Research
The AISL is one of the leading engineering laboratories in applied artificial intelligence, with a broad range of projects sponsored by industrial and governmental sources. AISL has state-of-the-art hardware and software facilities. The research team at the School of Nuclear Engineering at Purdue University performs active research dedicated towards muon imaging with special emphasis in nuclear applications and the application of innovative problem solving techniques to modeling, simulation and control of complex systems. These techniques include novel signal processing and machine learning methodologies including neural networks, fuzzy logic, genetic algorithms and evolutionary computing, Gaussian processes, wavelet analysis, Hilbert-Huang transform, expert systems and advanced signal/image processing tools. The research team has state-of-the-art hardware and software facilities and significant experience with industrial and safeguards applications. Competitively funded projects, among others, include: Smart Data Embedding and Inverse Solution Algorithms, Modeling and Simulation for Nonproliferation (sponsored by NNSA), Intelligent Management of the Electric Power Grid (sponsored by DOD and EPRI), Trend Identification in Nuclear Reactor Control (sponsored by ANL), Intelligent Flowmeter Diagnostics (sponsored by Emerson Electric), and Hazardous Material Identification (sponsored by Crane Naval Systems).
Intelligent systems refer to a collection of computational paradigms that include neural networks, fuzzy logic, expert systems, anticipatory systems and genetic algorithms. Intelligent systems research in nuclear engineering aims at improving safety and operation of current nuclear power reactors. In addition, research contributes to passive safety, enhanced autonomy, and remote monitoring diagnostics and control in future reactor systems.
ONR-BAA: Reactor Simulation Tool for Investigating the Resilience of a Cyberphysical Security Ecosystem
Office of Naval Research
Award Period: March 2018-April 2021
Purdue Research Team:
Principal Investigator: Prof. Lefteri H. Tsoukalas
Co-Principal Investigator: Prof. Miltos Alamaniotis
Description:
The focus of this study is to implement a simulation tool for the PUR-1, by modeling the plant physically and building a protection architecture for detecting and mitigating cyber attacks.
Completed Projects
CNEC: Consortium for Nonproliferation Enabling Capabilities
National Nuclear Security Administration (NNSA)
Award Period: October 2014-July 2020
Partners:
North Carolina State University (LEAD UNIVERSITY)
University of Michigan
University of Illinois at Urbana Champaign
Kansas State University
Georgia Institute of Technology
North Carolina A&T State University
Los Alamos National Laboratory
Oak Ridge National Laboratory
Pacific Northwest National Laboratory
Purdue Research Team:
Principal Investigator: Prof. Lefteri H. Tsoukalas
Co-Principal Investigator: Prof. Chan Choi
Collaborator: Prof. Miltos Alamaniotis
Research Work on:
i) Signatures and Observables Thrust Area
ii) Simulation and Modeling Thrust Area
Description:
Focused on developing novel methods and techniques using machine learning for analyzing gamma ray signals; identifying new isotopic signatures; modeling background radiation.
Machine Intelligence for Dynamic Data-Driven Morphing of Nodal Demand in Smart Energy Systems
US National Science Foundation
Award Period: September 2015-August 2017
Partners:
Purdue University (Lead)
University of Texas at San Antonio
Purdue Research Team:
Principal Investigator: Prof. Lefteri H. Tsoukalas
Co-Principal Investigator: Prof. Miltos Alamaniotis
Description:
The focus of this study will be the development of a set of new intelligent and self-adaptive algorithms for online big data processing and fast real-time decision-making in smart energy infrastructures. The main feature of the current research is the integration of machine learning dynamic data driven systems with dynamic optimization methods to solve the computational problem of forecasting optimal or near-optimal shapes of a load in a timely manner accounting for multiple streams of continuously incoming data and their inherent uncertainty.
Creation of a Geant4 Muon Tomography Package for Imaging of Nuclear Fuel in Dry Cask Storage
Nuclear Energy University Program (NEUP)
Award Period: January 2013-December 2015
Partners:
Purdue University (Lead)
Purdue Research Team:
Principal Investigator: Prof. Lefteri H. Tsoukalas
Description:
Monitoring spent nuclear fuel stored in dense shielded dry casks using cosmic ray muons has the potential to allow for non-destructive assessment of nuclear material accountancy with the aim to independently verify and identify weapons grade material, such as fuel pellets, fuel rods and fuel assemblies stored within those sealed dense dry casks. Cosmic ray muons are charged particles, having approximately 200 times the mass of electron, generated naturally in the atmosphere, and rain down upon the earth.Energetic muons have the unique ability to penetrate high density materials allowing the distribution of material within the object to be inferred from muon measurements. High energy cosmic rays continuously entering Earth's atmosphere generate a cascade of secondary rays and relativistic particles. Of those that eventually reach the surface are cosmic ray muons.
Intelligent Model-Assisted Sensing System (iMASS) for Fast and Accurate Nuclear Material Interrogation
US National Science Foundation
Award Period: August 2007 - September 2010
Partners:
Purdue University (Lead)
Description:
iMASS will combine Computer Simulated-Nuclear Resonance Fluorescence (CS-NRF) with real time Monte Carlo to iteratively and adaptively integrate gamma measurements and decision-making. Conventional sensing systems collect data, process and analyze it for the purpose of reaching a decision. This one-way pass of information is not effective for detection of nuclear materials where the signal-to-noise ratio is extremely low (particularly when the material is well-shielded). iMASS will use multiple information passes. In an iterative manner, measurement and real-time Monte Carlo will converge to the most likely scenario of concealment. This ultra fast process is guided by a perception unit, which regulates the overall detection process and makes decisions to adaptively and optimally tune system parameters.
Intelligent Component Monitoring for Nuclear Power Plants
Department of Energy (DOE)
Award Period: August 2006 - September 2009
Partners:
Purdue University (Lead)
Purdue Research Team:
Principal Investigator: Prof. Lefteri H. Tsoukalas
Co-Principal Investigator: Rong Gao
Description:
Develop and test in Purdue’s reactor PUR-1 an intelligent prognostics methodology for predicting aging effects impacting long-term performance of nuclear components and systems. The approach is particularly suitable for predicting the performance of nuclear reactor systems which have low failure probabilities (e.g., less than 10-6 year-1). Such components and systems are often perceived as peripheral to the reactor and are left somewhat unattended. That is, even when inspected, if they are not perceived to be causing some immediate problem, they may not be paid due attention. Attention to such systems normally involves long term monitoring and possibly reasoning with multiple features and evidence, requirements that are not best suited for humans.
Consortium for the Intelligent Management of the Electric Power Grid (CIMEG)
Electric Power Research Institute (EPRI)
Award Period: January 1997-December 2002
Partners:
University of Tennesse
Purdue University (Lead)
Fisk University
Commonwealth Edison
Tennessee Valley Authority
Description:
CIMEG seeks to investigate and advance innovative modeling, measurement and control approaches for the intelligent management of the vast and increasingly deregulated continental electric power grid. The focus of the consortium is a simple but potentially powerful idea: To protect itself from upset events, the grid should act proactively, that is, effect local control in anticipation (not just in response) of possible contingencies. The thesis that anticipation of the future is an integral part of control strategies originated in the field of mathematical biology and ethology by Rosen (1985) as a way of explaining how organisms cope with complexity. Advancing an anticipatory formalism for complex engineering systems such as the power grid, however, presents unique and challenging opportunities for research at the frontiers of modeling, measurement, forecasting and control.