Strategies for designing transformable production networks

Patrick A Brunese, Purdue University

Abstract

The effective design of a production network has a significant impact on the continued success of any manufacturing company, especially those that produce many products across numerous facilities. The production network design problem is composed of two distinct phases which are traditionally considered separately. The first phase is the location of facilities and the sourcing of products to the facilities. In the case where facilities already exist the first phase is just the sourcing of products to these facilities. Once products have been allocated to the facilities each facility can be designed. The specific sourcing of products to a facility is the major determinant of the production system design, and therefore, the overall facility design. By considering the sourcing of products and the facility design separately, there is the potential for facility designs that do not take advantage of efficient production system designs. Therefore, there is a natural trade-off that exists between product sourcing and facility design, especially when the facility is capable of transformation. Products should be sourced to facilities closest to their demand sources to minimize shipping costs, but should be also be located in a facility best equipped to produce efficiently. It may not always be the case where the same facility satisfies both desires. Furthermore, traditional modeling techniques do not make provision for the simultaneous consideration of product sourcing and production system design. Furthermore, since consumer demand changes frequently the ability of a production network to adapt to changes must be studied. Such a production network is classified as a transformable production network. A transformable production network is one where each facility within the network is capable of being redesigned regularly, and that resources can be shared among the facilities, as needed, to cope with changes in customer demand. In this research, the production network design problem is studied. A general model is developed for the static production network design problem that assumes a functional layout structure at each facility. An example case is generated and solved which illustrate the usage of the model. An additional static production network design model is developed to consider cellular layout structure at each facility. This model is an improvement on a previous (and, to the author, only known) production network design model. In addition, the model is extended to the time-domain to allow for network transformability. This research presents the first production network design models considering functional layout structure (for both static and time-dependent cases). In addition, a strategy for improving the machine layout problem (MLP) is developed based on the concept of variable bounding. It is well known that the MLP is a hard problem to solve due to its combinatorial and nonlinear structure. This is especially true when there is no assumption made of the facility design in which the machines are to be arranged. The bounding scheme is developed in the context of the single-row machine layout problem and extended to two dimensions. The developed bound is tighter than previously existing bounds. Statistical tests were performed on two different formulations for the MLP and the bounding scheme does not significantly reduce time to solution for the MLP; however, the majority of test cases experienced a reduction in time to solution.

Degree

Ph.D.

Advisors

Tanchoco, Purdue University.

Subject Area

Industrial engineering

COinS