표제지
목차
Abstract 8
Nomenclature 20
제1장 서론 24
1.1. 연구 배경 및 필요성 24
1.2. 사후 정비와 예방 정비 28
1.3. 선박 예측 정비 관련 선행 연구 33
1.4. 연구 목적 36
제2장 연구 대상 및 방법 39
2.1. 연구 대상 39
2.1.1. 선박 39
2.1.2. AMS(Alarm monitoring system) 데이터 41
2.1.3. 선박 엔진 제어 특성 45
2.2. 연구 방법 49
2.2.1. 기계 학습(Machine learning) 49
2.2.2. 기계 학습 평가 지표 66
2.2.3. 엔진 시뮬레이션 및 알고리즘 개발 도구 68
2.2.4. 엔진 시뮬레이션 모델 72
2.2.5. Data preprocessing 알고리즘 73
2.2.6. 표준 데이터 분석 및 비정상 운전 데이터 검출 알고리즘 76
2.2.7. 엔진 이상 징후 및 정비 판단 인자 79
2.2.8. 이상 징후 검출 알고리즘 88
2.2.9. 추진 엔진 이상 징후 판단 알고리즘 90
2.2.10. 발전기 엔진 이상 징후 판단 알고리즘 98
2.2.11. 정비 시간 예측 알고리즘 104
제3장 연구 결과 및 고찰 109
3.1. 엔진 기계 학습 데이터 확보를 위한 시뮬레이션 검증 109
3.2. 표준 데이터 분석 알고리즘 검증 114
3.3. 비정상 운전 데이터 검출 알고리즘을 활용한 추진 엔진의 비정상 운전 데이터 분석 결과 119
3.4. 이상 징후 검출 알고리즘을 이용한 추진 엔진의 이상 징후 데이터 검증 결과 125
3.5. 추진 엔진 정비 예측 알고리즘 개발 131
3.5.1. 이상 징후 판단 알고리즘 검증 131
3.5.2. 정비 시간 예측 알고리즘을 이용한 정비 시간 예측 결과 157
3.6. 발전기 엔진 데이터 분석 및 예측 정비 알고리즘 개발 158
3.6.1. 이상 징후 판단 알고리즘 구축 및 결과 분석 158
3.6.2. 정비 기준 인자에 따른 정비 시간 예측 알고리즘 검증 결과 174
제4장 결론 182
4.1. 엔진 기계 학습 데이터 확보를 위한 시뮬레이션 검증 182
4.2. 표준 데이터 분석 및 비정상 운전 데이터 검출 알고리즘 검증 183
4.3. 추진 엔진 이상 징후 판단 알고리즘 검증 및 정비 시간 예측 알고리즘 개발 184
4.4. 발전기 엔진 데이터 분석 및 예측 정비 알고리즘 개발 및 결과 분석 185
4.5. 향후 계획 186
참고문헌 188
Table 1.1. Annual operating cost and shipbuilding cost of container ship 25
Table 1.2. Example of time based maintenance(TBM) item on ships 31
Table 2.1. Specifications of ship detailed 40
Table 2.2. List of engine main monitoring and alarm on the AMS 43
Table 2.3. Specifications of main engine 45
Table 2.4. Engine control stage 47
Table 2.5. Specifications of generator engine 48
Table 2.6. Principal algorithms of supervised learning 53
Table 2.7. Principal algorithms of unsupervised learning 55
Table 2.8. Algorithm principal evaluation indicators 68
Table 2.9. Examples of major programming language grammars 71
Table 2.10. Engine PMS item 80
Table 2.11. Abnormal issues related to exhaust gas temperature during engine operation 82
Table 2.12. Abnormal issues related to engine RPM and exhaust gas color during engine operation 84
Table 2.13. Turbocharger related issues during engine operation 86
Table 2.14. Criteria for setup normal operation state database of the propulsion engine 91
Table 2.15. Criterion for MCCV abnormality judgment of propulsion engine 92
Table 2.16. Criteria for setup normal operation state database of the GE 99
Table 2.17. Criterion for GCCV abnormality judgment of GE 100
Table 2.18. Research method and process 107
Table 3.1. Comparison data of principle data on test report and simulation model 110
Table 3.2. Conditions of change for simulation 112
Table 3.3. Results data of exhaust gas temperature rise simulation 113
Table 3.4. No. 1 Cylinder condition data 133
Table 3.5. Cylinder exhaust gas temperature analysis data 134
Table 3.6. Cylinder combustion pressure analysis data 134
Table 3.7. Data of MCCV analysis of normal operating condition on No. 1 cylinder 136
Table 3.8. Index frequency by ME ORDER RPM and VOYAGE 139
Table 3.9. Criteria data of normal operation condition of average on No. 1 cylinder 140
Table 3.10. Revision criteria factor of MCCV on No. 1 cylinder 144
Table 3.11. Regression coefficient of the revision factor on No. 1 cylinder 144
Table 3.12. Comparison of MCCV and RMCCV analysis data of all normal operating data on No. 1 cylinder 146
Table 3.13. Comparison of MCCV and RMCCV analysis data of average normal operation data based on ME ORDER... 147
