You are here

Data Science Based Approaches for Modeling Optimal Controls and PHM for Power and Thermal System Components.

Description:

TECHNOLOGY AREA(S): Info Systems 

OBJECTIVE: Develop model order reduction strategies for electrical power and thermal components that retain high accuracy with reduced computational time for real time control and health monitoring applications. 

DESCRIPTION: Design and Verification of advanced propulsion and electric power controls requires reduced order models (ROMs) that run in real time. Calculation of power utilization, load factors, parameter estimates, and control mechanisms is a challenge as accurate, predictive algorithms may take an order of magnitude more time to execute (versus clock time) to reach a stable solution. It is desirable to reduce this computational burden to allow real time use of novel algorithms in control systems Propulsion Health Monitoring (PHM), and power and thermal architectures. It has been demonstrated that computational statistics (CS), machine learning (ML), and related artificial intelligence (AI) techniques that access large data sets can learn constrained domains without explicit programming. They can capture a large percentage of the requirements for accuracy of complex component and system loop (feedback) models of waste heat and transient power flow for electric actuation and high electric power usage components such as diode/fiber lasers. Use of key AI techniques such as CS and ML have the potential to reduce model/algorithm execution time by one or more orders of magnitude compared to the state-of-the-art. In the Phase I program, it is desirable to employ an Artificial Intelligence (AI) machine learning techniques to develop an integrated ROM for use in simulations of power and heat flow networks (feedback loops), electric actuation controls, and high power electric loads. The ROM should incorporate suitable component specific transient (high order) power and thermal sink characteristics expected in operational scenarios. The research should explore acceptable processing execution speed versus accuracy over the domain of interest. The ROM should consider future compatibility with relevant system demonstration hardware, such as execution on an engine control verifier bench, which interfaces a FADEC, and other real-time hardware in a closed loop. 

PHASE I: Select an AI machine learning methodology for prototype development of a Reduced Order Model (ROM) for control of high power electrical component energy flows (such as actuation, lasers) and waste heat. Ensure that the transient quality (high order effects) of the waste heat and electric power is considered. Evaluate suitable software and hardware architectures that reduce computational burdens, delays and communication uncertainties. Compare the performance of the ROM with a baseline representation of the system or component to determine the performance benefits and suitability for real-time applications. Participate in a workshop with the stakeholder (including all potential users of the tool) to insure that all requirements for the future prototype are clearly understood. 

PHASE II: Relevant modeling software will be coded, refined, and tested based on the Phase I design. Demonstration of the real-time high fidelity modeling capability will be performed on a state-of-the-art closed loop control system bench. Limitations and potential operational issues will be documented, as well as, applicability to targeted advanced propulsion, power and thermal systems. Develop a training manual and a transition plan to facilitate use of the tool in the design process by an engine or airframe company. 

PHASE III: Implementation and integration of the high fidelity capability will be accomplished. Real time and other performance issues with the Phase II design will be addressed and a fielded capability will be developed that meets the engine/aircraft power and thermal control system or thermal component operational requirements. Provide training to those identified in previous phases to accelerate transition to the field. 

REFERENCES: 

1. Konrad Rawlik, Marc Toussainty, “On Stochastic Optimal Control and Reinforcement Learning by Approximate Inference,” School of Informatics, University of Edinburgh, UK, Department of Computer Science, FU Berlin, Germany (2013).; 2. Neal Dawson-Elli, Seong Beom Lee, “Data Science Approaches for Electrochemical Engineers: An Introduction through Surrogate Model Development for Lithium-Ion Batteries,” Journal of the Electrochemical Society (2018).; 3. Jiequn Han, “Deep Learning Approximation for Stochastic Control Problems,” 1The Program of Applied Mathematics, Princeton University (2016).

KEYWORDS: Control Optimization, Machine Learning, Data Science And Optimization, Power And Thermal Optimal Control, AI Methodologies In Optimal Control, Supervisory Control Techniques 

US Flag An Official Website of the United States Government