Title Page
Contents
Abstract (국문) 9
1. Introduction 11
2. Experiment 14
2.1. SPR Experimental materials 14
2.2. SPR Experimental environment 18
2.3. SPR Experimental Results 21
2.4. FDS Experimental Materials 28
2.5. FDS Experimental environment 29
2.6. FDS Experimental Results 31
3. Development of prediction models 34
3.1. Data preprocessing 34
3.2. CNN Model architecture 36
3.3. Post-processing of CNN Predictions 41
3.4. Validation of the CNN models 45
4. Conclusion 47
5. References 50
Abstract (English) 55
Fig. 1. Configuration of self-piercing rivet machine. (a) SPR joining equipment (b) Sheets combination of heterogeneous materials 17
Fig. 2. Joining Quality Factors for SPR 19
Fig. 3. Experiment conditions of SPR. (a) Clamping condition (b) Gap condition (c) Thickness change condition (d) Strength change condition (e) Rivet change condition 21
Fig. 4. Interlock probability distribution of SPR for each condition. (a) Clamping condition (b) Gap condition (c) Thickness change condition (d) Strength change condition (e) Rivet change condition 26
Fig. 5. Force-displacement curves (Monitoring graphs) of SPR for each condition. (a) Clamping condition (b) Gap condition (c) Thickness change condition (d) Strength change... 27
Fig. 6. Configuration of flow drill screw machine. (a) FDS joining equipment (b) Sheets combination of FDS with prehole 29
Fig. 7. Joining Quality Factors for FDS 31
Fig. 8. Experiment conditions of FDS. (a) Normal condition (b) Internal gap condition (c) Misaligned condition (d) Scraping condition 32
Fig. 9. Torque-time curves (Monitoring graphs) of FDS for each condition. (a) Internal gap condition (b) Misaligned condition (c) Scraping condition 33
Fig. 10. Data preprocessing using Moving Average Method. (a) Before data preprocessing (b) After data preprocessing 35
Fig. 11. Convolutional Neural Network (CNN) Model architecture. (a) SPR CNN Model (b) FDS CNN Model 38
Fig. 12. Convolutional Neural Network (CNN) Model architecture. (a) SPR CNN Model (b) FDS CNN Model 39
Fig. 13. Confusion matrix for 2 class (a) Confusion matrix (b) Accuracy equation of confusion matrix 41
Fig. 14. Confusion matrix for test data set (a) SPR Confusion matrix and accuracy (b) FDS Confusion matrix and accuracy 42
Fig. 15. SPR Scatter plot graphs of shape factor -Interlock (a) Graph for boundary definition between OK and Close to NG (b) Graph for boundary definition between Close to NG and NG 44
Fig. 16. FDS Scatter plot graphs of shape factor-Gap & Head height (a) Graph for boundary definition between OK and Close to NG (b) Graph for boundary definition between Close to NG... 44
Fig. 17. K-fold Cross Validation structure of K=4 45
Fig. 18. Structure of a Decision Tree. 46
Fig. 19. CNN loss of validation data set (a) SPR loss of validation data set 1 (b) FDS loss of validation data set 1 46
Fig. 20. Confusion matrix of Decision Tree & refined model of CNN (a) SPR Confusion matrix and accuracy of Decision Tree (b) SPR Confusion matrix and accuracy of CNN (c) FDS... 49