Handbook of Dynamic Data Driven Applications Systems: Volume 1

✍ Scribed by Erik P. Blasch (editor), Frederica Darema (editor), Sai Ravela (editor), Alex J. Aved (editor)

• Publisher: Springer
• Year: 2022
• Tongue: English
• Leaves: 753
• Category: Library

✦ Synopsis

The Handbook of Dynamic Data Driven Applications Systems establishes an authoritative reference of DDDAS, pioneered by Dr. Darema and the co-authors for researchers and practitioners developing DDDAS technologies.

Beginning with general concepts and history of the paradigm, the text provides 32 chapters by leading experts in ten application areas to enable an accurate understanding, analysis, and control of complex systems; be they natural, engineered, or societal:

The authors explain how DDDAS unifies the computational and instrumentation aspects of an application system, extends the notion of Smart Computing to span from the high-end to the real-time data acquisition and control, and manages Big Data exploitation with high-dimensional model coordination.

The Dynamically Data Driven Applications Systems (DDDAS) paradigm inspired research regarding the prediction of severe storms. Specifically, the DDDAS concept allows atmospheric observing systems, computer forecast models, and cyberinfrastructure to dynamically configure themselves in optimal ways in direct response to current or anticipated weather conditions. In so doing, all resources are used in an optimal manner to maximize the quality and timeliness of information they provide.

Kelvin Droegemeier, Regents’ Professor of Meteorology at the University of Oklahoma; former Director of the White House Office of Science and Technology Policy

We may well be entering the golden age of data science, as society in general has come to appreciate the possibilities for organizational strategies that harness massive streams of data. The challenges and opportunities are even greater when the data or the underlying system are dynamic - and DDDAS is the time-tested paradigm for realizing this potential.

Sangtae Kim, Distinguished Professor of Mechanical Engineering and Distinguished Professor of Chemical Engineering at Purdue University

