Introduction
Building a discrete-event simulation model for planning adaptable material flow systems is still challenging, despite ongoing research. This is because simulation experts need to create and compare many structural variants based on key performance indicators like throughput, workload, and investment costs, while also ensuring the system remains flexible.
This paper introduces a method to simplify the process of creating structural variations during simulation studies. Starting with a valid initial simulation model, structural variants are automatically generated using component-based software synthesis and combinatorial logic. This approach provides users with a range of simulation models to choose from.
The paper demonstrates this methodology through a case study and discusses how it helps reduce the effort needed for simulation experiments.
Procedure model
The main goal is to create a systematic procedure that allows users of simulation tools to automatically generate different versions of their models. The authors developed a comprehensive procedure model, shown below using AnyLogic software.
AnyLogic was chosen for its dedicated material handling library and multimethod modeling environment. In this model, users specify the part or the entire model they want new versions of, and these parts are then regenerated. The coordinates of the selected region and the relevant model boundaries are sent to the synthesis framework.
The Combinatory Logic Synthesizer (CLS) framework uses a collection of modular building blocks, like machines or buffers, to find solutions that manage the material flow from input to output. CLS creates all possible solutions using these building blocks that meet the input and output requirements.
These solutions are then sent to a connector, which generates an XML file for each one. These XML files include the components and their coordinates, making them easy to use in AnyLogic.
As a result, the user receives a variety of new simulation models, each a different version of their original model. These variants can be run to thoroughly assess and evaluate their performance.
Results
A case study is used to show how different versions of a simulation model can be automatically created from an initial valid model.
Using the CLS-Framework, these versions are generated automatically based on a target description. The process starts with graph models of the initial simulation and its variants. Then, a connector translates the results into AnyLogic XML, making them directly usable in the simulation tool.
While the current solution can automatically create different versions of a model using specific operators, there are still some research challenges to address. Such as further researching the complexity and expressiveness of the synthesis approach and the evaluation of structural variants. This involves improving the method to allow for more precise specifications, creating more complex structures, and synthesizing advanced components like distribution strategies.
While the methodology reduces the effort needed to build different simulation model versions, these versions are only useful when used in simulation experiments. Their evaluation is based on performance metrics like throughput or utilization rate, but other criteria related to the flexibility of the structure should also be considered.