Challenges for the automatic generation of simulation models for production systems

This paper is intended as an introduction of the challenges that exist in the area of automatic simulation model generation in the production and logistics context. As a work-in-progress paper, it firstly analyzes and classifies previous work; it then introduces the most relevant challenges and lastly presents potential solutions currently being investigated by a PhD thesis. 1. MOTIVATION Simulation is used within many different disciplines and application areas. In the area of production and logistics, simulation is a well accepted tool for the planning, evaluation and monitoring of relevant processes. It is applied to ensure the feasibility of planning concepts, to discover rationalization possibilities and to assist in decision making. A wide variety of commercial simulation systems mostly based on discrete-event simulation paradigms exists in the area of production and logistics. Considering the lifecycle of production and logistic systems, simulation is traditionally applied in both the planning phase and the operation phase. In the planning phase, simulation supports the planning and dimensioning of production systems, the design of process alternatives as well as the design of control strategies. In the operation phase, simulation is used in the context of simulation based production control, as an early-warning-system [Hotz et. al 2006], and, more generally, as a tool for visualisation of complex processes to assist in management information systems and in decision making. More recently, the usage of simulation in the context of virtual commissioning as “emulation” for the real system has received an increased interest in production and logistics [Boer and Saanen 2008]. Here, it is used to test real control software and hardware. In any of these application contexts of simulation it is essential to adequately model reality in a simulation model in a way that allows sufficiently exact predictions about the real system. The quality of the simulation based predictions directly depends on the quality of the model. The modelling process for achieving high quality models which are verified and validated is time consuming and usually requires a simulation expert. Moreover, the entire simulation technology is a very interdisciplinary technology which requires expertise in different knowledge areas, including computer science, economic sciences, as well as mechanical and industrial engineering and statistics. The latter often constitutes a problem for the application of simulation in production and logistics, especially in smaller and medium sized enterprises. The benefit that can be achieved by simulation strongly depends on the capabilities of the modelling simulation expert. In this context [Fowler and Rose 2004] have discussed future challenges for modelling and simulation of complex production systems. Among others, they identified the reduction of the time and effort for simulation studies as well as the integration of simulation with the real production as future research areas. 2. AUTOMATIC MODEL GENERATION Given these challenges, approaches to automatically generate simulation models seem to be very appealing. In such approaches of automatic (or semi-automatic) model generation a simulation model is not created manually using the modelling tools of the chosen simulator, rather it is generated from external data sources using interfaces of the simulator and algorithms for creating the model. This is often also referred to as “data-driven model generation” [Eckardt 2002]. The promise of such approaches is that they, if successful, can reduce the amount of time needed to create a simulation model as well as the expertise needed for creating and conducting simulations. There is a wide variety of potentially relevant external data sources and IT systems (Figure 1). The relevant data of such sources can be roughly classified into technical data describing the topology of a production system as well as its components,