Research Article: Road screening and distribution route multi-objective robust optimization for hazardous materials based on neural network and genetic algorithm

Date Published: June 21, 2018

Publisher: Public Library of Science

Author(s): Changxi Ma, Wei Hao, Fuquan Pan, Wang Xiang, Xiaosong Hu.


Route optimization of hazardous materials transportation is one of the basic steps in ensuring the safety of hazardous materials transportation. The optimization scheme may be a security risk if road screening is not completed before the distribution route is optimized. For road screening issues of hazardous materials transportation, a road screening algorithm of hazardous materials transportation is built based on genetic algorithm and Levenberg–Marquardt neural network (GA-LM-NN) by analyzing 15 attributes data of each road network section. A multi-objective robust optimization model with adjustable robustness is constructed for the hazardous materials transportation problem of single distribution center to minimize transportation risk and time. A multi-objective genetic algorithm is designed to solve the problem according to the characteristics of the model. The algorithm uses an improved strategy to complete the selection operation, applies partial matching cross shift and single ortho swap methods to complete the crossover and mutation operation, and employs an exclusive method to construct Pareto optimal solutions. Studies show that the sets of hazardous materials transportation road can be found quickly through the proposed road screening algorithm based on GA-LM-NN, whereas the distribution route Pareto solutions with different levels of robustness can be found rapidly through the proposed multi-objective robust optimization model and algorithm.

Partial Text

Hazardous materials refer to products with flammable, poisonous, and corrosive properties that can cause casualties, damage to properties, and environmental pollution, and require special protection in the process of transportation, loading, unloading, and storage. In recent years, hazardous materials transportation accidents have occurred frequently, causing vehicle damages, fatalities, and environment pollution. Distribution route optimization of hazardous materials refers to the design of a safe and efficient distribution plan based on existing transportation network according to the characteristics of hazardous materials and transportation requirements. The result of this study can provide a direct reference for relevant decision-making departments. Preventing hazardous materials transportation accidents is crucial. Thus, research on hazardous materials distribution route optimization has great significance.

Multiple objectives of multi-objective optimization may be in conflict with each other, which is different from single-objective optimization. The improvement of a sub-goal will lead to a decrease in another sub-target, that is, multiple sub-goals achieving optimum are impossible. Therefore, multi-objective optimization obtains a non-inferior solution set, the elements of which are called Pareto optimal or non-inferior optimal solutions. The Pareto optimal solution can also be interpreted as no solution exists better than at least one of the goals and not worse than other goals. The elements of the Pareto optimal solution set are not comparable to each other in terms of all objectives. Using the obtained Pareto set, decision makers selected one or many solutions from the Pareto optimal solutions as the optimal solution of multi-objective optimization problem according to other information or personal preference. Therefore, the main task of solving multi-objective optimization problem is to obtain widely distributed Pareto optimal solutions. In this paper, a multi-objective GA is designed to solve this model according to the multi-objective robust optimization model characteristics. The algorithm uses an improved selection strategy to complete the operation, applies partial matching cross transposition and single ortho swap methods to complete the operation of crossover and mutation, and employs the selected method to construct the Pareto optimal solution set.

We study the Zhengzhou coal materials supply and marketing company, which is responsible for distributing explosives for the 15 coal mines of Zhengzhou Coal Group in China, such as Dragon, Cui Miao, Lu Gou, and so on. The company uses joint distribution method in which a vehicle can service multiple spots. A total of 32 roads are in the distribution area, and these roads must be selected to complete optimal distribution route choice. The maximum load of each vehicle is 8 tons, and supply is adequate. A total of 15 demand points are used, which are shown in Table 3.

Optimization of distribution route is an important link to ensure safe transportation of hazardous materials. Scientific and reasonable distribution route design of hazardous materials can make hazardous materials reach the customer demand point safely, quickly, and economically. However, the optimized scheme may incur serious security risks if road screening is not carried out before route optimization. This paper extensively studies the problem of road screening for hazardous materials transportation, and builds road screening algorithm based on GA-LM-NN and the multi-objective robust optimization model of transportation route with adjustable robustness based on Bertsimas. The improved elitist selection strategy is used to complete choice operation, partial matching cross shift method, and single ortho swap method is used to complete crossover and mutation operation. The Pareto optimal solution set is constructed based on the exclusive method. The study shows that the proposed GA-LM-NN road screening algorithm can determine quickly the suitable transportation section sets of hazardous materials. Furthermore, transportation path multi-objective robust optimization model and algorithm can determine rapidly the Pareto solution set of different robustness transportation route. Finally, decision makers can choose suitable transportation routes from better robust Pareto solutions based on actual situation or preferences through a case study.