| As the world economy continues to develop, more and more companies began to realize that companies will get a huge competitive advantage in the fierce competition from the development of reverse logistics. However, because of the shortage in researching the reverse logistics at all aspects, which will make the implementation of reverse logistics network need to pay a great cost, which will give the company a significant financial risk. therefore, the issue of reverse logistics network optimization studies such as how to reduce the costs of the implementation of reverse logistics network and how to improve the operational efficiency of the entire network, have an important practical significance, which is to be an very important key to implement reverse logistics.Basing on the predecessors'great works on reverse logistics network optimization studies , this paper will explore the optimization issues in the field of Refabrication.(1) We consider a single-product , single-cycle reverse logistics network model. In this model , basing on the original logistics network, we will consider how to extend the nodes in the forward logistics into forward/backward logistics networks to reduce costs and improve the implementation of the efficiency of circulation. Built the entire network nodes including: manufacturing plants, re-manufacturing plant, manufacturing / remanufacturing plants, distribution centers, recycling centres , distribution / collection centres, customer base, as well as the abandoned waste Centre dealing with non-use, recovery and recycling goods.(2)Basing on the above the model, the mathematical model is established. In order to establish the mathematical model expediently, certain assumptions are necessary . On the principle of guaranteeing the lowest cost of the whole network ,considering how to get a function to depict the lowest cost under certain restrictive conditions.We can set the optional nodes ,which will be possible to be choosen to build recycling centres or distribution / collection centres,as 0-1 variables , if we choose to set up a new building at this point , the variable will be evaluated with 1;if not, the variable will be evaluated with 0.The next step is to set the quantity in any possible routine as float variable . Then we can establish a mixed integer programming model (MILP)of single-product and single-cycle manufacturing / remanufacturing reverse logistics network:(3)Using the genetic algorithms solve the model. In the coding stage, considering the yi and y j,as well as xs si and xs sj have the different meanings, therefore,we use the mixed coding. yi and y j use the binary coding,0 means the address is not selected , and 1 means that the address is selected; xs si and xs sj use the float coding which means that the quantity of the recycled products coming from the customers and being sent to different distribution centers or distribution / collection centres are different. During the selection , the most important thing is to find a good fitness function. Because the model that the paper referred is restricted optimization problem conditions,so the first problem is How to lift restrictive conditions, in the paper, I will use punishment function to do it.In the selection process , use the roulette wheel selection.In the crossover stage , using single-point crossover. First, have the selected point crossovered. And then, have the related gen point corssovered. In the mutation stage, we use single-point mutation and still need to think about the two steps mutation. First, have the selected point mutated. And then have the related xs si and xs sj mutated. If the selected point is 1,change the value into 0,and have the whole related xs si and xs sj been 0. If the selected point is 0,change the value into 1,and have the whole related xs si and xs sj been evaluated. If the xs si and xs sj are selected, we use the following function to mutate: x ' = A + k ( A + B).(4)Using the C programming language to realize the the genetic algorithm, and to solve the problem. During the process, we have to pay attention to the following points.First, initiate the population. Because we choose using the mixed coding, we have to initiate the variables separately. Using a array in length of k×( I×S + J×S ) + ( I + J) and evaluate values randomly to the whole ( I + J) array elements. And then evaluate values to the following k×( I×S + J×S) elements.Getting fitness function for every body . First, getting the punishment function values: punish_function= d i( x )*(1+generation)/(2*a1) And then basing on the function (4-1),getting the whole network cost without restricted conditions minZ.Next, According to limit construction method, get the fitness function:Fit ( f ( x )) = C max? evalution _function.Mutation. Using the single-point mutation. Selecte a mutation point,if the points's value is 1,have it been 0,and have the related gen point been 0.If the point is 0,use the steps above to process too.(5)Consider a real example responsing the model above. The model is: 1 re-manufacturing centre, three recycling centres, two distribution / collection centres and six customer points. And use C programming to solve it, and the best selection is: select 3 recycle centres. To the 1st centres, the quantity coming from the 6 costomer points separately are: 847,734,1447,3537,2069,1;To the 2nd centres, the quantity coming from the 6 costomer points separately are: 988,703,2257,473,1093,1177;To the 3rd centres, the quantity coming from the 6 costomer points separately are: 401,1030,384,588,326,46.And to the whole network, the lowest cost is 1053600. |
