Title Page
Contents
Abstract 10
초록 12
CHAPTER ONE. [제목없음] 14
1.1. Background and Motivation 14
1.2. Thesis Contribution 17
1.3. Thesis Outline 17
CHAPTER TWO. [제목없음] 18
2.1. Machine Learning 18
2.1.1. Supervised Learning 19
2.1.2. Unsupervised Learning 20
2.1.3. Semi-supervised Learning 20
2.2. Deep Learning 21
2.3. Anomaly Detection 22
2.4. Working Principle of EPS 24
CHAPTER THREE. [제목없음] 26
3.1. Traditional Machine Learning Approach 26
3.2. Deep Learning Approach 33
CHAPTER FOUR. [제목없음] 39
4.1. Pre-processing of Data 40
4.2. Model Training 41
4.2.1. LSTM Network 41
4.2.2. Autoencoder 46
4.3. Anomaly Detection Approach 49
CHAPTER FIVE. [제목없음] 51
5.1. Dataset Collection 51
5.2. Experimental Setup and Model Hyperparameters 52
5.3. Evaluation Metrics 54
5.4. Performance Results 57
5.4.1. EPS Data Anomaly Detection 57
5.4.2. Models Result Analysis 59
CHAPTER SIX. [제목없음] 64
6.1. Conclusion and Future Work 64
Reference 66
Publications 73
Table 1. Model training Parameters 54
Table 2. provides an overview of the confusion matrix, when examining the output of an anomaly detection classification model. 55
Table 3. performance comparison considering the values of true positive (TP), false positive (FP), false negative (FN), and true negative (TN), as well as... 57
Table 4. Benchmarking performance against similar models 63
Figure 2.1. Different machine-learning approaches and data requirements. 19
Figure 2.2. The relationship between AI, ML, and DL. 21
Figure 2.3. A generalized framework for detecting anomalies. 23
Figure 2.4. Workflow of EPS system. 25
Figure 4.1. Layout of our proposed method for anomaly detection in EPS. 40
Figure 4.2. Schematic illustration of the LSTM architecture. 43
Figure 4.3. Schematic illustration of the autoencoder architecture. 46
Figure 4.4. Model training architectures 49
Figure 4.5. Patterns of detected anomalies in EPS torque sensor 50
Figure 5.1. Schematic illustration of dataset collection 52
Figure 5.2. Epoch graph of training and validation loss for EPS torque sensor 53
Figure 5.3. Anomaly detection on EPS torque sensor data 58
Figure 5.4. Confusion matrix of model detection on EPS data 60
Figure 5.5. Model receiver operation characteristic curve 62