✦ Table of Contents

Contents
About the Editors
1 Introduction to the Dynamic Data Driven Applications Systems (DDDAS) Paradigm
1.1 Introduction
1.2 What Is DDDAS?
1.3 State Estimation and Data Assimilation
1.3.1 DDDAS and Adaptive State Estimation
1.3.2 Does DDDAS Use Feedback Control?
1.4 DDDAS Methods
1.5 DDDAS Research Areas of Historical Development
1.5.1 Theory: Modeling and Analysis
1.5.2 Methods: Domain Applications
1.5.3 Analysis and Design: Systems and Architectures
1.6 Book Overview
1.7 DDDAS Future
1.8 Summary
References
Part I Measurement-Aware: Data Assimilation, Uncertainty Quantification
2 Tractable Non-Gaussian Representations in Dynamic Data Driven Coherent Fluid Mapping
2.1 Introduction
2.1.1 Systems Dynamics and Optimization
2.1.2 Dynamically Deformable Reduced Models
2.1.3 Nonlinear High Dimensional Inference
2.2 Ensemble Learning in Mixture Ensembles
2.2.1 Mixture Ensemble Filter and Smoother
2.3 Nonlinear Filtering Must Reduce Total Variance
2.4 Ensemble Learning with a Stacked Cascade
2.4.1 Application Example
2.5 Information Theoretic Learning in Filtering
2.5.1 Tractable Information Theoretic Approach
2.6 Application Example
2.7 Conclusions
References
3 Dynamic Data-Driven Adaptive Observations in Data Assimilation for Multi-scale Systems
3.1 Introduction
3.2 Dimensional Reduction and Homogenization
3.3 Data Assimilation in Multi-scale Systems
3.4 Information-Theoretic Sensor Selection Strategy
3.4.1 The Linear Case
3.4.2 Information Flow for the Coarse Grained Dynamics
3.4.3 Finite-Time Lyapunov Exponents and Singular Vectors
3.4.4 Sensor Selection and the Lorenz 1963 Model
3.4.4.1 Sensor Selection with Kullback-Leibler Divergence
3.4.4.2 Sensor Selection with Singular Vectors
3.4.4.3 Influence of Singular Values in Discrete-Time, Linear Gaussian Case
3.4.4.4 Numerical Results
3.5 Conclusions
References
4 Dynamic Data-Driven Uncertainty Quantification via Polynomial Chaos for Space Situational Awareness
4.1 Introduction
4.2 Gaussian Mixture Models
4.3 Polynomial Chaos
4.4 Polynomial Chaos with Gaussian Mixture Models
4.5 Global Ionosphere-Thermosphere Model
4.6 Results
4.6.1 Orbital Uncertainty Quantification
4.6.2 Initial Results for Atmospheric Density Forecasting
4.7 Conclusion
References
Part II Signals-Aware: Process Monitoring
5 Towards Learning Spatio-Temporal Data Stream Relationships for Failure Detection in Avionics
5.1 Introduction
5.2 Background
5.2.1 Error Detection and Correction Methods
5.2.2 Spatio-Temporal Data Stream Processing System
5.3 Design of Machine Learning Component
5.3.1 Prediction in PILOTS Programming Language
5.3.2 Prediction in PILOTS Runtime
5.4 Data-Driven Learning of Linear Models
5.4.1 Learning Algorithm
5.4.2 Linear Model Accuracy
5.5 Statistical Learning of Dynamic Models
5.5.1 Offline Supervised Learning
5.5.1.1 Gaussian Naïve Bayes Classifiers
5.5.1.2 Offline Learning Phase
5.5.2 Dynamic Online Unsupervised Learning
5.5.2.1 Major and Minor Modes
5.5.2.2 Online Learning Phase
5.6 Case Study: Airplane Weight Estimation
5.6.1 Experimental Settings
5.6.1.1 Data Generation
5.6.1.2 Implementation and Evaluation of Learning Algorithms
5.6.2 Aerodynamic Model Parameter Estimation by Linear Regression
5.6.2.1 Assumption
5.6.2.2 Linear Regression Model
5.6.3 Error Detection and Correction Using Error Signatures
5.6.3.1 PILOTS Program
5.6.3.2 Error Detection
5.6.3.3 Software Parameter Settings
5.6.3.4 Results
5.6.4 Error Detection Using the Dynamic Bayes Classifier
5.6.4.1 PILOTS Program
5.6.4.2 Mode Prediction Evaluation
5.6.4.3 Experimental Settings
5.6.4.4 Results
5.6.5 Comparison Between Error Signatures and Dynamic Bayes Classifier
5.7 Related Work
5.8 Discussion and Future Work
References
6 Markov Modeling via Spectral Analysis: Application to Detecting Combustion Instabilities
6.1 Motivation and Introduction
6.2 Background and Mathematical Preliminaries
6.3 Proposed Approach
