This book introduces some recent developments in Bayesian real-time system identification. It contains two different perspectives on data processing for system identification, namely centralized and distributed. A centralized Bayesian identification framework is presented to address challenging problems of real-time parameter estimation, which covers outlier detection, system, and noise parameters tracking. Besides, real-time Bayesian model class selection is introduced to tackle model misspecification problem. On the other hand, a distributed Bayesian identification framework is presented to handle asynchronous data and multiple outlier corrupted data. This book provides sufficient background to follow Bayesian methods for solving real-time system identification problems in civil and other engineering disciplines. The illustrative examples allow the readers to quickly understand the algorithms and associated applications. This book is intended for graduate students and researchers in civil and mechanical engineering. Practitioners can also find useful reference guide for solving engineering problems.
Author(s): Ke Huang, Ka-Veng Yuen
Publisher: Springer
Year: 2023
Language: English
Pages: 285
City: Singapore
Preface
Contents
Nomenclature
1 Introduction
1.1 Dynamical Systems
1.1.1 Time-Invariant Systems
1.1.2 Time-Varying Systems
1.2 System Identification
1.2.1 Problems in System Identification
1.2.2 Real-Time System Identification
1.2.3 Bayesian System Identification
1.3 Uncertainty
1.4 Organization of the Book
References
2 System Identification Using Kalman Filter and Extended Kalman Filter
2.1 Introduction
2.2 Standard Kalman Filter
2.2.1 Derivation of the Discrete-Time Kalman Filter
2.3 Applications to State Estimation
2.3.1 Vehicle Tracking Problem
2.3.2 Sixty-Story Building
2.4 Extended Kalman Filter
2.4.1 Derivation of the Extended Kalman Filter
2.4.2 Extended Kalman Filter with Fading Memory
2.5 Application to State Estimation and Model Parameter Identification
2.5.1 Single-Degree-of-Freedom System
2.5.2 Three-Pier Bridge
2.5.3 Bouc-Wen Hysteresis System
2.6 Application to a Field Inspired Test Case: The Canton Tower
2.6.1 Background Information
2.6.2 Identification of Structural States and Model Parameters
2.7 Extended Readings
2.8 Concluding Remarks
References
3 Real-Time Updating of Noise Parameters for System Identification
3.1 Introduction
3.2 Real-Time Updating of Dynamical Systems and Noise Parameters
3.2.1 Updating of States and Model Parameters
3.2.2 Updating of Noise Parameters
3.3 Efficient Numerical Optimization Scheme
3.3.1 Training Phase
3.3.2 Working Phase
3.3.3 Uncertainty Estimation of the Updated Noise Parameters
3.4 Applications
3.4.1 Bouc-Wen Hysteresis System
3.4.2 Three-Pier Bridge
3.5 Concluding Remarks
References
4 Outlier Detection for Real-Time System Identification
4.1 Introduction
4.2 Outlier Detection Using Probability of Outlier
4.2.1 Normalized Residual of Measurement
4.2.2 Probability of Outlier
4.3 Computational Efficiency Enhancement Techniques
4.3.1 Moving Time Window
4.3.2 Efficient Screening Criteria
4.4 Outlier Detection for Time-Varying Dynamical Systems
4.4.1 Training Stage
4.4.2 Working Stage
4.5 Applications
4.5.1 Outlier Generation
4.5.2 Single-Degree-of-Freedom Oscillator
4.5.3 Fourteen-Bay Truss
4.6 Concluding Remarks
References
5 Bayesian Model Class Selection and Self-Calibratable Model Classes for Real-Time System Identification
5.1 Introduction
5.2 Bayesian Real-Time Model Class Selection
5.3 Real-Time System Identification Using Predefined Model Classes
5.3.1 Parametric Identification with a Specified Model Class
5.3.2 Parametric Identification Using Multiple Model Classes
5.3.3 Parametric Identification Using the Most Plausible Model Class
5.3.4 Predefined Model Classes
5.4 Self-Calibratable Model Classes
5.4.1 Parameterization and Model Classes
5.4.2 Self-Calibrating Strategy
5.4.3 Procedure of the Real-Time System Identification with Self-Calibratable Model Classes
5.5 Hierarchical Interhealable Model Classes
5.5.1 Hierarchical Model Classes
5.5.2 Interhealing Mechanism
5.5.3 Triggering Conditions
5.5.4 Procedure of the Real-Time System Identification Using Hierarchical Interhealable Model Classes
5.6 Applications to Bayesian Real-Time Model Class Selection for System Identification
5.6.1 Identification of High-Rise Building with Predefined Model Classes
5.6.2 Identification of Bouc-Wen Nonlinear Hysteresis System with Self-Calibratable Model Classes
5.6.3 Identification of Three-Dimensional Truss Dome with Hierarchical Interhealable Model Classes
5.7 Concluding Remarks
References
6 Online Distributed Identification for Wireless Sensor Networks
6.1 Introduction
6.2 Typical Architectures of Wireless Sensor Network
6.2.1 Centralized Networks
6.2.2 Decentralized Networks
6.2.3 Distributed Networks
6.3 Problem Formulations
6.4 Compression and Extraction Technique at the Sensor Nodes
6.4.1 Compression and Extraction of the Updated State Vector
6.4.2 Compression and Extraction of the Covariance Matrix
6.5 Bayesian Fusion at the Central Station
6.5.1 The Product of Univariate Gaussian PDFs
6.5.2 The Product of Multivariate Gaussian PDFs
6.5.3 Fusion of the Compressed Local Information
6.6 Illustrative Examples
6.6.1 Example 1: Forty-Story Building
6.6.2 Example 2: Bridge with Two Piers
6.7 Concluding Remarks
References
7 Online Distributed Identification Handling Asynchronous Data and Multiple Outlier-Corrupted Data
7.1 Introduction
7.2 Online Distributed Identification Framework
7.2.1 Local Identification at the Sensor Nodes
7.2.2 Bayesian Fusion at the Central Station
7.3 Online Distributed Identification Using Asynchronous Data
7.4 Application to Model Updating of a Sixteen-Bay Truss
7.5 Hierarchical Outlier Detection
7.5.1 Local Outlier Detection at the Sensor Nodes
7.5.2 Global Outlier Detection at the Central Station
7.6 Application to Model Updating of a Forty-Story Building
7.7 Concluding Remarks
References
