Search

This invited article reflects the particular preoccupations of the author, which concern the parallel and connected development of academic models for managing inventories and their actual applications in industry. Somecentralisedproduction-inventory models that have been of service to contemporary business practices indecentralised decision-making settingsare discussed here. I have limited my examples to those in which I have had direct experience. I have attempted to highlight four key points. (a) The strategic role of inventories in capturing revenue and market share, in addition to their traditional role as buffers, in our contemporary customer-scarce Schumpeter’s market. (b) Facilitating the acceptance of model outputs by decision makers across organisational silos and even across firm boundaries using management mechanics. (c) Development and use of a general-purpose algorithm utilising infinitesimal perturbation analysis (IPA) derivatives. (d) The emergence of enterprise inventory optimisation (EIO) as mainstream software. [ABSTRACT FROM PUBLISHER]

Stochastic inventory system simulation is often the tool of choice by industry practitioners who struggle with the evaluation of the quality of proposed inventory targets using service levels. However, driving simulations with unknown input distribution parameters has its own challenges. In this paper, we focus on the newsvendor problem and quantify the amount of demand parameter uncertainty – the uncertainty around the unknown demand distribution parameters which are estimated from the limited historical demand data – in the confidence interval of the mean service level. We use this quantification to understand how the variance of the mean service level, due to the amount of the demand parameter uncertainty in the simulation output process, changes with the choice of Type-1 and Type-2 service-level criteria, the historical data length, the ratio of the unit shortage cost to the unit holding cost, and the distributional shape of the demand's density function. [ABSTRACT FROM AUTHOR]

In an environment where a buyer procures capacity from a capacitated supplier through a wholesale price contract, the impact of timing of the decisions and of additional demand information on the supplier's pricing and the buyer's procurement decisions are investigated. As the selling season approaches, the buyer and the supplier have better information about the demand process. It is questioned as to what degree waiting until the demand uncertainty is resolved—given that such an option is feasible—is the best alternative for the supply chain partners. The main model is a two-period dynamic pricing/procurement game where the supplier sets wholesale prices dynamically; he sets the second period price after seeing the demand state as well as the buyer's first period procurement quantity. Two single-period models based on the information available to the players prior to their decisions are also studied. For each model, the subgame perfect Nash equilibrium is found in a closed form and the players' behaviors are analyzed to determine the impact of the additional information and trading periods on their welfare. The supplier's optimal capacity decision is also considered. Several additional insights are obtained through numerical experiments. [ABSTRACT FROM AUTHOR]

We address several aspects related to managing re-entrant lines in a unified manner capacity allocation, inventory management and production control. Our approach to study these systems is through simulation based optimization. Simulation offers the flexibility to model the complexities adequately while the gradient computation (via infinitesimal Perturbation Analysis (IPA)) helps identify good solutions quickly. Our framework is a discrete time capacitated multiple product production-inventory system operating under a base stock policy. We analyze several different production and capacity allocation rules. We develop expressions for and validate the appropriate derivatives. These derivatives can then be used in an optimization tool which enables the determination of optimal parameters. For the several policies proposed. We also present a summary of insights obtained through an extensive computational study. [ABSTRACT FROM AUTHOR]

We present an efficient solution method -- Direct Derivative Estimation (DDE) -- for computing optimal order-up-to levels for a discrete time non-stationary inventory control model. We generalize a number of non-stationary inventory control models (which underlie larger models of supply chains) that include forecast updates. seasonality, information sharing, and exchange-rate fluctuations. This procedure is different from the existing ones in that it computes, in a recursive manner, the derivative of the cost function and not the cost function itself. It can also handle a much wider variety of fluctuations in the problem parameters. In our computational testing it was found to be considerably faster than Dynamic Programming and Infinitesimal Perturbation Analysis. [ABSTRACT FROM AUTHOR]

The integration of product design and operations could potentially enhance the ability of manufacturers to provide large variety in the product line while keeping their operations competitive. At the plant level, an improved assembly sequence can lead to superior operational response to customer demands. While this has been intuitively accepted, there is a scarcity of models that can adequately reflect the operational benefits of a costlier yet better design. In this paper, we present integrated models applicable at a plant level that can provide quantification of certain operational benefits. Further, through a computational study, we also provide qualitative insights on several issues such as the effect of : (i) features in the product line; (ii) variance of demand for the different features on the optimal assembly design; (iii) set-up times; and (iv) life cycle of the product line. [ABSTRACT FROM AUTHOR]

This paper continues our study of the design of inventory control policies for serial systems. Using simulation, we compare alternate control policies where a target of mean throughput has to be met under three different objectives -- minimize the maximum inventory, minimize the average inventory and minimize the variance of the output -- for balanced lines and lines with one bottleneck. The main objectives of this paper are: (1) to study several objectives not amenable to sample path analysis and (2) to show some sets of conditions and objectives under which Constant-Work-in Progress (CONWIP) is preferred over the traditional kanban mechanism, and vice-versa. [ABSTRACT FROM AUTHOR]

The design of inventory control policies for serial systems is a topic currently being explored by a number of researchers. Our goal -- in two papers -- is to synthesize and extend some of these efforts. We consider, simultaneously, four sources of variability in production lines -- processing time variability, machine breakdowns, rework and yield loss -- and show some similarities and differences in their effects on the performance of the line. In this paper we introduce an enhanced model that accommodates raw material and demand uncertainty and demonstrate that many of the sample path results obtained previously can be extended. The main objectives of this paper are: (1) to demonstrate that Constant-Work-in-Process (CONWIP) and the traditional kanban control are just two extremes in a finite family of implementable pull controls; (2) to show that while machine breakdowns, rework and random processing times have a similar effect in terms of the optimal decisions, yield losses in the line may have to be managed differently. [ABSTRACT FROM AUTHOR]

Most past approaches to job shop scheduling are either highly myopic, or they are unable to adapt effectively to the stream of unforeseen disruptions that characterizes almost all manufacturing systems. We propose a two-module scheduling system that is both robust and non-myopic. Periodically a Planning Module develops a global view of what is likely to happen in the facility over the next one to four weeks. It then passes information down to the Dispatching Module. The Dispatching Module is based on microprocessors that are located on the factory floor. Whenever a machine is about to become idle, a fast optimization algorithm is run on the closest microprocessor to decide which job should be done next. The data provided by the Planning Module allows this decision to be made in a non-myopic manner. [ABSTRACT FROM AUTHOR]