การแก้ปัญหาการจัดส่งคอนกรีตผสมเสร็จแบบมีการแบ่งส่วนจัดส่ง โดยใช้วิธีการวิวัฒนาการโดยใช้ผลต่าง

Abstract

The objective of this study is solved ready-mixed concrete dispatching problem with split delivery. The goal is to construct a ready-mixed concrete delivery schedule with minimization of total costs. This research begun by collecting data from a case study. Then, a mathematical model of an integer non-linear programming (INLP) with deterministic of all parameters is proposed. The constraints of this consist of concrete delivery for one working day, a single depot, two vehicle sizes, accessible eligibility, and customer rejection. Fourteen instances are solved by IBM ILOG CP Optimizer (CPO), an optimizer in IBM ILOG CPLEX Optimization Studio 12.10.0. The results showed that optimal solutions can be obtained swittly in small-sized instances. An increasing of problem size resulted in increased time-consumed, cannot obtain a solution within acceptable time, or due to out of memory. Thus, we have to implementing a meta-heuristic called Differential Evolution algorithm (DE) for solving a larger-sized instances and implementing in real-world problem. Differential Evolution algorithm was implemented using nine instances generated based on delivery capacity. The suitable parameter settings are determined for different problem sizes. It was found that there is no single best settings for small-sized instances . The medium-sized instances suitable setting is DE/best/1 - 20 vectors 500 iterations F0.5 CR0.5. The large-sized instances suitable setting is DE/best/1 - 50 vectors 200 iterations F0.25 CR0.25/0.5. A performance benchmark of IBM ILOG CP Optimizer and Differential Evolution algorithm for solving instances of the mathematical programming using DE/rand/1 - 50 Vectors 200 Iterations F0. 5 CR0.5 parameter setting showed that a few solution differences between two approaches in several instances. Differential Evolution algorithm had better solution in some instances. Similarly, CPO had better solution in some instances too. However, Differential Evolution algorithm consumed much less time than the CPO in most instances. Moreover, this study developed a package software for RMC delivery scheduling. The software supports adding, deleting, or editing the data and providing results responded to uncertainty situations and ease of use in real-world application.