A Lagrangian Relaxation approach for production planning with demand uncertainty

A production planning problem with stochastic demands is considered in this paper. The problem is to determine over a given time horizon the production quantity of each intermediate/final product at each facility of finite capacity so that a system-wide total cost is minimised while meeting given service level requirements for the final products. After reformulating the stochastic decision problem as a multiitem, multistage capacitated lot-sizing problem with a non-linear cost function using deterministic equivalence, it is solved by using a Lagrangian Relaxation (LR) approach enhanced with a local search method based on a modified simplex algorithm. Numerical experiments show that the approach can find high quality near-optimal solutions for randomly generated problems of realistic sizes in a computation time much shorter than that of an exact algorithm. [Received on 2 February 2007; Revised 28 May 2007; Accepted 7 June 2007]