Collaborative Planning in Supply Chains pp 165-213 | Cite as

# Computational Evaluation

## Abstract

This final chapters deals with the evaluation of the collaborative planning scheme developed in chapters 4 and 5 by computational tests. The purpose is to determine the quality of solutions attainable with the scheme on the one hand and the computational efforts necessary for realizing these solutions on the other. The focus of the computational analysis is on the basic version of the scheme as described in chapter 4, i.e. one-time planning between a single buyer and supplier. However, a somewhat smaller number of tests also considers a more general SC structure with a single supplier but several buyers, as well as planning on a rolling basis between two SC partners.

### Keywords

Expense Posite Peri Lution Eosine## Preview

Unable to display preview. Download preview PDF.

### References

- 344.C.f. ILOG (2000), p. 17.Google Scholar
- 345.See in particular section 4.3, pp. 90.Google Scholar
- 346.Details on test instances and input parameters used in the computational study follow in the next section 7.2, pp. 168.Google Scholar
- 347.See e.g. Model 6, p. 72.Google Scholar
- 348.See p. 96.Google Scholar
- 349.See pp. 23.Google Scholar
- 350.C.f. Derstroff (1995), pp. 90.Google Scholar
- 351.See e.g. Tempelmeier / Derstroff (1993), pp. 68, Tempelmeier / Derstroff (1996), pp. 750, Ertogral / Wu (2000), pp. 937, Stadtler (2003), pp. 23.Google Scholar
- 352.I.e. units of item j required to produce one unit of a successor item k (see Model 1, p. 30).Google Scholar
- 353.C.f. Stadtler (2003), p. 24.Google Scholar
- 354.I.e. capacity units of resource r required to process one unit of item j (see Model 1, p. 30).Google Scholar
- 355.In test class L, where the planning interval is limited to 10 periods, available capacity changes in periods 3 and 9 rather than 4 and 10 as shown in the table.Google Scholar
- 356.See Model 1, p. 30. Variable production costs are neglected, assuming that unit production costs per item do not change during the planning interval and hence do not affect planning results.Google Scholar
- 357.With respect to capacity expansion, the concept used here differs from Derstroff (1995) who does not foresee capacity expansion at all. Stadtler (2003) allows for overtime only in the prohibitive expediting mode (c.f. Stadtler (2003), p. 26).Google Scholar
- 358.C.f. Tempelmeier (2003), p. 213.Google Scholar
- 359.C.f. Derstroff (1995), p. 92.Google Scholar
- 360.The abbreviations of cost rates follow the declarations laid out in Model 1, p. 30.Google Scholar
- 361.See e.g. Silver et al. (1998), pp. 151, Chase et al. (1998), pp. 587.Google Scholar
- 362.Of course, a (percentage) surplus could be added for overtime operation due to higher overtime wages etc. This is however omitted for the sake of simplicity.Google Scholar
- 363.C.f. Simpson / Erengüc (2001), p. 123. Since the negotiation scheme is intended to “close the cost gap” between Upstream Planning and centralized optimization, significant initial gaps are desired.Google Scholar
- 364.The problem structures considered here are identical to test set A+ used by Stadtler (2003) (see Stadtler / Sürie (2000), p. 5).Google Scholar
- 365.See p. 30.Google Scholar
- 366.See 7.2, pp. 168.Google Scholar
- 367.C.f. Stadtler (1996), p. 572, (non-negativity restrictions on variable values are ignored).Google Scholar
- 368.See Stadtler (1996), p. 570.Google Scholar
- 369.C.f. Stadtler (1996), pp. 574.Google Scholar
- 370.Also, the total time to find solutions for all test instances of test class L was seen at a reasonable limit with 63 hrs.Google Scholar
- 371.C.f. Stadtler (2003), pp. 1.Google Scholar
- 372.Strictly speaking, without optimal solution to the global MLCLSP, the best known solution no longer represents a lower bound on total SC costs. Nonetheless, best solutions are still used as comparison benchmarks in all test cases.Google Scholar
- 373.See 7.2, pp. 168.Google Scholar
- 374.This information is not taken from Table 23, but from the detailed records available to the author.Google Scholar
- 375.In case of S2, this observation is however not valid.Google Scholar
- 376.C.f. Simpson / Erengüc (2001), p. 123.Google Scholar
- 377.See 7.2, pp. 168.Google Scholar
- 378.See p. 170.Google Scholar
- 379.See pp. 125 for details.Google Scholar
- 380.See pp. 103.Google Scholar
- 381.See pp. 127.Google Scholar
- 382.In the two-partner scenario, in contrast, only 94 of 756 test instances proved capacity infeasible in Upstream Planning.Google Scholar
- 383.See pp. 103.Google Scholar
- 384.See p. 184.Google Scholar
- 385.See p. 185.Google Scholar
- 386.See p. 174.Google Scholar
- 387.For details see the description is 5.2, pp.115.Google Scholar
- 388.See pp. 115 (implementation and computational tests are limited to the version with full exchange of cost information as laid out in 5.2).Google Scholar
- 389.See p. 178.Google Scholar
- 390.Alternatively, costs only up to period 12 could be considered, but would need to be reduced by a “bonus” depending on inventory levels at the end of period 12, since production of the inventory positions leads to costs that are actually caused by demand occurring in period 13 or later.Google Scholar
- 391.See Table 22, p. 179.Google Scholar
- 392.See p. 180.Google Scholar
- 393.See Table 24, p. 181.Google Scholar
- 394.See e.g. the schematic overview in Fig. 33, p. 119.Google Scholar
- 395.See Model 14, p. 121.Google Scholar
- 396.See p. 170.Google Scholar