6.3.1 Estimation of Reduced-Order Markov Model
6.3.2 Estimation of Parameters for the Reduced-Order Markov Model
6.3.3 Pseudocode of the Main Algorithm
6.4 Combustion Experiment Details
6.5 Results and Discussion
6.6 Conclusions and Future Work
References
7 Dynamic Space-Time Model for Syndromic Surveillance with Particle Filters and Dirichlet Process
7.1 Background
7.2 Methodology
7.2.1 Dirichlet Process
7.2.2 Particle Filter
7.2.3 Evaluation of the Method
7.3 Applications with Indiana Surveillance Data
7.4 Conclusions and Future Work
References
Part III Structures-Aware: Health Modeling
8 A Computational Steering Framework for Large-Scale Composite Structures: Part I—Parametric-Based Design and Analysis
8.1 Introduction
8.2 Elements of the DISCERN Framework
8.2.1 Parametric-Based Design and Interactive Visual Programming
8.2.2 Analysis
8.2.2.1 Modeling of Thin-Shell Composites Using Isogeometric Analysis
8.2.2.2 The Structural Health Monitoring (SHM) System
8.3 NREL Phase VI Wind Turbine Blade
8.3.1 Setting Material Properties, Loads, and Boundary Conditions
8.3.2 Simulation Results
8.3.3 Visualization of IGA Results
8.3.4 Parametric Design Modification
8.4 Conclusions
References
9 Development of Intelligent and Predictive Self-Healing Composite Structures Using Dynamic Data-Driven Applications Systems
9.1 Introduction
9.1.1 Overview of the Proposed Intelligent Structure
9.2 Experimental Section
9.2.1 Double-Cantilever Beam (DCB) Test Specimen Fabrication
9.2.1.1 Materials
9.2.2 Manufacturing of DCB Test Specimens
9.2.3 Fracture and Healing Protocols
9.2.4 Fracture Analysis
9.3 Results and Discussions
9.3.1 Fracture Test Results
9.3.2 Quantification of Healing Efficiency
9.3.3 Fractography Using Scanning Electron Microscopy (SEM)
9.3.4 Parametric Sensitivity Analysis
9.4 Concluding Remarks
References
10 Dynamic Data-Driven Approach for Unmanned Aircraft Systems Aero-elastic Response Analysis
10.1 Introduction
10.2 Framework
10.3 Aero-Elastic Simulation
10.4 Data-Driven Prediction Framework
10.5 Case Study
10.6 Decision Support
10.7 Concluding Remarks
References
Part IV Environment-Aware: Earth, Biological, and Space Systems
11 Transforming Wildfire Detection and Prediction Using New and Underused Sensor and Data Sources Integrated withModeling
11.1 Introduction
11.2 Background
11.2.1 Forecasting Approaches
11.2.2 The 2015 Canyon Creek Wildfire Complex
11.3 Methods
11.3.1 Wildland Fire Detection, Mapping, and Monitoring
11.3.1.1 The Visible and Infrared Imaging Radiometer Suite (VIIRS)
11.3.1.2 Landsat
11.3.2 Coupled Weather-Wildland Fire Modeling
11.4 Experiment Design and Results
11.4.1 Dynamic Data Driven Model Invocation
11.4.2 Results: Impact on Fire Detection
11.4.3 Results: Impact on Fire Prediction
11.4.4 Integrated Results
11.5 Discussion
11.6 Conclusions
References
12 Dynamic Data Driven Application Systems for Identification of Biomarkers in DNA Methylation
12.1 Introduction
12.2 DNA Methylation Data
12.3 Proposed DDDAS-Based Learning Framework (3D-HCL)
12.3.1 Initialization Algorithm: Principal Component Analysis
12.3.2 Clustering Algorithm: Hierarchical Clustering
12.3.3 Orchestration Procedure: Cluster Membership Score Based Algorithm
12.3.4 Outlier Detection Algorithm
12.3.5 Dimension Reduction Algorithm: Locus Information Score Based Algorithm
12.4 Results and Discussion
12.4.1 Learning from Training Data
12.4.2 Learning from Test Data
12.5 Conclusion
References
13 Photometric Stereopsis for 3D Reconstruction of Space Objects
13.1 Introduction
13.2 Problem Statement and Background
13.3 Photometric Stereo
13.3.1 Formulation
13.3.2 Modified Photometric Stereo
13.3.3 Surface Reconstruction and Depth Estimation
13.4 Photometric stereo In Motion
13.5 Covariance Analysis
13.5.1 Raw Sensor Noise and the Intensity Uncertainty
13.5.2 Covariance of the Normal Vector Estimates
13.5.3 Error Covariance of The Surface Points
13.6 Simulation and Experiment
