• Tidak ada hasil yang ditemukan

application of an artificial neural network model to

N/A
N/A
Protected

Academic year: 2023

Membagikan "application of an artificial neural network model to"

Copied!
12
0
0

Teks penuh

(1)

PREDICT PARAMETER OF FRICTION STIR SPOT WELDING ON ALUMINUM SHEET

By

Albertus Aan Dian Nugroho 21952061

MASTER’S DEGREE in

MECHANICAL ENGINEERING (MECHATRONICS)

FACULTY OF ENGINEERING AND INFORMATION TECHNOLOGY

SWISS GERMAN UNIVERSITY The Prominence Tower

Jalan Jalur Sutera Barat No. 15, Alam Sutera Tangerang, Banten 15143 - Indonesia

JANUARY 2021

Revision After Thesis Defense on January 28th, 2021

(2)

Albertus Aan D Nugroho STATEMENT BY THE AUTHOR

I hereby declare that this submission is my own work and to the best of my knowledge, it contains no material previously published or written by another person, nor material which to a substantial extent has been accepted for the award of any other degree or diploma at any educational institution, except where due acknowledgement is made in the thesis.

Albertus Aan Dian Nugroho

_____________________________________________

Student Date

Approved by:

Edy Sofyan, B.Eng., M.Eng., Ph.D.

_____________________________________________

Thesis Advisor Date

Dena Hendriana, B.Sc., S.M., Sc.D.

_____________________________________________

Thesis Co-Advisor Date

Dr. Maulahikmah Galinium, S. Kom, M.Sc.

_____________________________________________

Dean Date

(3)

Albertus Aan D Nugroho ABSTRACT

APPLICATION OF AN ARTIFICIAL NEURAL NETWORK MODEL TO PREDICT PARAMETER OF FRICTION STIR SPOT WELDING ON ALUMINUM

SHEET By

Albertus Aan Dian Nugroho

Edy Sofyan, B.Eng., M.Eng., Ph.D., Advisor Dena Hendriana, B.Sc., S.M., Sc.D., Co-Advisor

SWISS GERMAN UNIVERSITY

This research was conducted to predict the maximum load in the Friction Stir Spot Welding process using Aluminum Alloy 1050 material. Process parameters there are 4 variations, namely tool pin diameter, tool rotation speed, welding speed which each has 3 levels, and plunge depth which has 2 levels. The experimental design in this research used the Taguchi method with 54 experiments. The results of the training Backpropagation Neural Network have a 4-8-8-1 network architecture that consists of from 4 input layers, 2 hidden layers with 8 neurons, and 1 neuron on output layer. The activation function used is "tansig" and the training function is "trainrp". In addition, regression analysis was also carried out on the 4 parameters of the Friction Stir Spot Welding which are the input variables. From the results of the regression analysis, it is known that the parameters of welding speed (46.68%) and tool diameter (36.85%) have the most influence on the magnitude of maximum load.

Keywords: Friction Stir Spot Welding, Maximum Load, Regression, Taguchi, tansig, trainrp, Backpropagation

(4)

Albertus Aan D Nugroho

© Copyright 2021 by Albertus Aan Dian Nugroho

All rights reserved.

(5)

Albertus Aan D Nugroho DEDICATION

This study is wholeheartedly dedicated to my beloved wife and two children, who have been our source of inspiration and gave me strength when I thought of giving up.

To Polman Astra who has provided funding and support during the study period.

And lastly, I dedicated to the God, thank you for the guidance, strength, power of mind, protection and for giving us a healthy life.

(6)

Albertus Aan D Nugroho ACKNOWLEDGEMENTS

Praise be to God the Almighty for the blessing of His grace I can complete this thesis.

The writing of this thesis is done to fulfill one of the requirements to achieve a master’s degree majoring in Master of Mechanical Engineering, Swiss German University. I realize that in the writing process until the completion of this thesis many people have helped and encouraged me in writing this thesis. Therefore, I would like to thank:

1. My beloved wife and children who always provide support every day.

2. Mr. Edy Sofyan, B.Eng., M.Eng., Ph.D. as the Advisor who has provided the time, energy, and thoughts to direct the author in the preparation of this thesis.

3. Mr. Dena Hendriana, B.Sc., S.M., Sc.D. as the Co-Advisor who has provided the time, energy, and thoughts to direct the author in the preparation of this thesis.

4. Astra Manufacturing Polytechnic extended family for the support that has provided learning opportunities and higher motivation.

