Earthworks planning using optimization techniques: literature analysis and solution proposal

Abstract

Earthmoving operations account for approximately one-third of construction costs in large engineering projects and require efficient resources management. Since the 1980s, researchers have suggested computational optimization techniques to improve decision-making in earthworks and proposed mathematical models for material and equipment allocation. However, these computational applications are generally ignored by road construction professionals, who plan earthworks through estimations based on mass haul diagrams. Consequently, this dissertation has the objective of investigating the usage of optimization techniques in earthmoving operations to propose a novel mathematical programming approach for cost minimization on road construction projects. This research was divided into two distinct parts: A systematic mapping study and an original research article. At first, I presented a mapping study on the topic of optimization of earthmoving planning and operation. I analyzed 5,134 papers in total, selecting 72 relevant studies through consistent selection criteria. As a result, I could map the research field by identifying the most investigated subjects, optimization techniques, and research gaps. I found that allocation, fleet planning, routing, and scheduling problems were the most commonly explored topics, and linear programming, mixed-integer linear programming, and genetic algorithms were the most used optimization methods. I also observed that studies related to road construction have focused on improving well-known mathematical models, incorporating specific engineering features such as temporary haul roads, paving operations, and material mixing and recycling. Based on these research trends, I proposed a mixed-integer linear programming model to plan material allocation in earthmoving and paving operations, including geotechnical constraints and construction of haul roads. This optimization approach was validated by applying the model to a real road project with 121 cut sections, 257 fill sections, 272 pavement segments, 26 borrow pits, and five quarries. After structuring and modeling the proposed case study, I obtained the optimized solution in 2.98 seconds, indicating that realistic instances can be solved in reasonable processing times.