Table 3.14. Comparison of MCCV and RMCCV analysis data in average normal operating condition based on VOYAGE... 150
Table 3.15. Revision factors and Result of analysis of all Cylinder MCCV correction factors 152
Table 3.16. RMCCV of average data based on ME ORDER RPM and VOYAGE of all cylinder 154
Table 3.17. RMCCV of average data based on VOYAGE of all cylinder 154
Table 3.18. Criterion for RMCCV abnormality judgment of propulsion engine 155
Table 3.19. Calculation factor for the MTRHR 157
Table 3.20. Analysis data of cylinder exhaust gas temperature on No. 2 GE 159
Table 3.21. Analysis data of normal operating condition GCCV on No. 2 GE No. 1 cylinder 162
Table 3.22. Regression coefficient of the GCCV revision factor on No. 2 GE No. 1 cylinder 165
Table 3.23. Revision criteria factor of GCCV on No. 2 GE No. 1 cylinder 165
Table 3.24. Comparison of GCCV and RGCCV analysis data of all normal operating data on the No. 2 GE No. 1 cylinder 166
Table 3.25. Comparison of GCCV and RGCCV analysis data of average normal operating data based on VOYAGE on... 167
Table 3.26. Revision factors and Result of analysis of all Cylinder GCCV correction factors on the No. 2 GE 170
Table 3.27. RGCCV of all Cylinder on the No. 2 GE 171
Table 3.28. RGCCV of average data based on VOYAGE of all cylinder on the No. 2 GE 171
Table 3.29. Criterion for RGCCV abnormality judgment of No. 2 GE 172
Table 3.30. Calculation factor for the MTRHRG 176
Table 3.31. Calculation factor for the MTRHE 179
Figure 1.1. Comparison of economic feasibility between building cost and operating cost for the ship 24
Figure 1.2. Current state of marine accidents by accident type 26
Figure 1.3. Current state of marine accidents by ship type 27
Figure 1.4. Type of corrective maintenance and preventive maintenance 29
Figure 2.1. Ship for sailing data collection 39
Figure 2.2. Screen of AMS for the propulsion engine 42
Figure 2.3. Engine telegraph 46
Figure 2.4. Configuration of generator engine 48
Figure 2.5. Machine learning 49
Figure 2.6. Classification and regression in supervised learning 50
Figure 2.7. Clustering in unsupervised learning 51
Figure 2.8. Dimensionality reduction of principal component analysis 57
Figure 2.9. K-nearest neighbors algorithm 58
Figure 2.10. K-means standard algorithm 60
Figure 2.11. Regression model 62
Figure 2.12. Difference of prediction result 63
Figure 2.13. Simple linear regression model 63
Figure 2.14. Residuals in linear regression 65
Figure 2.15. Module structure and characteristics of CRUISE™ M 69
Figure 2.16. Generator engine simulation model 72
Figure 2.17. Data preprocessing algorithm for the propulsion engine 74
Figure 2.18. Data preprocessing algorithm for the generator engine 76
Figure 2.19. Preprocessing data of the main engine 77
Figure 2.20. PCA/KNN algorithm for standard data analysis and abnormal operation data detection 78
Figure 2.21. PCA/K-means algorithm for anomaly symptom data detection 89
Figure 2.22. Correlation analysis table of No. 1 cylinder at all mode 94
Figure 2.23. Correlation analysis table of No. 1 cylinder at navigation full mode 95
Figure 2.24. Correlation analysis table of No.2 GE No. 1 cylinder 102
Figure 3.1. Comparison of principle data on test report and simulation model 109
Figure 3.2. Simulation exhaust port modeling factors 111
Figure 3.3. Results of exhaust gas temperature rise simulation 113
