Description of MLCLSP Test Instances

The test instances provided here serve as a basis for computational tests on lot-sizing problems of the MLCLSP type without backlogging (Multi-Level Capacitated Lot-sizing Problem, see Sect. 2 for a model formulation). They are structured according to the test instances of Tempelmeier and Derstroff (1996) (available at http://www.uni-koeln.de/wiso-fak/spw/publikationen/index.htm). The test instances described here are characterised by a longer planning horizon and seasonality in the demand data. Test set C(m) in addition includes a second ressource for each operation. However, some instances of test sets C, D and E with corresponding attributes match the ones of Tempelmeier and Derstroff (1996) exactly.

The solutions made available via this website are not proven to be optimal, because of the problem size and the solution approach used. However, they are the best solutions known to the authors. The research community is encouraged to submit better or optimal solutions, which then will be made public, too.

The test instances are grouped into sets of test instances by the number of periods in the planning horizon (T), the number of products (items, operations) (J), the number of resource groups (M) and whether or not setup times are included in the model. Resulting are the eight sets shown in table 1.

Test set	T	J	M	Setup times	Test data	Solutions
A+	24	10	3	No	classap.zip	solutionsap.xls
B+	24	10	3	Yes	classbp.zip	solutionsbp.xls
C	16	40	6	No	classc.zip	solutionsc.xls
C+	48	40	6	No	classcp.zip	solutionscp.xls
C(m)	16	40	6	No	classcm.zip	solutionscm.xls
D	16	40	6	Yes	classd.zip	solutionsd.xls
D+	48	40	6	Yes	classdp.zip	will be available soon
E	16	100	10	No	classe.zip	solutionse.xls
E+	48	100	10	No	classep.zip	solutionsep.xls

Full download: testdata.zip

The test instance data of each test set is characterised by the following criteria:

The demand series is described by the number of periods in the planning horizon (T), the length and amplitude of the seasonal demand pattern and different coefficients of variation for the demand data reflecting stochastic variations.

The products/operations are defined by their number (J), their structure (general or assembly), different TBO profiles and whether or not setup times are modelled.

The resources are described by their number (M), their assignment (cyclic or non-cyclic) and by different utilization profiles.

A comprehensive introduction to the test sets is available in PDF-format (120KB) or PS-format (113KB, zip-file).

Please note, that computational tests have not been performed for all possible combinations of the test instance defining criteria and hence solutions are - so far - only available for a subset of the test instances.

More test data on lotsizing problems is available at http://www.core.ucl.ac.be/wolsey/lotsizel.htm

State: 03/07/2002

Contact: Christopher Sürie