13.6.1 Stationary Observation of Lambertian Surface
13.6.2 Observation of Lambertian Surface From Non-Stationary View Point
13.6.3 Observation of Non-Lambertian Surface From Non-Stationary View Point
13.7 Conclusion
References
Part V Situation Aware: Tracking Methods
14 Aided Optimal Search: Data-Driven Target Pursuit from On-Demand Delayed Binary Observations
14.1 Introduction
14.2 Related Work
14.3 Notation and Preliminaries
14.4 Problem Statement
14.5 Target Trajectory Estimation Via Sparse Gaussian Mixture Model
14.5.1 Prediction
14.5.2 Update
14.6 Ground-Sensor-Aided Search Via Mixed-Integer Convex Programming
14.6.1 Ground-Sensor-Aided Optimal Search
14.6.2 Sampling-Based Aided Search via Convex Mixed-Integer Programming
14.7 Numerical Experiments
14.7.1 The Scenario
14.7.2 The Algorithms: Implementation Details
14.7.3 A Typical Result
14.7.4 Monte Carlo Analysis
14.8 Conclusion
A Prediction Equations
B Update Equations
References
15 Optimization of Multi-target Tracking Within a Sensor Network Via Information Guided Clustering
15.1 Introduction
15.2 Target Tracking and Motivating Problem
15.2.1 Problem Formulation
15.2.1.1 Prediction
15.2.1.2 Update
15.2.2 Motivation: The Euclidean Cluster
15.3 Information Guided Rapid Clustering Algorithm
15.3.1 Sensing Feasibility
15.3.1.1 A Note on Feasibility
15.3.2 Information Utility
15.3.2.1 Mahalanobis Distance
15.3.2.2 Sensing Mode and Quality Cluster Creation
15.3.3 Communication Cost
15.3.4 Final Optimal Sensor Selection
15.3.4.1 A Note on the Use of Two Separate Information Gain Metrics
15.3.5 Procedure of the IGRCA
15.4 Target Dynamics and Sensor Measurement Models
15.4.1 System Model
15.4.2 Measurement Model
15.5 Analysis
15.5.1 Frame of Reference
15.5.2 Extension to Bearing Sensors
15.6 Simulation Results
15.6.1 Performance Metrics
15.6.2 Algorithm Win Probability: Simulation
15.6.3 Multiple Target Tracking
15.6.3.1 Comparative Posterior CRLB: Various Sensor Densities
15.6.3.2 Computational Expenditure
15.6.3.3 Comparative Posterior CRLB: Various Number of Targets
15.7 Conclusion
References
16 Data-Driven Prediction of Confidence for EVAR in Time-Varying Datasets
16.1 Introduction
16.2 Preliminaries
16.2.1 Probabilistic Inequalities
16.2.2 Lévy Process
16.2.3 Information Gain and Exploration
16.2.4 Entropic Value at Risk (EVAR) Risk Measure
16.3 Formulation: Exploration as Multi-play N-Armed Restless Bandits
16.3.1 Multi-play N-Armed Restless Bandit Formulation
16.3.2 Data-Driven EVAR
16.3.2.1 Requirement for Predicting Data-Driven EVAR
16.3.3 Modeling the Information Gain
16.3.3.1 Poisson Exposure Distribution (Ped) Likelihood
16.3.3.2 Poisson Exposure Process (Pep) Model
16.4 Algorithms and Probabilistic Guarantees
16.4.1 The Time-Invariant Case: Ped-Based Exploration
16.4.2 The Time-Varying Case: Pep-Based Exploration
16.5 Experimental Evaluation
16.5.1 Evaluation Setup
16.5.2 Assumptions on Prior and Posterior Distributions
16.5.3 EVAR-Seeking Algorithms and Results
16.6 Conclusion
References
Part VI Context-Aware: Coordinated Control
17 DDDAS for Attack Detection and Isolation of Control Systems
17.1 Introduction
17.2 Problem Formulation
17.2.1 Cyber-Attacks in Control Systems
17.3 DDDAS Anomaly Isolation and Response
17.3.1 Anomaly Detection
17.3.2 Anomaly Isolation
17.4 Obtaining a Simulation Model
17.5 Case Study
17.5.1 Description of the System
17.5.2 Obtaining the Model System from I/O Data
17.5.3 Detection of Sensor Attacks
17.5.4 Isolation of Sensor Attacks
17.6 Conclusions and Future Work
References
18 Approximate Local Utility Design for Potential Game Approach to Cooperative Sensor Network Planning
18.1 Introduction
18.2 Background
18.2.1 Information Measures
18.2.1.1 Entropy
18.2.1.2 Mutual Information
18.2.2 Game-Theoretic Architecture
18.2.2.1 Strategic Form Game
18.2.2.2 Potential Game
18.3 Sensor Network Planning as a Potential Game
