • Regulatory/Historical Data: Analysis of 32 national and local standards in China and historical accident statistics.
  • Expert Knowledge Base: Two rounds of expert surveys (30 experts for initial screening; 70 experts/66 valid responses for weighting).
  • Engineering Case Study: Real-time telemetry data from a long-tube trailer trip was used to demonstrate the model's efficacy in a "Project Demonstration" section.

Key Insights and Ideas

  • Spatiotemporal Sensitivity: A major contribution is the inclusion of "Sensitivity" as a primary risk dimension. A trailer isn't just risky because of its pressure; it is more risky when it passes a school during peak hours or a protected water source.
  • Visualizing Consensus: The use of Cloud Maps (Figures 2 and 3) provides a fascinating look at expert agreement. A "line-shaped" cloud indicates a unified understanding of a parameter's importance, while a "foggy" cloud reveals a lack of consensus, leading to the exclusion of that parameter.
  • Measurability over Theory: The authors made the pragmatic choice to delete "Fatigue degree" as a direct indicator because it is difficult to measure objectively in real-time. Instead, they used "Continuous driving hours" as a measurable proxy.

Key Results

  • The 12-Indicator System: The model successfully condensed complex safety variables into 12 actionable metrics, including equipment inspection status, vessel operating parameters, and surrounding social/natural environments.
  • Weighting Distribution: Under the "Likelihood" dimension, Vessel operating parameters (0.2441) and Continuous driving time (0.2037) were identified as the most critical factors.
  • Real-time Validation: In a case study (Figure 5), the model successfully identified a risk spike (Point A) when a vehicle entered a school zone, demonstrating that the model can provide dynamic, location-aware warnings rather than just static equipment alerts.
  • Risk Grading: The authors established a four-level risk scale (Level I to IV) based on the "80/20 rule," providing a clear framework for enterprise managers to trigger emergency responses.

Limitations and Future Work

  • Automation Gaps: The authors acknowledge that while they prioritized measurable indicators, some data points are still not perfectly automated.
  • Human Factors: Future research aims to incorporate more sophisticated safety state parameters, such as direct driver fatigue monitoring (e.g., via computer vision), to improve the model's granularity.
  • Visualization: While the model works mathematically, the authors suggest that better real-time visualization tools are needed for fleet dispatchers to act on the data effectively.

Summary updated

Abstract

Road transport safety has always been paid attention to by the safety production managers of enterprises. In this study, cloud model and analytic hierarchy process were applied to the safety of long-tube trailer transport. The opinions of 30 experts were analyzed, from which 29 key parameters were selected. The study addressed the relevance of the parameters and the possibility of automatic collection and transmission to obtain 12 core risk factors. The macro-safety risk indicator system for long-tube trailers was established based on the identified risk indicators. Finally, a risk assessment model for road transport of long tube trailers consisting of 3 dimensions of likelihood, severity and sensitivity was constructed. This model provides a technical method for strengthening the risk control of road transport of long-tube trailers.
BibTeX
@article{dai2023research,
  author = {Caiyan Dai and Wenkun Wang and Ming Xu and Chenglong Ma and Lianqing Yang and Hong ZHAO and Yuan He},
  title = {{Research on Risk Assessment Method of Long-Tube Trailer Road Transportation}},
  journal = {Journal of Software Engineering and Applications},
  year = {2023},
  doi = {10.4236/jsea.2023.168021},
}
Figure