Using hybrid SD-DES for the simulation of the manufacturing enterprise

Full Citation
Helal, M., Rabelo, L., Jones, A. 2006. Using hybrid SD-DES for the simulation of the manufacturing enterprise. The French-US Workshop on ICT and Standards for Supply Chains and PLM. Nov 6-7; NIST, Gaithersburg, MD

Extended Abstract
For decades, discrete event simulation (DES) has been the modeler’s choice for simulating the manufacturing systems. It has effectively allowed analysts model systems to the deepest level of details. Yet with the adoption of integration and system approaches in managing the manufacturing system and the pressure imposed by the increased competition and rapidly changing business environment, the need has arisen for different simulation modeling tools. Managers need to overcome the traditional organizational barriers and run their facilities in a more flexible, coordinated and dynamic manner.

However, more integration in manufacturing systems increases the levels of complexity. The need is for simple comprehensive simulation tools that help achieving coordination while being able to handle the dynamics of the complex system processes. Such tools must be holistic, easy to comprehend, and able to offer decision makers at the various management levels the appropriate levels of details that suit each of them, and reflect the impact of decisions in an enterprise-wide fashion.

We propose to combine the system dynamics (SD) methodology (Forrester, 1965) and DES in a hybrid approach to simulate the manufacturing enterprise. This combines the effectiveness of DES with the simplicity and overall system thinking approach of SD. Manufacturing enterprises consist of manufacturing and non-manufacturing functions. Including aggregate and operational levels of management in the same simulation model has become essential to correctly assess the enterprise performance. Using the most advanced equipment and producing the same product quality as competitors do not offer a competitive advantage (Wu, 1992) unless marketing, customer relations, financial aspects and other professional supporting functions are coordinated. Implementing a total quality management (TQM) program can dramatically improve the operational level performance while leading to significant decline in financial performance (Sterman et al., 1997) unless coordination with an overall simulation model of the organization is utilized. Further, approximating continuous system parameters with DES simulation models would lead to overestimating or underestimating the performance of the system (Lee et al., 2002). Many published reports have clearly indicated the need to combine the aggregate and operational levels of management in simulating the system.

There are always detailed data available for the manufacturing functions. But for the aggregate management level non-manufacturing functions, data is not usually available and in most cases only rough estimates exist. DES has been successful in conducting detailed manufacturing systems analyses such that the status of individual entities and resources can be monitored. Its outputs are given as estimates of and correlations among variables and performance measures using statistics, which are not easily understandable, especially for the case of an integrated large sized manufacturing system. Further, DES allows the evaluation of the performance for specific values of decision variables or control policies, but analyzing the stability of the system in any region or neighborhood of those values or policies is not supported (Rabelo et al., 2005). Add to that DES models rapidly become very complex with the size of the system being modeled.

SD on the other hand is a system thinking approach that targets top management levels. It addresses the impact of the system structure and inherent feedback interactions among its component, on performance. Its data requirements are minimal and it offers aggregate level managers an integrative approach to analyze the nonlinear, overlapping cause-and-effect relationships in their systems. Nevertheless, using SD at the operational levels has failed to offer the desirable outcomes.

In our hybrid SD-DES approach the enterprise model consists of a comprehensive SD model for the total enterprise system connected with a number of DES models for the parts in the system that require detailed analyses. This offers managers the needed comprehensive, simple, yet scalable approach to design and test management policies in an integrated system perspective.

However, due to the differences between the two modeling paradigms, coordinating the interactions among the SD and DES models and synchronizing the simulation time pose a challenge. Traditionally, hybrid discrete-continuous simulations have been based on control situation requirements in which a discrete system controls a continuous variable based on a preset threshold value or on/off-like discrete events (Ziegler et al., 2002; Maler et al., 1992).

We argue that such control approaches are not fully appropriate for the business systems as they normally result in discontinuous behaviors of the continuous parameters and unavoidable dominance of the discrete components. We describe the conceptual design of the SD-DES controller whose functions are the coordination and synchronization of interconnected SD and DES simulation models without having one approach dominating the other. We here discuss the considerations and challenges in achieving the synchronization of the interacting models. We also, investigate the perspectives for implementing the SD-DES controller as a generic software application.