18.3.1 Cooperative Sensor Planning for Maximum Information
18.3.2 Sensor Selection as Potential Game
18.4 Approximate Local Utility Design
18.4.1 Neighbors with Correlation
18.4.2 Determination of the Neighbor Set
18.4.3 Computation Time Analysis
18.5 Numerical Example
18.5.1 Sensor Targeting for Weather Forecast
18.5.2 Comparative Results
18.6 Conclusion
References
19 Dynamic Sensor-Actor Interactions for Path-Planning in a Threat Field
19.1 Introduction
19.1.1 Literature Review
19.1.2 Proposed Work and Contributions
19.2 Problem Formulation
19.3 Actor-Driven Sensor Reconfiguration
19.4 Results and Discussion of Numerical Simulation Experiments
19.5 Conclusions
References
Part VII Energy-Aware: Power Systems
20 Energy-Aware Dynamic Data-Driven Distributed Traffic Simulation for Energy and Emissions Reduction
20.1 Introduction
20.2 Research Components: Models and Architecture
20.2.1 Cellular Automata Modeling
20.2.1.1 Methodology
20.2.2 Bayesian Inference Approach
20.3 Energy and Emission Modeling with MOVES-Matrix
20.4 Distributed Simulation Middleware
20.4.1 G-RTI Architecture
20.4.2 Energy Consumption Measurements
20.5 Concluding Comments and Future Work
References
21 A Dynamic Data-Driven Optimization Framework for Demand Side Management in Microgrids
21.1 Introduction
21.2 Proposed Framework
21.2.1 Simulation Module
21.2.2 Optimization Module
21.2.3 ε-Constraint Method for Multi-objective Optimization
21.2.4 Real-Time Decision Making Module
21.3 Experiments and Results
21.4 Conclusion
References
22 Dynamic Data Driven Partitioning of Smart Grid for Improving Power Efficiency by Combinining K-Means and Fuzzy Methods
22.1 Introduction
22.2 Methodology
22.2.1 k-means Partitioning Algorithm
22.2.2 Fuzzy Logic Decision Making Model
22.3 Simulations Configuration
22.4 Simulation Results
22.5 Conclusion
References
Part VIII Process-Aware: Image and Video Coding
23 Design of a Dynamic Data-Driven System for Multispectral Video Processing
23.1 Introduction
23.2 Related Work
23.3 Lightweight Dataflow (LD) Spectral
23.4 Run-Time System Model
23.5 Band Subset Processing
23.6 Band Subset Selection
23.7 Experimental Results
23.7.1 Experimental Setup
23.7.2 Accuracy Metric
23.7.3 Example Images
23.7.4 Accuracy Evaluation
23.7.5 Execution Time Evaluation
23.8 Conclusions
References
24 Light Field and Plenoptic Point Cloud Compression
24.1 Lenslet-Based Light Field Image Compression
24.1.1 Self-similarity-Based Light Field Image Compression
24.1.2 Pseudo-Sequence-Based Light Field Image Compression
24.1.2.1 The 2-D Hierarchical Coding Structure
24.1.2.2 Distance-Based Reference Frame Selection and Motion Vector Scaling
24.1.3 Dictionary-Learning-Based Light Field Image Compression
24.2 Camera-Array-Based Light Field Image Compression
24.2.1 Compression of Dense-Camera-Array-Based Light Field Images with Obvious Perspective Motions
24.2.1.1 The 2-D Hierarchical Coding Structure
24.2.1.2 Global Perspective Model
24.2.1.3 Local Four-Parameter Affine Motion Model
24.2.2 Compression of Dense-Camera-Array-Based Light Field Images with Translational Motions
24.3 Surface Light Field Image Compression
24.3.1 Interpolation-Based Surface Light Field Image Compression
24.3.2 Plenoptic Point Cloud Compression
24.4 Conclusion
References
25 On Compression of Machine-Derived Context Sets for Fusion of Multi-modal Sensor Data
25.1 Introduction
25.2 Learning Context from Data
25.3 Cardinality Reduction of Context Sets
25.3.1 Graph-Theoretic Compression
25.3.2 Compression by Subset Selection
25.4 Experiments and Results
25.5 Conclusion
References
Part IX Cyber-Aware: Security and Computing
26 Simulation-Based Optimization as a Service for Dynamic Data-Driven Applications Systems
26.1 Introduction
26.2 Problem Statement and Overview of SBOaaS
26.2.1 Motivating Case Study: Dynamic Traffic Light Control System
26.2.2 DDDAS-Specific Problem Statement and the SBOaaS Approach
26.2.3 Key Features of SBOaaS
