Structural Health Monitoring by Time Series Analysis and Statistical Distance Measures (PoliMI SpringerBriefs)

This book conducts effective research on data-driven Structural Health Monitoring (SHM), and accordingly presents many novel feature extraction methods by time series analysis and signal processing, to extract reliable damage sensitive features from vibration responses. In this regard, some limitations of time series modeling are dealt with. For decision-making, innovative distance-based novelty detection techniques are presented to detect, locate, and quantify different damage scenarios. The performance of the presented methods is demonstrated via laboratory and full-scale structures along with several comparative studies. The main target audience of the book includes scholars, graduate students working on SHM via statistical pattern recognition in terms of feature extraction and classification for damage diagnosis under environmental and operational variations; it would also be beneficial for practicing engineers whose work involves these topics. Foreword 6 Preface 8 Contents 10 1 An Introduction to Structural Health Monitoring 14 1.1 Background and Motivation 14 1.2 Levels of SHM 15 1.3 Methods of SHM 16 1.3.1 Model-Driven Approaches 16 1.3.2 Data-Driven Approaches 17 1.4 Statistical Pattern Recognition 17 1.4.1 Operational Evaluation 18 1.4.2 Sensing and Data Acquisition 19 1.4.3 Feature Extraction 22 1.4.4 Statistical Decision-Making 23 1.5 Machine Learning 23 1.6 Environmental and Operational Variability 24 1.7 Aim and Scope of the Book 24 1.8 Organization of This Book 25 1.9 Conclusions 25 References 26 2 Feature Extraction in Time Domain for Stationary Data 29 2.1 Introduction 29 2.2 Types of Time Series Data 32 2.2.1 Stationary Versus Non-stationary Time Series 32 2.2.2 Linear Versus Non-linear Time Series 32 2.2.3 Univariate Versus Multivariate Time Series 33 2.2.4 Gaussian Versus Non-Gaussian Time Series 33 2.3 Types of Time-Invariant Linear Models 34 2.4 Model Identification Based on Engineering Aspect 35 2.5 Model Identification Based on Statistical Aspect 36 2.5.1 A Conventional Approach to Output-Only Conditions 36 2.5.2 An Automatic Approach to Output-Only Conditions 37 2.6 Proposed Order Selection Algorithms 39 2.6.1 Robust Order Selection by an Iterative Algorithm 40 2.6.2 Robust and Optimal Order Selection by a Two-Stage Iterative Algorithm 42 2.6.3 Robust and Optimal Order Selection by an Improved Two-Stage Algorithm 43 2.7 Parameter Estimation 46 2.8 Types of Feature Extraction via Time Series Analysis 47 2.9 Proposed RBFE Methods 47 2.9.1 A Developed RBFE Approach 47 2.9.2 A Fast RBFE Approach 50 2.10 Proposed Spectral-Based Method 52 2.11 Conclusions 53 References 54 3 Feature Extraction in Time-Frequency Domain for Non-Stationary Data 58 3.1 Introduction 58 3.2 Adaptive Time-Frequency Data Analysis Methods 60 3.2.1 Empirical Mode Decomposition 60 3.2.2 Ensemble Empirical Mode Decomposition 61 3.2.3 Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise 62 3.2.4 Selection of Noise Amplitude and Ensemble Number 63 3.3 A New Hybrid Feature Extraction Method by ICEEMDAN and ARMA 64 3.3.1 A Novel Automatic IMF Selection Approach 64 3.3.2 Proposed Hybrid Algorithm 65 3.4 Conclusions 66 References 66 4 Statistical Decision-Making by Distance Measures 69 4.1 Introduction 69 4.2 Statistical Distance Measure 71 4.3 Conventional Distance Approaches 71 4.3.1 General Problem 71 4.3.2 Kullback–Leibler Divergence 72 4.3.3 Mahalanobis-Squared Distance 72 4.3.4 Kolmogorov–Smirnov Test Statistic 73 4.4 Proposed Univariate Distance Methods 74 4.4.1 KLD with Empirical Probability Measure 74 4.4.2 Parametric Assurance Criterion 76 4.4.3 Residual Reliability Criterion 77 4.5 Proposed Multivariate Distance Methods 78 4.6 Proposed Hybrid Distance Method 80 4.6.1 Partition-Based Kullback–Leibler Divergence 80 4.6.2 PKLD-MSD 81 4.7 Proposed Multi-Level Distance Method 83 4.8 Threshold Estimation 86 4.8.1 Standard Confidence Interval Limit 86 4.8.2 Monte Carlo Simulation 87 4.9 Conclusions 87 References 88 5 Applications and Results 90 5.1 Introduction 90 5.2 Validation of the DRBFE, PAC, and RRC Methods by the LANL Laboratory Frame 91 5.2.1 Model Identification by Box-Jenkins Methodology 92 5.2.2 AR Order Selection by the Proposed Iterative Algorithm 93 5.2.3 AR Order Selection by the Proposed Two-Stage Iterative Algorithm 95 5.2.4 Feature Extraction by the CBFE and RBFE Approaches 97 5.2.5 Feature Extraction by the Proposed DRBFE Approach 98 5.2.6 Damage Diagnosis by PAC and RRC 99 5.3 Validation of the FRBFE and KLDEPM Methods by the IASC-ASCE Structure Under the Shaker Excitation 100 5.3.1 Feature Extraction by FRBFE 104 5.3.2 Damage Localization by KLDEPM 108 5.4 Validation of the ICEEMDAN-ARMA, SDC, and MDC Methods by the IASC-ASCE Structure Under Ambient Vibration 111 5.4.1 Parameter Selection and IMF Extraction from ICEEMDAN 113 5.4.2 Optimal IMF Extraction from MPF 113 5.4.3 ARMA Modeling 114 5.4.4 Damage Diagnosis by SDC and MDC 117 5.5 Validation of the Automatic Model Identification and PKLD-MSD Methods by the SMC Cable-Stayed Bridge 120 5.5.1 Automatic Model Identification 122 5.5.2 ARARX Modeling 125 5.5.3 The Conventional RBFE Approach 128 5.5.4 Early Damage Detection by PKLD-MSD 128 5.6 Validation of Spectral-Based and Multi-level Distance-Based Methods by the Wooden Truss Bridge 134 5.6.1 Response Modeling and Feature Extraction 135 5.6.2 Damage Detection Under Limited Sensors and Environmental Effects 137 5.7 Conclusions 138 References 139 6 Summary and Conclusions 140 6.1 Introduction 140 6.2 Main Conclusions of Feature Extraction 141 6.3 Main Conclusions of Damage Diagnosis 142 6.4 Further Developments 143 References 144