31.08. – 03.09.2021 Bern
M. Sc. Jana Plitt
Parameterization of optimization-based order release in the context of engineer-to-order systems
Ralf Gössinger and Jana Plitt
Order release is the sub-task of production planning that initializes the execution of already accepted orders. Besides the uncertainty of order arrival, in engineer-to-order systems two further kinds of uncertainty become relevant for this task: the period of complete specification of an order (specification time uncertainty) and the manufacturing process of an order (capacity requirements uncertainty). To decide on the release of an order with incomplete specification, the order is subdivided into components such that completely specified (incompletely specified) components can be released finally (provisionally). For incompletely specified components, more information is provided at a later, yet unknown point in time. Since further information on order arrival, specification time, capacity requirements and order fulfillment becomes available with the passage of time, a MILP model embedded in a rolling horizon approach has been proposed. Thereby the values of the stochastic variables “specification time” and “capacity requirements” are estimated using chance-constraints, for which probability thresholds need to be specified in advance as parameters.
In order to obtain clues for setting the parameter values in a favorable way, we conduct a full-factorial numerical study that reveals the model behavior in terms of costs and robustness. Statistical analyzes confirm a significant connection between these attributes of release plans and the parameter values. Therefore, the results of the numerical study are used to determine costs and robustness as functions of the parameter values. Analyzes of such functions enable the delimitation of areas of parameter value combinations that are dominant with respect to single or multiple objectives.