5. All my friends in MME Batch 10 in arms who have contributed to writing this Thesis.

.

Tangerang, January 2021

Albertus Aan Dian Nugroho

(7)

Albertus Aan D Nugroho TABLE OF CONTENTS

Page

TITLE ... 1

STATEMENT BY THE AUTHOR ... 2

ABSTRACT ... 3

DEDICATION ... 5

ACKNOWLEDGEMENTS ... 6

TABLE OF CONTENTS ... 7

LIST OF FIGURES ... 10

LIST OF TABLES... 12

CHAPTER 1 - INTRODUCTION ... 13

1.1. Background ... 13

1.2. Research Problem ... 14

1.3. Research Objectives ... 15

1.4. Significance of Study ... 15

1.5. Research Questions ... 15

1.6. Hypothesis ... 15

CHAPTER 2 - LITERATURE REVIEW ... 16

2.1. Friction Stir Spot Welding ... 16

2.2. Mechanical Properties ... 22

2.3. Aluminum Alloy ... 25

2.4. Taguchi Method Experimental Design ... 27

2.5. Artificial Neural Network (ANN) ... 29

2.6. Neural Models ... 31

2.7. Backpropagation Neural Network (BPNN) ... 33

2.8. Previous Study ... 37

CHAPTER 3 – RESEARCH METHODS ... 39

3.1. Research Design ... 39

3.2. Scope of Study ... 40

(8)

Albertus Aan D Nugroho

3.5. Experimental Design ... 46

3.6. Experimental Preparation ... 49

3.7. Experimental Response Data ... 50

3.8. Regression Analysis ... 50

CHAPTER 4 – RESULTS AND DISCUSSIONS ... 56

4.1. Result Friction Stir Spot Welding Process ... 56

4.2. Data Analysis ... 56

4.3. Normalization Process Input and Output Data ... 58

4.4. Stopping Criteria ... 60

4.5. Network Form BPNN ... 60

4.6. Training Data and Testing Data ... 62

4.7. Weight Value and Bias Value Criteria ... 62

4.8. Weight Value and Bias Value Result ... 63

4.9. BPNN Result ... 65

4.10. Regression Analysis ... 67

4.11. Regression Coefficient ... 71

4.12. Comparison Backpropagation and Regression ... 73

4.13. Experimental Confirmation ... 74

CHAPTER 5 – CONCLUSIONS AND RECOMMENDATIONS ... 75

5.1. Conclusions ... 75

5.2. Recommendations ... 75

GLOSSARY ... 76

REFERENCES ... 77

APPENDIX ... 80

1. Datasheet Aluminum Alloy 1050 ... 80

2. Composition of HSS ... 81

3. Ortogonal Array ... 82

4. Three Level Orthogonal Array ... 83

5. Degree of Freedom ... 84

7. Summary Multiple Regression ... 86

(9)

Albertus Aan D Nugroho 9. Regression ... 92 CURRICULUM VITAE ... 93

(10)

Albertus Aan D Nugroho LIST OF FIGURES

Figures Page

Figure 1 Schematic of Friction Stir Welding (ASM International, 2007) ... 16

Figure 2 Schematic of Friction Stir Welding (Khaled, Terry, 2005) ... 17

Figure 3 Heat zone on Friction Stir Welding (ASM International, 2007) ... 17

Figure 4 Scheme Stir Weld (Khaled, Terry, 2005) ... 18

Figure 5 Design tools configuration (Thomas, WM., 1991) ... 20

Figure 6 Example of design tools (Rowe, C E D; Thomas, Wayne, 2006) ... 20

Figure 7 Tools Load of Friction Stir Welding (ASM International,2007) ... 21

Figure 8 Micro Vickers Hardness Tester ... 23

Figure 9 Profile hardness welding area (ASM International, 2007) ... 24

Figure 10 Scheme and result tensile test ... 24

Figure 11 Cast and wrought aluminum (I. J. Polmear, 1995) ... 25

Figure 12 Aluminum forming method (Arifin, Bustanul, 2007)... 26

Figure 13 Taguchi Loss Target Model (Pyzdek, 2003) ... 27

Figure 14 Single layer network (Kusumadewi, 2014) ... 30

Figure 15 Multi-layer network (Sutojo, 2018) ... 31

Figure 16 Neural Network method (Kusumadewi, 2014) ... 32

Figure 17 Research Flowchart... 39

Figure 18 Tool Geometry and tool variation... 43

Figure 19 Makino KE55 CNC Milling Machine (makino catalog,2001) ... 44

Figure 20 Tensile Testing Machine Galdabini ... 45

Figure 21 Specimen test dimension ... 49

Figure 22 Relationship between load and tool pin diameter ... 51

Figure 23 Relationship between load and rotation speed ... 51

Figure 24 Relationship between load and welding speed ... 52

Figure 25 Relationship between load and plunge depth ... 52

Figure 26 Chart of MR tool pin diameter ... 53

Figure 27 Chart of MR tool rotation speed ... 54

(11)

