An open tool-set for simulation, design-space exploration and optimization of supply chains and inventory problems

Abstract

This paper presents the design overview and work-in-progress status for InventOpt-a Python-based, open tool-set for simulation, design space exploration and optimization of supply chains and inventory systems. InventOpt consists of a Python library of component models that can be instantiated and connected together to model and simulate complex supply chains. In addition, InventOpt contains a GUI-based tool to assist the user in planning design of experiments, visualizing the objective functions over a multi-dimensional design space, building and tuning meta-models and performing meta-model assisted optimization to identify promising regions in the design space. We present a detailed case study that illustrates the current prototype implementation, planned features and utility of the tool-set. The case study consists of simulation-based optimization of inventory threshold levels in a particular supply chain system with 8 decision parameters. We present our observations from the case study that lead to design decisions for building InventOpt such as the choice of the meta-model type, number of simulation measurements for building the meta-model, the choice of optimizer and the trade-off between computational cost and quality of results. A significant aspect of this work is that each step of the process has been implemented using open Python libraries.