System Analysis Using Multiple Expert Tools

Abstract

Many of today‟s advanced simulation tools are suitable for modeling specific systems; however, they provide rather limited support for model building and management. Setting up a detailed vehicle simulation model requires more than writing down state equations and running them on a computer. In this paper, we describe how modern software techniques can be used to support modeling and design activities, with the objective of providing better system models more quickly by assembling these system models in a “plug-and-play” architecture. Instead of developing detailed models specifically for Argonne National Laboratory‟s Autonomie modeling tool, we have chosen to place emphasis on integrating and re-using the system models, regardless of the environment in which they were initially developed. By way of example, this paper describes a vehicle model composed of a detailed engine model from GT Power, a transmission from AMESim, and with vehicle dynamics from CarSim. The paper will explain the different options available for the interface and how each of these options can be implemented. It will use a simple case study to show how the detailed expert simulation models can be used with simpler Simulink models to address different vehicle design issues. INTRODUCTION One of the basic capabilities that an enterprise-wide modeling and simulation tool needs is to provide a framework for integrating models developed in different languages/tools. Because different expert tools have distinct advantages for modeling specific phenomena, numerous tools are used across companies, such as automotive original equipment manufacturers (OEMs). When these models are available, nevertheless the attempt to integrate them into a single environment to build an accurate system model is very onerous. As a result, most vehicle models used for analysis are usually based on steady-state look-up tables. The lack of a common framework to link such expert models results in isolated pockets of model-rich design groups. The open architecture of Autonomie [1, 2], a modeling tool developed by Argonne National Laboratory (Argonne) over the past four years in collaboration with General Motors (GM), supports the rapid integration and analysis of powertrain/propulsion systems and technologies for rapid technology sorting and evaluation of fuel economy improvement under dynamic/transient testing conditions. The tool, developed in the MathWorks environment, is based on a forward-looking model architecture. Its plug-and-play architecture provides a common framework in which expert tools can be linked to build detailed vehicle models. In this paper, we will explain how Autonomie is used to act as the interface between models developed with expert tool by taking the GT-Power engine model, AMESim transmission model, and CarSim vehicle dynamics models as examples. THE NEED FOR DETAILED SYSTEM MODELS Each type of model has its own advantages and disadvantages. While Simulink models based on simple look-up tables are easy to develop and use, they lack the ability to simulate the details of the plant‟s physical interactions. As a result, simple plant models cannot support control systems development. More detailed models are also needed to evaluate the effect of new technology. Different expert tools are available to meet such needs. It is far more practical to use those tools than to build and validate an equally detailed model in Simulink. While the detailed models add accuracy, the input data needed for those models are difficult to obtain, and the models run slower. Hence, engineers must switch between the detailed and simpler models depending on the analysis and what level of accuracy in the simulation results is desired. As shown in Fig. 1, Autonomie provides a convenient way to use models of appropriate fidelity by enabling a plug-and-play function with the models and by interfacing with expert system tools.