Title Page
Contents
NOMENCLATURES 11
ABSTRACT 15
제1장 INTRODUCTION 17
1.1. Background and State of the Art 17
1.2. Research Objectives 18
1.3. Structure of the Paper 18
제2장 LITERATURE REVIEW 22
2.1. Framework and Stages of Decision Support 22
2.2. Unique Maritime Transportation Environments 24
2.3. Key Attributes of Vessel Operation 26
제3장 GENERAL RESEARCH METHODOLOGY 33
3.1. Research Overview 33
3.2. Data Acquisition 34
3.3. Data Pre-processing 34
3.4. Feature Engineering 35
3.5. Analysis (Tool) 36
제4장 VARIABLE ASSOCIATION AND DECISION SUPPORT MODELS FOR VESSEL PREDICTIVE MACHINERY MAINTENANCE (STUDY 1) 37
4.1. Abstract 37
4.2. Introduction 38
4.3. Types of Maintenance 39
4.4. Acquired Dataset 42
4.5. Data Variance 47
4.6. Variable Correlation 49
4.7. Stepwise Regression Technique 49
4.8. Multivariate Regression Development Technique 50
4.9. Variable Association Case Study 50
4.9.1. Data Description 50
4.9.2. Correlation Coefficient 51
4.9.3. Feature Selection 51
4.9.4. Feature Rank 53
4.9.5. Stepwise Regression 55
4.9.6. Decision Support Case Study 56
4.10. Discussion 59
제5장 NAVIGATION TRAFFIC ELEMENT FEATURES FOR DECISION SUPPORT MODELS OF VESSEL NAVIGATION ANOMALY BEHAVIOR BASED ON WATER AREA (STUDY 2) 61
5.1. Abstract 61
5.2. Introduction 62
5.3. Dataset Acquired 63
5.4. Dataset Parameter Setting 64
5.5. AIS Data Pre-processing 65
5.6. Feature Selection 65
5.7. Traffic Element Features 66
5.7.1. Ship's Course 66
5.7.2. Ship's Speed 67
5.7.3. Rate of Course Change 67
5.7.4. Acceleration and Deceleration 68
5.7.5. Distance from Obstacles 68
5.7.6. Course Alteration Frequency 69
5.7.7. Deviation from Trend Trajectories 69
5.8. Sliding Window and Sampling 71
5.9. Trajectory Clustering 71
5.9.1. Principal Component Analysis (PCA) 71
5.9.2. Optimal Value of K using Elbow Method 71
5.9.3. K-Means 72
5.9.4. Predictor Importance 72
5.10. Case Study #1 74
5.10.1. Vessel Trajectory Extraction 74
5.10.2. Extracted Feature-Rate of Course Change 77
5.10.3. Extracted Feature-Acceleration & Deceleration 78
5.10.4. Extracted Feature-Course Alteration Frequency 79
5.10.5. Extracted Feature-Distance from Obstacles 80
5.10.6. Water Area 81
5.10.7. Deviation from Trend Trajectories 82
5.10.8. Trajectory Point Cluster Analysis 83
5.11. Case Study #2 87
5.10.1. Vessel Trajectory Extraction 87
5.10.2. Extracted Feature-Rate of Course Change 90
5.10.3. Extracted Feature-Acceleration & Deceleration 91
5.10.4. Extracted Feature-Course Alteration Frequency 92
5.10.5. Extracted Feature-Distance from Obstacles 93
5.10.6. Water Area 94
5.10.7. Deviation from Trend Trajectories 95
5.10.8. Trajectory Point Cluster Analysis 96
5.12. Case Study #3 100
5.11.1. Vessel Trajectory Extraction 100
5.11.2. Extracted Feature-Rate of Course Change 103
5.11.3. Extracted Feature-Acceleration & Deceleration 104
5.11.4. Extracted Feature-Course Alteration Frequency 105
5.11.5. Extracted Feature-Distance from Obstacles 106
5.11.6. Water Area 107
5.11.7. Deviation from Trend Trajectories 108
5.11.8. Trajectory Point Cluster Analysis 109
5.13. Discussion 113
제6장 AFTERWORD 115
6.1. Summary of Thesis Findings 115
6.2. Limitations and Future Works 116
REFERENCES 118
부록 126
APPENDIX A. Major Maintenance Log Data (Period of 2020)[원문불량;p.126] 126
APPENDIX B. Vessel Operation NMEA Data 127
APPENDIX C. Vessel Navigation AIS Data 128
APPENDIX D. GIS Mapping Data 129
Table 1-1. Overview of aims, datasets, methods, and limitation designed for this multi-study data-driven decision support models for vessel operation 20
Table 2-1. Key Attributes of Vessel Operation 31
Table 4-1. Categorization of Maintenance 40
Table 4-2. Acquired Dataset to Build the Decision Support Models for Vessel Predictive Maintenance 44
Table 4-3. Selected Predictors and Response for Gearbox Failure Case Study 52
Table 4-4. Stepwise Regression Result for Gearbox Failure Case Study 55
Table 5-1. Acquired Dataset to Build the Decision Support Models for Vessel Anomaly Navigation Behavior Detection 63
