Simulation self-diagnoses

Abstract

After an initial simulation model is created for a construction process, it still involves a time-consuming and difficult debugging process to identify and correct errors in the model until a valid experiment is obtained. Both comprehensive knowledge and hands-on experience are essential in achieving efficiency in this labor-intensive process. This research presents the simulation self-diagnosis methods based on the general role that simulation entities play in advancing a simulation, in which entities dynamically flow in the model and activate activities to operate as the simulation time advances. A valid simulation requires that all entities must flow in correct patterns and all activities must be correctly executed in the experiment. The presented self-diagnosis methodology consists of two separate stages: model compilation and runtime diagnosis. Compiling a model intends to examine the matching relations between modeling elements in the model. Diagnosing an experiment at runtime explores any abnormally executed activities and then to search for corresponding causes. Both stages can pinpoint errors in the model and suggest corresponding corrective measures. An example is used to illustrate the improved debugging process with the enhanced self-diagnosis function. © 2003 Elsevier Science B.V. All rights reserved.