Apply Modelica Language to Multi-physics Applications on HPC and Cloud Platforms

Period of Performance: 02/17/2015 - 11/16/2015


Phase 1 SBIR

Recipient Firm

Sentient Science Corp.
672 Delaware Ave. Array
Buffalo, NY 14209
Firm POC
Principal Investigator


Project Summary/Abstract The object-oriented modeling language Modelica allows users to analyze the performance of their complex systems consisting of mechanical, electrical, hydraulic, control, etc. components. Many manufacturing industries and national laboratories are increasingly using Modelica to develop the next generation of energy efficient systems. However, simulating complex systems that includes components from several domains are computationally inefficient. Sentient and Xogeny are proposing an automated model reduction environment, called Mercator, that takes detailed subsystem models and, using high performance computing resources, automatically generates reduced-order Modelica models (ROMMs). ROMMs replace the original subsystem models to quickly assess overall system performance. Mercator will be a cloud-based product with dynamically adjustable computational resources. Users can access and up-/download sub-systems via a standard web-browser. Anticipated Benefits/Potential Commercial Applications of the Research or Development Several groups across multiple national laboratories (INL, LBNL, ORNL), as well as industrial companies (Boeing) are using Modelica in their modeling efforts and all of them have been extremely enthusiastic about the prospect of a tool that would allow them to easily and automatically generate reduced-order models. Mercator will provide the national laboratories and the manufacturing and engineering markets with the following benefits: 1. Increase computational efficiency of Modelica Models (allows for faster optimization of systems) 2. Use of High Performance Computers (quick creation of reduced-order models) 3. Scalable (Mercator will work on large as well as small clusters) 4. Domain independent 5. Encapsulated (reduced-order models can be easily shared between institutions) 6. Preservation of proprietary information (reduced-order models do not reveal the topology or detailed design information of the original models) 7. User-friendliness (Mercator will allocate computational resources and recommend reduction algorithms) Keywords: Modelica; High-performance computing; Cloud computing; Reduced-order models; Numerical simulation; Computational Efficiency Summary for Members of Congress There have been several important European led initiatives in the area of modeling and simulation in the last decade (e.g. Modelica and FMI) that are clearly having an impact on our shores and even in our national laboratories. This project is an attempt to cultivate such innovations led by US-based companies while benefiting US companies and national laboratories.