Simulation of emergent behavior in manufacturing systems

Abstract

When we build a factory, we construct a building, purchase and install some amount of equipment, and hire and train some number of personnel. When we mix the equipment and the personnel, some behavior emerges and it is very difficult to predict what this behavior will be. In part, this is because the equipment will exhibit less than perfect availability due to pseudo-random breakdowns. But a major contribution to this unpredictability is the fact that humans who have been exposed to the same training will behave differently on the factory floor. In fact, the same people will behave differently over time given their ability to learn from experience. Clearly the behavior of the manufacturing system will depend on the integral of the behavior of the individuals and the interactions between the individuals, and will change with time. Current discrete event simulations of manufacturing are equipment-centered and can be run without any modeling of floor personnel since the equipment models include decision making capability. This is obviously not an accurate reflection of the manner in which current manufacturing systems operate. What is needed is a simulation that includes both equipment and personnel, and includes them with accurate emphasis. This paper describes a prototype simulation for a very small, but relatively complete factory. Equipment is modeled in objects that can accept commands and issue status while personnel are modeled in objects that can accept status and issue commands. An information system connects these two distinct types of objects. We demonstrate that the details of human behavior are as important as the details of machine behavior to factory performance.