Figure 3.4. Results of PCA of preprocessing data of the propulsion engine 115
Figure 3.5. Results of PCA according to mode classification criteria of the preprocessing data 116
Figure 3.6. Results of abnormal data analysis of PCA/KNN algorithm 118
Figure 3.7. Results of abnormal data processing of PCA/KNN algorithm 119
Figure 3.8. Data for confirmation for engine operating condition 120
Figure 3.9. Results of PCA for normal operating data 120
Figure 3.10. Results of KNN analysis for normal operating data 121
Figure 3.11. Results of abnormal operating data analysis of PCA/KNN algorithm 123
Figure 3.12. Confirmation of abnormal operating data by PCA/KNN algorithm 124
Figure 3.13. Results of PCA of normal operating data for 8 voyages 126
Figure 3.14. Results of PCA for anomaly symptom data 127
Figure 3.15. Results of KNN analysis for anomaly symptom data 128
Figure 3.16. Results of PCA for 9 voyage data including anomaly symptom data 129
Figure 3.17. Results of K-Means elbow method 130
Figure 3.18. Results of K-Means clustering 130
Figure 3.19. No. 1 Cylinder condition 132
Figure 3.20. MCCV of normal operating condition on No. 1 cylinder 135
Figure 3.21. MCCV including anomaly symptom data on No. 1 cylinder 138
Figure 3.22. Results of regression analysis of exhaust gas temperature correction factor on No. 1 cylinder 142
Figure 3.23. Results of regression analysis of combustion maximum pressure correction factor on No. 1 cylinder 143
Figure 3.24. Comparison of MCCV and RMCCV of normal operating data on No. 1 cylinder 145
Figure 3.25. Comparison of MCCV and RMCCV of average of normal operating data based on ME ORDER RPM and VOYAGE... 147
Figure 3.26. RMCCV including anomaly symptom data on No. 1 cylinder 148
Figure 3.27. RMCCV of average of data including anomaly symptom based on ME ORDER RPM and VOYAGE of the No. 1 cylinder 149
Figure 3.28. RMCCV of average data including anomaly symptom based on VOYAGE of No. 1 cylinder 151
Figure 3.29. RMCCV of average data including anomaly symptom based on VOYAGE of all cylinder 156
Figure 3.30. Prediction result of maintenance time based on RMCCV of No. 1 cylinder 157
Figure 3.31. Exhaust gas temperature of No. 2 GE No. 1 cylinder 160
Figure 3.32. GCCV of normal operating condition on No. 2 GE No. 1 cylinder 161
Figure 3.33. GCCV including anomaly symptom data on No. 2 GE No. 1 cylinder 163
Figure 3.34. Results of regression analysis of correction factor on No. 2 GE No. 1 cylinder 164
Figure 3.35. Comparison of GCCV and RGCCV of normal operating data on No. 2 GE No. 1 cylinder 166
Figure 3.36. Comparison of GCCV and RGCCV of average of normal operating data based on VOYAGE of No. 2 GE No. 1 cylinder 167
Figure 3.37. Comparison of anomaly symptom data detection of GCCV and RGCCV based on all data of No. 2 GE No. 1 cylinder 168
Figure 3.38. Comparison of anomaly symptom data detection of GCCV and RGCCV of average data based on VOYAGE of No. 2... 169
Figure 3.39. RGCCV of average data including anomaly symptom based on VOYAGE of No. 2 GE all cylinder 173
Figure 3.40. Prediction result of maintenance time based on RGCCV of No. 2 GE No. 1 cylinder 177
Figure 3.41. Prediction result of maintenance time based on exhaust gas temperature of No. 2 GE No. 1 cylinder 179
Figure 3.42. Comparison of result on generator maintenance time prediction algorithm of No. 2 GE No. 1 cylinder 181