A Multi-Criteria Decision Matrix and Regression-Based Optimization Tool for Sustainable Gravel Road Performance Enhancement

Abstract: Nearly 30% of roads in the United States are unpaved, serving as critical links for rural transportation, agricultural operations, and local connectivity. Despite their importance, the performance of gravel roads remains highly dependent on material composition, gradation, and maintenance practices. Many local agencies rely on empirical guidelines or experience, often lacking systematic evaluation of available materials. This study develops a performance-based data-driven framework to improve gravel road design, reduce maintenance frequency, and optimize material use through laboratory evaluation, decision analysis, and computational modeling. Seventeen materials—comprising surface aggregates and subgrade soils—were characterized through comprehensive index property testing, including gradation, Atterberg limits, specific gravity, and compaction behavior. Repeated triaxial tests were conducted to determine resilient modulus (MR) and permanent deformation (PD), representing stiffness and long-term durability, respectively. Virgin materials exhibited highly variable mechanical behavior, with MR values ranging from 10 ksi for sand-dominant samples to 30 ksi for well-graded aggregates. Virgin materials also showed poor deformation resistance, indicating that they are unsuitable for direct surface application. To improve performance, granular stabilization was achieved by blending materials such as gravel, crusher run, and fine-grained soils to correct particle size distribution, shape, and plasticity deficiencies. A total of 31 optimized blends were prepared and tested, showing increased dry unit weights (average 134 pcf), improved stiffness (MR between 8 ksi and 27 ksi), and nearly 100% reduction in accumulated permanent strain compared to virgin materials. To objectively assess performance and economic feasibility, a multi-criteria decision matrix (MCDM) was developed. The framework incorporates mechanical performance (MR), durability (inverse of PD), and material cost as primary evaluation criteria. Each blend was scored and weighted according to project priorities, allowing transparent ranking and selection of the most balanced gradation combinations. An Excel-based optimization tool employing multi-linear regression models was developed to predict and design optimal gradation blends. The tool integrates available in-situ materials including existing surface, base, and subgrade materials with selected new aggregates, computing the required proportions to achieve target gradation curves and desired mechanical performance outcomes. This process promotes material reuse, reduces dependence on new aggregates, and enhances the sustainability and cost-efficiency of gravel road rehabilitation. Results confirm that the combined decision matrix and regression-based tool provide a transparent, adaptable, and performance-based methodology for gravel road design. The framework effectively captures trade-offs between stiffness, durability, and cost, supporting data-driven decisions for improved material utilization and long-term performance management of unpaved road networks.