2022. In order to analyze the behavior of the proposed network and mathematical model in the future, we generated different scenarios using ELVs projections. For projecting the number of ELVs, we used the framework adopted by Dargay and Gately (1999) and Andersen and Larsen (2008). We combined the historical data, mainly from OECDSTAT and TURKSTAT on population, GDP and the number of cars per capita and the vintage distribution of cars to project the number of ELVs in the future. Firstly, since the relationship between the car density and the number of ELVs cannot be ignored, we made projections of the car ownership in Ankara to the year 2022. As in the literature, GDP-dependent Gompertz function is used to model the development in car ownership which is an S-curve that in-creases towards a saturation level and Weibull distribution is used to determine the attrition rates of car vintages (Dargay and Gately, 1999; Andersen and Larsen, 2008). We assumed zero import and export of old cars for simplicity. The Gompertz equation relating car ownership per capita (Ct) to the income per capita (GDPt) can be expressed in Eq. (1) where g is the saturation level and a and b are negative parameters defining the shape, or curvature of the func-tion (Dargey and Gately, 1997).
Ct ¼ g$ea$e
|
b$GDPt
|
Eq. (1)
|
|
|
|
The estimated saturation level (g) of 246.15 per thousand capita for Turkey by Medlock and Soligo (2002) is used and a and b are estimated as 9.761 and 1.785 depending on this g.
Having Ct, the stock of cars is calculated as:
St ¼ Ct $Pt Eq. (2)
In this section we aimed to model the generation of ELVs and presented a projection of the number of ELVs from the 2012 to
Table 8
Estimation results.
where Pt is the population.
For a specific vintage of cars, the lifetime is obtained by Weibull distribution given by:
|
Year
|
Per-capita income (2005
|
$ PPP)
|
Population of Ankara
|
Car ownership per 1000 population
|
Total vehicles in Ankara
|
No of ELVs in Ankara
|
|
|
|
|
|
|
|
|
|
2012
|
13,643
|
|
4,965,542
|
117.0700
|
581,316
|
2304
|
|
2013
|
13,901
|
|
5,056,126
|
121.2559
|
613,085
|
3262
|
|
2014
|
14,378
|
|
5,146,307
|
129.3429
|
665,638
|
4504
|
|
2015
|
15,160
|
|
5,235,807
|
143.6328
|
752,034
|
6084
|
|
2016
|
15,972
|
|
5,324,705
|
159.8741
|
851,283
|
8068
|
|
2017
|
16,751
|
|
5,413,000
|
176.9358
|
957,753
|
10,532
|
|
2018
|
17,499
|
|
5,500,577
|
194.7431
|
1,071,200
|
13,565
|
|
2019
|
18,227
|
|
5,587,439
|
213.5728
|
1,193,325
|
17,270
|
|
2020
|
18,949
|
|
5,673,544
|
233.7608
|
1,326,252
|
21,766
|
|
2021
|
19,674
|
|
5,758,868
|
255.6208
|
1,472,087
|
27,189
|
|
2022
|
20,404
|
|
5,843,435
|
279.4207
|
1,632,776
|
33,691
|
|
|
|
|
|
|
|
|
Please cite this article in press as: Demirel, E., et al., A mixed integer linear programming model to optimize reverse logistics activities of end-of-life vehicles in Turkey, Journal of Cleaner Production (2014), http://dx.doi.org/10.1016/j.jclepro.2014.10.079
10 E. Demirel et al. / Journal of Cleaner Production xxx (2014) 1e13
Table 10
Solutions of the model with different scenarios for the amount of ELVs.
Scenarios
|
Dismantler
|
|
Shredder
|
Obj. Function
|
CPU time (s)
|
|
|
locations
|
|
locations
|
( 106)
|
|
|
0
|
24
|
|
|
8
|
4.0397
|
12.7
|
|
1
|
24
|
|
|
8
|
4.2805
|
7.6
|
|
2
|
24
|
|
|
8
|
4.7610
|
6.9
|
|
3
|
24
|
|
|
8
|
5.3841
|
2.8
|
|
4
|
2,17
|
|
|
8
|
6.7954
|
23.8
|
|
5
|
2,17
|
|
|
8
|
7.7941
|
14.4
|
|
6
|
2,7,17
|
|
|
8
|
9.6534
|
40.2
|
|
7
|
2,7,17
|
|
|
8
|
11.1771
|
26.8
|
|
8
|
2,7,17,22
|
|
8
|
13.6640
|
187.7
|
|
9
|
2,7,17,22,27
|
|
8,9
|
19.0520
|
3329.0
|
|
10
|
2,7,17,22,24,27
|
|
8,9
|
22.4091
|
2743.6
|
|
11
|
2,3,7,17,22,24,27
|
8,9
|
26.3170
|
2860.8
|
|
|
|
|
|
|
|
|
|
|
ð
|
|
Þ l
|
k
|
|
|
|
|
|
T
|
|
|
|
|
|
|
F T ¼ e
|
q=
|
and F
|
T
|
¼ 1 for T q
|
Eq. (3)
|
|
|
|
|
where l and k are the positive scale and shape parameters and q is the location parameter of Weibull distribution. T is the age of cars, F(T) is the lifetime function determining the fraction of cars of vintage n still in operation in year t, (T ¼ t n). We regarded the fraction of cars being scrapped in the first year because of accidents and set q ¼ 0. With the assumption that the lifetime of individual vintages of cars follows the same Weibull distribution, we consid-ered 30 years mean lifetime of cars in Turkey and calibrated the parameter l as 33.44. For a symmetric and bell-shaped weibull distribution the shape parameter k is taken as 3.3.
In the year t, the remaining stock of a given vintage of cars is calculated by Eq. (4).
Sv;t ¼ Sv;v$F t v Eq. (4)
where S v,v is the initial stock of vintage n cars. ELVs of vintage n cars in the year t is calculated as given in Eq. (5).
ELVv;t ¼ Sv;t 1 Sv;t
|
Eq. (5)
|
The total number of ELVs in year t is calculated as:
|
|
ELVt ¼ X ELVv;t
|
Eq. (6)
|
v
|
|
The number of new cars in year t is calculated as given in Eq. (7).
St;t ¼ St St 1 þ ELVt Eq. (7) Share with your friends: |