MULTI-source Learning-Accelerated Design of high-Efficiency multi-stage compRessor (MULTI-LEADER)

Default ARPA-E Project Image

Program:
DIFFERENTIATE
Award:
$1,548,000
Location:
East Hartford, Connecticut
Status:
ALUMNI
Project Term:
06/19/2020 - 12/18/2022
Website:

Critical Need:

The DIFFERENTIATE program seeks to leverage the emerging artificial intelligence (AI) revolution to help resolve the energy and environmental challenges of our time. The program aims to speed energy innovation by incorporating machine learning (ML) into the energy technology development process. A core part of AI, ML is the study of computer algorithms that improve automatically through experience. This approach is expected to facilitate a rapid transition to lower-carbon-footprint energy sources and systems. To organize the proposed efforts, the program uses a simplified engineering design process framework to conceptualize several ML tools that could help engineers execute and solve these problems in a manner that dramatically accelerates the pace of energy innovation.

Project Innovation + Advantages:

The United Technologies Research Center (UTRC) will work to accelerate the design of high-efficiency multi-stage compressors, via machine learning (ML), with considerations of aerodynamics, structures and additive manufacturability through their framework, MULTI-LEADER. The framework addresses four design challenges in current industrial practices: (1) concurrent optimization of multiple stages under non-linear constraints; (2) evaluation of high-fidelity and expensive solvers and their gradients during optimization convergence in high-dimensional design spaces; (3) multi-disciplinary design to maximize aerodynamic performance while guaranteeing structural integrity and additive manufacturability; and (4) use of multiple fidelity of solvers with disparate parameterization and modeling assumptions. MULTI-LEADER has the potential to cut design costs by 80% while generating more energy-efficient designs of multi-stage compressors through faster and fewer design iterations, improved empiricism and performance evaluation, and quicker concurrent design processes. The proposed framework will deploy novel machine learning algorithms for multi-source learning of universal surrogate, physics-constrained data augmented modeling, generative manifold embedding, and budget-constrained fidelity-adaptive sampling to achieve the project goals.

Potential Impact:

DIFFERENTIATE aims to enhance the productivity of energy engineers in helping them to develop next-generation energy technologies. If successful, DIFFERENTIATE will yield the following benefits in ARPA-E mission areas:

Security:

Seek U.S. technological competitive advantage by leading the development of machine-learning enhanced engineering design tools.

Environment:

Use these tools to solve our most challenging energy and environmental problems by facilitating an economically-attractive transition to lower carbon-footprint energy sources and systems.


Economy:

Reap the economic productivity benefits associated with the commercial adoption of the resulting higher-value energy technologies and associated products.

Contact

ARPA-E Program Director:
Dr. David Tew
Project Contact:
Dr. Soumalya Sarkar
Press and General Inquiries Email:
ARPA-E-Comms@hq.doe.gov
Project Contact Email:
sarkars@utrc.utc.com

Partners

University of Michigan
University of Maryland
University of Pennsylvania

Related Projects

• Carnegie Mellon University (CMU) - Predicting Catalyst Surface Stability Under Reaction Conditions Using Deep Reinforcement Learning and Machine Learning Potentials
• Carnegie Mellon University (CMU) - High-fidelity Accelerated Design of High-performance Electrochemical Systems
• General Electric (GE) Global Research - Probabalistic Machine Learning for Inverse Design of Aerodynamic Systems (Pro-ML IDeAS)
• General Electric (GE) Global Research - IMPACT: Design of Integrated Multi-physics, Producible Additive Components for Turbomachinery
• IBM T.J. Watson Research Center - Model-based Reinforcement Learning with Active Learning for Efficient Electrical Power Converter Design
• Iowa State University (ISU) - Context-Aware Learning for Inverse Design in Photovoltaics
• Julia Computing - Accelerating Coupled HVAC/Building Simulation with a Neural Component Architecture
• Lawrence Berkeley National Laboratory (LBNL) - Deep Learning and Natural Language Processing for Accelerated Inverse Design of Optical Metamaterials
• Los Alamos National Laboratory (LANL) - Machine Learning-Based Well Design to Enhance Unconventional Energy Production
• Massachusetts Institute of Technology (MIT) - Machine Learning Assisted Models for Understanding and Optimizing Boiling Heat Transfer on Scalable Random Surfaces
• Massachusetts Institute of Technology (MIT) - Global Optimization of Multicomponent Oxide Catalysts for OER/ORR
• National Renewable Energy Laboratory (NREL) - End-to-End Optimization for Battery Materials and Molecules by Combining Graph Neural Networks and Reinforcement Learning
• National Renewable Energy Laboratory (NREL) - INTEGRATE - Inverse Network Transformations for Efficient Generation of Robust Airfoil and Turbine Enhancements
• Northwestern University - Adaptive Discovery and Mixed-Variable Optimization of Next Generation Synthesizable Microelectronic Materials
• Pacific Northwest National Laboratory (PNNL) - Machine Learning for Natural Gas to Electric Power System Design
• Princeton University - MLSPICE: Machine Learning based SPICE Modeling Platform for Power Magnetics
• Stanford University - Energy Efficient Integrated Photonic Systems based on Inverse Design
• United Technologies Research Center (UTRC) - MULTI-source Learning-Accelerated Design of high-Efficiency multi-stage compRessor (MULTI-LEADER)
• United Technologies Research Center (UTRC) - Learning Enabled Network Synthesis (LENS)
• University of Maryland (UMD) - Invertible Design Manifolds for Heat Transfer Surfaces (INVERT)
• University of Michigan, Dearborn - ML-ACCEPT: Machine-Learning-enhanced Automated Circuit Configuration and Evaluation of Power Converters
• University of Missouri - Deep Learning Prediction of Protein Complex Structures
• University of Texas at Austin (UT Austin) - Learning Optimal Aerodynamic Designs

Release Date:
04/05/2019