Albertus Aan D Nugroho

Figure 29 Chart of MR tool plunge depth ... 55

Figure 30 Specimen Dimension ... 56

Figure 31 Sample of FSSW result. ... 56

Figure 32 BPNN Network Architecture ... 62

Figure 33 Weight and Bias values from Input Layer to Hidden Layer 1 ... 63

Figure 34 Weight and Bias values from Hidden Layer 1 to Hidden Layer 2 ... 64

Figure 35 Weight and Bias values from Hidden Layer 2 to Output Layer ... 64

Figure 36 Graph of BPNN's result data. ... 65

Figure 37 Graph SLR of Tool Pin Diameter ... 67

Figure 38 Graph SLR of Tool Rotation Speed ... 68

Figure 39 Graph SLR of Welding Speed ... 68

Figure 40 Graph SLR of Tool Plunge Depth ... 69

(12)

Albertus Aan D Nugroho LIST OF TABLES

Table Page

Table 1 FSW Material Tool and its applications (ASM International, 2007)... 19

Table 2 Orthogonal Matrix Selection (Roy, 2020) ... 28

Table 3 Process parameter and variation ... 41

Table 4 Chemical Composition of AA 1050 ... 42

Table 5 Mechanical Properties AA 1050 H14 ... 43

Table 6 Mechanical Properties of HSS ... 44

Table 7 Specification Makino KE55 (makino catalog, 2001) ... 45

Table 8 Degree of Freedom of Process Parameters and Levels ... 46

Table 9 Orthogonal Arrays three level, L27 ( Ranjit K. Roy 2010) ... 47

Table 10 Experiment Design based on taguchi method. ... 48

Table 11 Experiment response data ... 50

Table 12 Aluminum Alloy Specimen Test Result Data ... 57

Table 13 BPNN Preprocessing Result Data ... 58

Table 14 Stopping Criteria on BPNN ... 60

Table 15 Training BPNN Result Data ... 61

Table 16 Detail of Network Architecture Model ... 61

Table 17 Random training data and testing data. ... 62

Table 18 Weight and Bias Value Criteria ... 63

Table 19 Comparison between Backpropagation and Experiment ... 65

Table 20 Multiple Regression Summary Output ... 70

Table 21 Regression Coefficient ... 71

Table 22 Significant effect with t test ... 72

Table 23 Significant effect with P-value ... 72

Table 24 Comparison between Predicted Regression and Experiment ... 73

Table 25 Comparison between Backpropagation and Regression ... 74

Referensi

Dokumen terkait

The objective of this work is to model the muscle EMG signal to torque for a motor control of the arm rehabilitation device using Artificial Neural Network

The results of permeability calculations at uncored intervals using artificial neural networks produce values that are not much different from core data at the same depth after

Judul: ESTIMATING MISSING PRECIPITATION TO OPTIMIZE PARAMETERS FOR PREDICTION OF DAILY WATER LEVEL USING ARTIFICIAL.. NEURAL

The model parameters, such as the geometry model, flow model, hydrodynamic model, seabed morphology model, and sediment transport model, determine the accuracy of

Twenty seven 27 architectural combinations of three activation function logsig, tansig, purelin have been trained using 126 data set 87.5% of biodiesel yield observed at three different

APPLICABILITY OF ARTIFICIAL NEURAL NETWORK MODEL FOR SIMULATION OF MONTHLY RUNOFF IN COMPARISON WITH SOME OTHER TRADITIONAL MODELS Le Van Duc University of Technology, VNU-HCM

2, Oktober 2023, pp 152-163 152 doi : 10.31849/siklus.v9i2.14221 Pengaruh Variasi Hidden Layer Terhadap Nilai MAPE Pada Pengembangan Model Estimasi Biaya Menggunakan Artificial

The results obtained are that the best method for predicting direct economic losses due to earthquakes in Indonesia is to use the Backpropagation Neural Network BPNN method, because it