| Hazardous materials are important raw materials for chemical production.95% of hazardous materials need to be transported off-site,80% of which depend on road transportation.Hazardous materials,as flammable and explosive,have a high potential danger coefficient in road transportation,which may cause significant losses to surrounding people and the environment.Based on this background,the dissertation studies the risk assessment model of hazardous materials road transportation,and optimizes the distribution route of hazardous materials road transportation based on the risk analysis,in order to achieve the dual goals of the lowest cost and the lowest risk in the distribution system.Firstly,the dissertation figures out hazardous materials road transport accidents in the past three years,combines with the classification of accident risk factors and the grey correlation analysis method is used to take the total number of accidents.Take the first-level risk factors and total number of accidents as the reference sequence,then take first-level and second-level risk factors as the comparison sequence.The identification results show that management factors and environmental factors are the main factors that induce the occurrence of hazardous materials transportation accidents.Among them,facilities and equipment factors,transportation supervision factors are the main aspects of management factors.Environmental factors are equivalent to operational factors.It is recommended that hazardous materials logistics enterprises and transportation management departments should strengthen cooperation to create a good working environment for transport personnel.Secondly,combined with previous risk identification results,the indicators of the risk assessment model are proposed and quantified.Traditional SGD algorithm trains BP neural networks with problems such as reduced accuracy and difficulty in parallel implementation,and other existing training algorithms also have problems such as slow training processes and large computer resources,so the dissertation introduces the most commonly used Adam optimization algorithm in deep learning as training algorithm to update the network weights and thresholds,then clarify the Adam-BP neural network risk assessment model process.The risk values of the training samples and the test samples are assigned by AHP after index weights are determined.Thirdly,a multi-objective optimization model of hazardous materials road transportation was constructed using the secondary distribution of refined oil as the abstraction of dangerous goods road transportation.The model considered constraints such as soft time window,vehicle capacity,and number of vehicles to minimize the total cost and risk of the distribution system.The principle of the solution algorithm NSGA-? was elaborated,and the algorithm implementation process was clarified.A greedy algorithm was proposed to improve the initial population,and the distance gene bank method was used to improve the initial population quality.Then the Or-opt algorithm was used to improve the cross method and reduce the traditional cross method randomness.Finally,take the distribution of W Company's gas station in Wuhan as an example.After determining the distribution data and the risk attributes of the road section,the Adam-BP neural network model is used to train samples and then determine the number of nodes in the hidden layer.By comparison,the results show that Adam's optimization algorithm has good training results and outputs a risk matrix for the link.The MATLAB software was used to program the NSGA-? algorithm before and after the improvement,which confirmed the advantages of the improved algorithm,such as high operation efficiency,strong stability,and difficulty in precocity.After obtaining the Pareto optimal solution,in order to satisfy the decision makers' preferences for the target considerations,the distribution schemes with different weights were obtained for reference. |
PDF Full Text Request