Table 5-2. Traffic Element Features Designed for Decision Support Models for Vessel Anomaly Navigation Behavior Detection Model 70
Table 5-3. Descriptive Statistics of Extracted Features for 1st Case Study[이미지참조] 86
Table 5-4. Descriptive Statistics of Extracted Features for 2nd Case Study[이미지참조] 99
Table 5-5. Descriptive Statistics of Extracted Features for 3rd Case Study[이미지참조] 112
Figure 2-1. Framework of Decision Support 22
Figure 2-2. Stages of Decision Support 23
Figure 3-1. General Research Methodology 33
Figure 4-1. Details of Ship and Sensors: (a) 295 G/T Coastal Passenger Ship, (b) Location of Temporary-installed Multisensors 46
Figure 4-2. Data Variance of Vessel Operation Streamed Raw Historical Data: (a) External Forces, (b) Machinery and Systems, (c) Navigation 48
Figure 4-3. Variable Correlation of Vessel Operation Streamed Raw Historical Data 51
Figure 4-4. Predictor Importance Rank Results for Gearbox Failure Case Study using:... 54
Figure 4-5. Goodness of Fit Result for Gearbox Failure Case Study 56
Figure 4-6. Polynomial Fitting Result for Gearbox Failure Case Study 57
Figure 4-7. Anomaly Detection based on Parameter Setting for Gearbox Failure Case Study 58
Figure 5-1. Initial Trajectory Extraction for 1st Case Study[이미지참조] 74
Figure 5-2. Trimmed Extracted Trajectory for 1st Case Study[이미지참조] 75
Figure 5-3. Cleaned Extracted Trajectory for 1st Case Study[이미지참조] 76
Figure 5-4. Rate of Course Change Value for 1st Case Study[이미지참조] 77
Figure 5-5. Acceleration and Deceleration Value for 1st Case Study[이미지참조] 78
Figure 5-6. Course Alteration Indication for 1st Case Study[이미지참조] 79
Figure 5-7. Nearest Points to Obstacles (Coastlines) for 1st Case Study[이미지참조] 80
Figure 5-8. Water Area Indication for 1st Case Study[이미지참조] 81
Figure 5-9. Anomaly-labelled Vessel Trajectory which Deviates from Trend Trajectories for 1st Case Study[이미지참조] 82
Figure 5-10. Trajectory Point Clusterization for 1st Case Study: (a) Scatter Plot of Clusters, (b) Trajectory Point Drawn based on Clusters 83
Figure 5-11. Dynamic Feature/Predictor Importance for 1st Case Study[이미지참조] 85
Figure 5-12. Initial Trajectory Extraction for 2nd Case Study[이미지참조] 87
Figure 5-13. Trimmed Extracted Trajectory for 2nd Case Study[이미지참조] 88
Figure 5-14. Cleaned Extracted Trajectory for 2nd Case Study[이미지참조] 89
Figure 5-15. Rate of Course Change Value for 2nd Case Study[이미지참조] 90
Figure 5-16. Acceleration and Deceleration Value for 2nd Case Study[이미지참조] 91
Figure 5-17. Course Alteration Indication for 2nd Case Study[이미지참조] 92
Figure 5-18. Nearest Points to Obstacles (Coastlines) for 2nd Case Study[이미지참조] 93
Figure 5-19. Water Area Indication for 2nd Case Study[이미지참조] 94
Figure 5-20. Anomaly-labelled Vessel Trajectory which Deviates from Trend Trajectories for 2nd Case Study[이미지참조] 95
Figure 5-21. Trajectory Point Clusterization for 2nd Case Study: (a) Scatter Plot of Clusters, (b) Trajectory Point Drawn based on Clusters[이미지참조] 96
Figure 5-22. Dynamic Feature/Predictor Importance for 2nd Case Study[이미지참조] 97
Figure 5-23. Initial Trajectory Extraction for 3rd Case Study[이미지참조] 100
Figure 5-24. Trimmed Extracted Trajectory for 3rd Case Study[이미지참조] 101
Figure 5-25. Cleaned Extracted Trajectory for 3rd Case Study[이미지참조] 102
Figure 5-26. Rate of Course Change Value for 3rd Case Study[이미지참조] 103
Figure 5-27. Acceleration and Deceleration Value for 3rd Case Study[이미지참조] 104
Figure 5-28. Course Alteration Indication for 3rd Case Study[이미지참조] 105
Figure 5-29. Nearest Points to Obstacles (Coastlines) for 3rd Case Study[이미지참조] 106
Figure 5-30. Water Area Indication for 3rd Case Study[이미지참조] 107
Figure 5-31. Anomaly-labelled Vessel Trajectory which Deviates from Trend Trajectories for 3rd Case Study[이미지참조] 108
Figure 5-32. Trajectory Point Clusterization for 3rd Case Study: (a) Scatter Plot of Clusters, (b) Trajectory Point Drawn based on Clusters[이미지참조] 109
Figure 5-33. Dynamic Feature/Predictor Importance for 3rd Case Study[이미지참조] 111