Progress towards achieving the vision of smart manufacturing systems requires the abilities to conduct detailed analytics on current performance, evaluate potential future courses of actions, and set the course that best leads towards the goals.
These abilities can be respectively termed as diagnostic, predictive and prescriptive analytics. Diagnostic analytics assesses past and current performance and cause and effect relationships among major control factors and performance metrics. Predictive analytics evaluates future performance of a system operating under selected policies and forecasted requirements such as demand scenarios. Prescriptive analytics helps develop future courses of actions using approaches such as optimization and combined si mulation-optimization.
The efforts to move towards smart manufacturing thus need to be supported by diagnostic, prescriptive and predictive analytics. Jain and Shao proposed the virtual factory, a high fidelity simulation of the manufacturing system, to support data analytics.
This paper represents a small step towards building a complete virtual factory prototype by exploring what capabilities that prototype needs to estimate the feasibility of multi-resolution modeling. Our research used a limited scenario - a small job shop with a single manufacturing cell comprising four turning machines.
Our virtual prototype captures this scenario at three levels of detail. The top layer, the cell, has a model can that tracks the processing of each part as a single block of time. Typically, a cell model is implemented in a discrete event simulation (DES). At the machine level, each machine can be modeled at a greater granularity level of detail to:
- track the granular movements needed to process the part
- predict characteristics such as temperature and energy use.
Typically, a machine model is implemented using the agent-based simulation (ABS)
paradigm.