Dean C. Chatfield

SISCO: an object-oriented supply chain simulation system

We present SISCO, the Simulator for Integrated Supply Chain Operations, an object-oriented supply chain simulation tool. SISCO advances the concept of a supply chain simulator in a number of ways. With SISCO, we introduce a fundamentally new approach to supply chain specification, storage, and model generation. The user specifies the structure and policies of a supply chain with a GUI-based application and then saves the supply chain description in the open, XML-based Supply Chain Modeling Language (SCML) format. SISCO automatically generates the simulation model when needed by mapping the contents of the SCML file to a library of supply-chain-oriented simulation classes. SISCOs object-oriented, agent-style system architecture and detailed output improve upon current supply chain simulation tools.

A multi-formalism architecture for agent-based, order-centric supply chain simulation

Abstract Agent-based simulation models can effectively represent decentralized systems. However, many supply-chains are order-driven, and agent modeling cannot effectively represent the order life-cycle. We present a conceptual architecture that combines simulation formalisms, allowing an agent representation of the supply-chain infrastructure while enabling a process-oriented approach to representing orders. This architecture allows for a natural, realistic representation of different supply-chain constructs and subsystems while following a consistent overall viewpoint. Our approach provides for excellent representation of supply-chain operations, allows for very detailed operational data to be gathered, and provides efficient representation of concurrent supply-chain activities in a manner that avoids preemption.

The economic lot scheduling problem: A pure genetic search approach

Abstract The economic lot scheduling problem (ELSP) is the challenge of accommodating several products to be produced on a single machine in a cyclical pattern. A solution involves determining the repetitive production schedule for N products with a goal of minimizing the total of setup and holding costs. We develop the genetic lot scheduling (GLS) procedure. This method combines an extended solution structure with a new item scheduling approach, allowing a greater number of potential schedules to be considered while being the first to explicitly state the assignment of products to periods as part of the solution structure. We maintain efficient solution feasibility determination, a problematic part of ELSP solution generation and a weakness of several other methods, by employing simple but effective sequencing rules that create “nested” schedules. We create a binary chromosomal representation of the new problem formulation and utilize a genetic algorithm to efficiently search for low cost ELSP solutions. The procedure is applied to a benchmark problem suite from the literature, including Bombergers stamping problem [Bomberger, A dynamic programming approach to a lot scheduling problem. Management Science 1966; 12:778–84], a problem that has been under investigation since the mid 1960s. The genetic lot scheduling procedure produces impressive results, including the best solutions obtained to date on some problems. Scope and purpose The need to utilize a single facility to produce several items is a common situation. We study the fundamental form of this situation, the economic lot scheduling problem (ELSP). The ELSP involves determining the minimal cost production schedule for several items that must be produced on a single machine while meeting all demands for each product. Setup costs and times are incurred each time the machine is changed (setup) to produce a different item. Holding costs are incurred based on the average amount of an item held in inventory. We must determine how much of each item should be made in each production run as well as the order of production. Examples of ELSP situations include producing several different colors of paint on the same equipment, and producing several types of stamped metal parts with the same stamping press. Since the ELSP is NP-complete, research, dating back to the 1950s, has focused on developing efficient heuristics that yield quality solutions. Our purpose is to develop an efficient heuristic search procedure for the ELSP that addresses shortcomings of previous methods. We present an extended and more definitive solution structure than previous ELSP methods, one that allows a direct mapping of solution parameters to an implementable production schedule. In addition, we address solution feasibility testing issues and structure our problem formulation in a manner that allows efficient search via a genetic algorithm (GA). We create a non-hybridized GA-based procedure that produces ELSP solutions of excellent quality, in some cases the best solutions to date.

Underestimating the bullwhip effect: a simulation study of the decomposability assumption

We investigate the assumption of decomposability as it pertains to modelling the bullwhip effect in multi-stage supply chains. Decomposing a multi-stage supply chain into a set of node pairs, each of which can be efficiently represented with a two-stage model, is a common modelling technique when analysing the bullwhip effect in supply chains. This approach depends on the validity of the decomposability assumption since most supply chains are coupled systems that are a logical fit for singular, or ‘monolithic’, multi-stage models. We utilise a simulation study to compare decomposition-based supply-chain models with monolithic models and determine if decomposition modelling significantly alters the predicted severity of the bullwhip effect. We find decomposition-based models exhibit a significantly lower level of bullwhip effect than monolithic models of the same supply chain. The systematic underestimation of the bullwhip effect by decomposition-based models indicates that the assumption of decomposability is flawed. Our study also confirms previous work showing the significant benefit of using actual, instead of approximate, lead-time demand information. We discuss implications for supply-chain modelling, supply-chain design, and data collection.

