Song Xiaoyu

Emergency goods scheduling model and algorithm during initial stage of disaster relief

During initial stage of disaster relief, real time requirement and limited relief source quantity are the main features of relief work. To improve the efficiency and precision of emergency goods scheduling, a model is established for the multi-depot and multi disaster places, single category material requirement emergency scheduling problem. The objective of the model includes minimizing task complete time and relief participated vehicles. The model is under the constraints of the relief vehicle number and the specified time by disaster places. Genetic algorithm is adopted to optimize the routes, to determine the number of participated vehicles and routes of varied emergency relief centers, thus the minimum emergency time and vehicles can be obtained. Finally an example is employed to validate the validity of the model and algorithm.

Demand Feature Identifying for Emergency Goods under Earthquake Rescue Logistics by Vague Set Based BP Neural Network

During the preliminary stage of rescue for earthquake disaster, some important information needs to be provided by identifying the key demand features such as quantity, necessary degree and urgent degree for emergency goods. To dispose uncertain information, especially unknown information for identifying above features, vague set is introduced into the conventional BP neural network. The relation among supportive degree, counteractive degree and unknown degree are analyzed. For disposing unknown information for vague set, the method for transforming vague set information into fuzzy set information is proposed. A series of rules for transforming vague values into fuzzy values are presented. Hereby, the fuzzy membership degree function is given. The structure of four-layer multi output VBP neural network is designed, and its implement steps are studied. To validate the validity of VBP neural network, it is applied to identify demand feature for emergency goods under earthquake. The result by VBP neural network is compared with those by conventional BP neural network and the VBP neural network which is based on conventional fuzzy membership degree transforming formula. The result shows that, the test precision by above VBP neural network is higher than those by the other two methods. As a novel learning method, VBP neural network is more suitable for training samples with uncertain information.

A multi-category emergency goods distribution model and its algorithm

When large scale emergency incident happens, the multi-category emergency goods distribution is a complex problem. To improve structure and efficiency of emergency goods distribution, a multi-objective non-linear optimization programming model for emergency goods distribution is established. Modeling precondition is analyzed in detail, some objectives such as minimizing the waiting time, the transportation cost, the preference of selecting emergency commodity distribution center (ECDC) and so on are taken into account. The constraints of the model including total demand quantity and structure of disaster area, the transportation quantity and time from each ECDC are considered. The penalty factor method is employed to transform above model into an easy one, then GA is employed. The encoding system for multi-category emergency goods distribution is introduced, and the basic implement steps of GA are given in detail. The typical instance abstracted from practice is adopted to validate the validity of above model and algorithm, the experiment result shows that above model can describe the practical demand of multi-category emergency goods distribution, and GA is suitable for solving some complex non-linear programming problems such as multi-category emergency goods distribution. Above model and algorithm will benefit for the emergency goods distribution practice.