Mathematical programming approach for the long-distance truck loading problem considering flexible demand requirements and multiple performance measures.

MARIA ANALÍA RODRIGUEZ; JUAN MATIAS NOVAS; JUAN IGNACIO RAMELLO

Valdivia

Workshop; Land Translog 2019; 2019

Instituto de Sistemas Complejos de Ingeniería - CONICYT

Truck loading is a key activity in logistic operations that has drawn the attention from both the academic community and the industry practitioners, due to its inner combinatory complexity and its impact on logistic costs. Even when there exists a significant number of contributions addressing this type of problems, further research is required to cope with issues and features usually found in real settings. In this respect, the truck loading problem is addressed at a real context, which is one of the major nonalcoholic & ready-to-drink bottlers in Argentina. The company produces and delivers soft drink products for more than 50% of the country territory. The logistic area is the responsible for the planning and the execution of the delivery of final products, by dispatching between 50 and 100 trucks per day. One of the critical activities is the loading of the trucks, by means of which the assignment of products to the trucks is carried out. Trucks of different dimensions and weight loading capacities are provided by the transportation suppliers. Each truck is fully loaded in the main plant, with single-product pallets varying in weigh and dimension, and completely unloaded in a given distribution center. Before proposed improvements, the truck load planning process was manually carried out and it was assumed that all trucks had the same dimensions and weight capacities. Given that most of the operational restrictions were ignored, the resulting plans had to be adjusted to the actual fleet by the warehouse operators, causing several drawbacks. Moreover, the load distribution on the trailer was not considered, thus, most trucks were dispatched overloaded. This issue has an economic impact, since a penalty fee is charged when the excess is detected. The objective of this work was to improve the truck loading planning process automatizing it. A first linear disjunctive formulation was developed in order to model most of the operational practices and constraints. By this, a better use of their capacities is obtained in comparison to the manual strategy. In a second stage, this formulation was extended in order to include the balance of forces in the truck. Both formulations are modeled as Generalized Disjunctive Programming (GDP) problems which are transformed into ILP and MILP, respectively. The performance of the models was tested considering data from the company. Although the complex nature of the formulations, good quality results are obtained that are implemented in the every-day practice.