A Solution for the Intractable Inventory Model When Both Demand and Lead Time are Stochastic

We consider the reorder point, order quantity inventory model where the demand, D, and the lead time, L, are independently and identically distributed (iid) random variables. This model is analytically intractable because of order crossover. However, we show how to resolve the intractability by empirical means, for example, by regression relationships produced by simulation and factorial experiments. Using a normal approximation, we show how to obtain regression equations for the optimal cost and the optimal policy parameters (here the order quantity and the safety stock factor) in terms of the problem parameters (ordering cost per order, holding cost per unit per unit time, shortage cost per unit, the standard deviation of demand, and the standard deviation of lead time).

SCML: An information framework to support supply chain modeling

We develop an open information standard to assist supply chain modeling, analysis, and decision support. The Supply Chain Modeling Language (SCML) is a platform- and methodology-independent Extensible Markup Language (XML)-based markup language for storing supply chain structural and managerial information. SCML enables supply chain problem instance reuse and sharing, provides a common format for analytical software interoperability, and can improve the quality of the supply chain description. We develop several pieces of software to aid both users and developers in the utilization of SCML. We demonstrate SCMLs applicability by developing a supply chain simulation tool (SISCO) that utilizes the SCML format.

All-zero forecasts for lumpy demand: a factorial study

Lumpy demand is a phenomenon encountered in manufacturing or retailing when the items are slow-moving or too expensive, for example fighter plane engines. So far, the seminal procedure of Crostons (1972), with or without modifications, has been the preference for forecasting lumpy demand. Nevertheless, Croston (1974) and others, such as Venkitachalam et al . (2002), have suggested the use of zero forecasts when the demand contains many zeros. In this paper, we put to the test this idea by doing a full factorial study comparing five forecasting methods, including all-zero, under several levels of demand lumpiness, demand variation and ordering, holding and shortage cost. We evaluate the forecasting methods by three measures of forecast error and two measures of inventory cost. We find that all-zero forecasts yield the lowest cost when lumpiness is high; is it also best for mid-lumpiness, if the shortage cost is much higher than the holding cost. We also find that the lowest forecasting error does not necessarily lead to the lowest system cost. And contrary to the assertions in Chen et al . (2000b) and Dejonckheere et al . (2003, 2004), our factorial experiment reinforces the intuition that simple exponential smoothing is superior to an equivalent moving average.

Stockout propagation and amplification in supply chain inventory systems

We investigate the existence and magnitude of stockout propagation and stockout amplification in the context of supply chain inventory systems. Stockout amplification is a stage-to-stage increase in overall stockout rates. Stockout propagation is the tendency for stockout at one node to instigate a stockout at a neighbouring node and is conceptually related to the idea of cascading failures in physical systems, such as electrical power grids. We study these concepts in both upstream (‘supply side’) and downstream (‘demand side’) directions in the context of normal operating conditions for an adaptive R, S (periodic, order-up-to) inventory policy. We build a simulation model of a 5-stage serial supply chain that experiences normally distributed customer demands and gamma distributed lead times. We find that stockout propagation exists, but contrary to conventional wisdom, it occurs in the upstream direction. There is little indication that stockout propagation is occurring to any significant degree in the downstream direction. We also find stockout amplification occurring in the upstream direction in scenarios where more aggressively adaptive inventory parameter updating is performed. We discuss implications of this work in the areas of supply chain inventory modelling, ordering decisions, safety stock determination, and the use of adaptive inventory policies.

Exploring the structural properties of the ( D , 0) inventory model

We consider the (D, 0) inventory model, where the demand per unit time, D, is stationary and independently and identically distributed (iid) and the lead time, L, is deterministic. In the case of the re-order point, order quantity (s, Q) system, where the cost function is convex in s and Q, this yields two equations that are solved iteratively to yield the optimal policy. The question that we address here concerns the effect of the variability in lead time demand on the total cost and the policy parameters. Simply stated: given other parameters, such as ordering, holding, and shortage costs, can we write the optimal total cost and the policy parameters as linear or quadratic functions of the standard deviation of demand during lead time?