26.3 Anytime Optimization Using Parallel Greedy Algorithm
26.3.1 Coordinate Greedy
26.3.2 K-Coordinate Greedy
26.3.3 Adaptive K-Coordinate Greedy
26.4 System Architecture
26.4.1 Runtime Architecture
26.4.2 Design Time Architecture
26.4.3 User Interaction Framework
26.5 Evaluation
26.5.1 Online Simulation-Based Optimization for Dynamic Traffic Light Control System
26.5.1.1 Environment
26.5.1.2 Experiment 1
26.5.1.3 Results
26.5.1.4 Experiment 2
26.5.1.5 Results
26.5.2 System Evaluation
26.5.2.1 Result
26.6 Related Work
26.6.1 Coordinate Greedy Algorithm
26.6.2 Cloud-Based Simulation Service
26.6.3 Traffic Light Optimal Control Problem
26.7 Conclusions
References
27 Privacy and Security Issues in DDDAS Systems
27.1 Introduction
27.2 Background
27.3 Overview and Goals
27.4 Conceptual PREDICT Model
27.4.1 System Model
27.4.2 Privacy Model
27.5 PREDICT Framework: Technical Approaches and Results
27.5.1 Privacy Preserving Data Collection and Data Aggregation with Feedback Control
27.5.2 Dynamic Data Modeling with Uncertainty Quantification
27.5.3 Secure Data Aggregation and Feedback Control Without Trusted Aggregator
27.6 Conclusion
References
28 Multimedia Content Analysis with Dynamic Data Driven Applications Systems (DDDAS)
28.1 Introduction
28.2 Multimedia Analysis
28.2.1 QuEST
28.2.2 Unexpected Query
28.3 Multimedia Contextual Reality
28.4 Modeling for Multimedia Content
28.4.1 Data Oriented Models (Cyber)
28.4.2 Analytical Models (Physical)
28.4.3 Cyber-Physical Models
28.5 Results: Activity Analysis
28.5.1 Interface
28.5.2 Case 1: Intersection
28.5.3 Case 2: Parking Lot
28.6 Conclusions
References
Part X Systems-Aware: Design Methods
29 Parzen Windows: Simplest Regularization Algorithm
29.1 Introduction
29.2 Related Work
29.3 Regularized Least Squares Method
29.4 Approximate Regularized Least Squares
29.5 Error Bound for SR
29.6 Bias, Variance and Regularization Constant
29.6.1 Regularization Constant
29.6.2 Regularization Constant and Simplest Regularization
29.7 Computational Complexity
29.8 SR and Parzen Windows
29.9 Experiments
29.9.1 Simulated Data Experiment
29.9.2 Real Data Experiment
29.9.2.1 Methods
29.9.2.2 Data Sets
29.9.2.3 Experimental Results
29.10 Summary
References
30 Multiscale DDDAS Framework for Damage Prediction in Aerospace Composite Structures
30.1 Introduction
30.2 The Multiscale DDDAS Framework
30.3 Computational Structural Model
30.3.1 Progressive Damage Model
30.4 Fatigue Damage Simulation of a Full-Scale CX-100 Wind Turbine Blade Driven by Test Data
30.4.1 Blade Structure and Its IGA Model
30.4.2 Blade Fatigue-Test Setup and Sensor Layout
30.4.3 Blade Fatigue Simulation Driven by Test Data
30.5 Numerical Simulation of the Orion UAV
30.5.1 Parametric UAV Model
30.5.2 Landing Simulation
30.6 Conclusions
References
31 A Dynamic Data-driven Stochastic State-Awareness Framework for the Next Generation of Bio-inspired Fly-by-feel Aerospace Vehicles
31.1 Introduction
31.2 Bio-inspired Sensor Networks and Wing Integration
31.2.1 The Composite Wing
31.3 The Wind Tunnel Experimental Process
31.3.1 The Wind Tunnel
31.3.2 The Experiments
31.4 Stochastic Global Identification Under Multiple Flight States
31.4.1 Baseline Modeling Under a Single Flight State
31.4.2 Global Modeling Under Multiple Flight States
31.5 Results
31.5.1 Numerical Simulations
31.5.2 Non-parametric Analysis
31.5.3 Baseline Parametric Modeling
31.5.4 Global Modeling Under Multiple Flight States
31.6 Concluding Remarks
References
32 The Future of DDDAS
32.1 *-5pt
32.2 DDDAS Has Universal Appeal
32.2.1 Paradigm for Theory-Data Symbiosis
32.2.2 Mitigates the Curse of Dimensionality
32.2.3 A Prediction and Discovery Instrument
32.3 Emerging Opportunities
32.3.1 Applications Systems
32.3.2 Instrumentation
32.3.3 Modeling and Simulation Methodology
32.3.4 Systems Software Computation
32.4 Example: Hurricane Prediction
32.5 Conclusions
References
Index