Efficient buffer design algorithms for production line profit maximization

Abstract

A production line is a manufacturing system where machines are connected in series and separated by buffers. The inclusion of buffers increases the average production rate of the line by limiting the propagation of disruptions, but at the cost of additional capital investment, floor space of the line, and inventory. Production lines are also a special case of assembly/disassembly systems as well as closed-loop systems. This thesis makes contributions to production system profit maximization. The profit of a production line is the revenue associated with the production rate minus the buffer space cost and average inventory holding cost. We assume that machines have already been chosen and therefore our only decision variables are the buffer sizes and the loop population. The difficulties of the research come from evaluation and optimization. We improve evaluation of loop systems. The optimization problem is hard since both the objective function and the constraints are nonlinear. Our optimization problem, where we consider the nonlinear production rate constraint and average inventory cost, is new. We present an accurate, fast, and reliable algorithm for maximizing profits through buffer space optimization for production lines, and extend the algorithm to closed-loop systems and production lines with an additional maximum part waiting time constraint. A nonlinear programming approach is adopted to solve the optimization problem. Two necessary modifications are proposed to improve the accuracy of the existing loop evaluation method before optimization of loops is studied. An analytical formulation of the part waiting time distribution is developed for two-machine one-buffer lines. It is used in the profit maximization for production lines with both the production rate constraint and the maximum part waiting time constraint. Numerical experiments are provided to show the accuracy and efficiency of the proposed algorithms. Finally, a segmentation method and an additive property of production line optimization are studied. They enable us to optimize very long lines rapidly and accurately.