Abstract
In order to reduce the operation risks of freeway maintenance, this study utilizes the Apriori algorithm to explore interconnected factors in deployment, control, and construction processes that affect operational safety. An analysis model was established based on an improved LEC (Likelihood, Exposure, Consequence) method. The model was used to calculate and analyze the risk level of a maintenance workzone, yielding a risk value between 47-56, categorized as low and consistent with on-site safety investigations. This approach effectively pinpoints key risk factors, providing theoretical guidance for improving freeway maintenance safety.
Introduction
Freeway maintenance is inherently dangerous. As global roadwork volumes increase, so do the associated safety risks. Traditionally, research has focused on external traffic flow—such as speeding vehicles or lane changes—as the primary drivers of accidents. However, a significant gap exists in understanding "internal" operational risks: worker behaviors, signage placement, and equipment deployment. Traditional risk assessment methods often rely on subjective expert intuition, which can lack the precision needed for complex workzone environments. This paper proposes a quantitative model to identify internal factors and their interconnections to improve safety outcomes.
Methodology: The Apriori-LEC Approach
The researchers developed a hybrid framework called Apriori-LEC, which combines data mining with a modified safety assessment metric.
The methodology consists of three primary stages:
- Apriori Algorithm: A classic data mining technique used to identify "association rules" between different risk factors. This helps uncover hidden links, such as how the absence of a traffic guide might correlate with improper transition lengths.
- Improved LEC Method: The standard LEC method calculates risk ($D$) as $D = L \times E \times C$. The authors refined this by creating objective, data-driven scoring tables specifically for freeway maintenance. For instance, the Consequence ($C$) score is linked to specific physical outcomes like "vehicle intrusion" rather than generic severity levels.
- Comprehensive Evaluation Model: Risks from deployment/control ($D_1$), construction operations ($D_2$), and other factors ($D_3$) are aggregated using an Independence Weighting Method. This ensures that each category contributes proportionally to the final safety score based on its statistical significance.
Risk Factor Identification and Datasets
The study utilized a real-world dataset from a provincial freeway maintenance project, covering 33 days of statistical data. Sources included rectification notices and safety inspection reports. Risk factors were categorized into specific zones (warning, transition, buffer, termination) and operational types (personnel, vehicles, material management).
As shown in the factor coding, specific issues like "Insufficient length of upstream transition" ($X8$) and "No traffic guide" ($X24$) are identified as critical variables. By shifting focus to these internal mismanagement precursors—such as "personnel having their backs to oncoming traffic"—the model provides a granular view of safety hazards.
Key Insights and Results
The application of the Apriori algorithm revealed strong associations between systemic setup failures. For example, $X8$ and $X24$ showed a high correlation, suggesting that negligence in initial workzone setup often leads to further safety lapses.
One of the most critical findings involved cone marker placement. "Excessive spacing of cone markers" reached a risk value of 720, placing it in the "Extremely Dangerous" (Level I) category. This highlights how seemingly minor details can become primary risk drivers.
Model Validation
The model was validated against three historical accidents ($S1, S2, S3$). It correctly identified that the conditions surrounding accident $S3$ represented a "Major Hazard," matching the actual severity of that incident. For the primary project analyzed, the total risk value was calculated between 47 and 56, indicating a Level IV (Low Risk) status, which aligned with on-site safety audits.
The Independence Weighting Method determined that "Risk of control factors" (36.80%) carries more weight in overall safety than "Construction factors" (33.74%), emphasizing the importance of the initial workzone deployment phase.
Conclusion and Future Directions
The Apriori-LEC model provides a more objective framework for freeway maintenance safety. However, limitations exist regarding data specificity to a single province and the potential for professional bias in initial importance scoring. Future research should focus on integrating this model into real-time monitoring systems, utilizing IoT or AI-enabled cameras to provide live risk alerts to managers, thereby transitioning from post-hoc analysis to active accident prevention.
Summary updated