Generative Adversarial Network-Based Fault Detection in Semiconductor Equipment with Class-Imbalanced Data

Choi, Jeong Eun; Seol, Da Hoon; Kim, Chan Young; Hong, Sang Jeen

doi:10.3390/s23041889

Cited by 7 publications

(4 citation statements)

References 40 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Another study presented generative adversarial networks (GANs) of deep learning to solve fault detection with imbalanced data in the plasma etching process. It verified the benefits of GANs solving imbalance OES data problems [21]. This study aims to identify the optimal pulse parameters for producing high-quality AlN films.…”

Section: Introductionmentioning

confidence: 70%

Impact of Pulse Parameters of a DC Power Generator on the Microstructural and Mechanical Properties of Sputtered AlN Film with In-Situ OES Data Analysis

et al. 2023

View full text Add to dashboard Cite

In the present study, the sputtered aluminum nitride (AlN) films were processed in a reactive pulsed DC magnetron system. We applied a total of 15 different design of experiments (DOEs) on DC pulsed parameters (reverse voltage, pulse frequency, and duty cycle) with Box–Behnken experimental method and response surface method (RSM) to establish a mathematical model by experimental data for interpreting the relationship between independent and response variables. For the characterization of AlN films on the crystal quality, microstructure, thickness, and surface roughness, X-ray diffraction (XRD), atomic force microscopy (AFM), and field emission-scanning electron microscopy (FE-SEM) were utilized. AlN films have different microstructures and surface roughness under different pulse parameters. In addition, in-situ optical emission spectroscopy (OES) was employed to monitor the plasma in real-time, and its data were analyzed by principal component analysis (PCA) for dimensionality reduction and data preprocessing. Through the CatBoost modeling and analysis, we predicted results from XRD in full width at half maximum (FWHM) and SEM in grain size. This investigation identified the optimal pulse parameters for producing high-quality AlN films as a reverse voltage of 50 V, a pulse frequency of 250 kHz, and a duty cycle of 80.6061%. Additionally, a predictive CatBoost model for obtaining film FWHM and grain size was successfully trained.

show abstract

Section: Introductionmentioning

confidence: 70%

Impact of Pulse Parameters of a DC Power Generator on the Microstructural and Mechanical Properties of Sputtered AlN Film with In-Situ OES Data Analysis

et al. 2023

View full text Add to dashboard Cite

show abstract

“…The target process was silicon etching using SF 6 , O 2 , and Ar gas plasma. Previous studies have shown that small changes in plasma conditions can significantly affect the Si trench etching results for high aspect ratio [19,21,24,25]. Those studies used 20 × 20 mm 2 silicon pattern samples with an SiO 2 etch mask grown on a 300 mm silicon wafer, whereas in this study, 300 mm silicon wafers were used without a mask.…”

Section: Experiments and Data Acquisitionmentioning

confidence: 99%

Explainable artificial intelligence-based evidential inferencing on process faults in plasma etching

Choi,

An,

Lee

et al. 2024

J. Phys. D: Appl. Phys.

View full text Add to dashboard Cite

The fault detection and classification (FDC) modeling proposed in this study is a research approach that is intended to improve the performance of plasma process models by leveraging optical emission spectroscopy (OES) data containing plasma information (PI) and enhancing model interpretability using explainable artificial intelligence (XAI) algorithms. Status variable identification data that included normal and abnormal states of bias power, pressure, SF6 gas flow, and O2 gas flow were collected during a silicon etching process with SF6, O2 gas plasma. Additional variables were derived from the OES data and included additional PI, such as O and F radicals, which were computed using actinometry, and electron temperature and electron density computed using the line ratio method. By building a high-performance FDC model and interpreting its results using XAI algorithms, we propose solutions to the limitations of the FDC model in semiconductor plasma processes.

show abstract

“…Synthetic Datasets Goodfellow et al ( 2014) proposed GAN as a new generative modeling framework [14] to synthesize new data with the same characteristics from training examples, visually approximating the training data set. Various GANbased methods have been proposed for image synthesis in recent years [15], [16], [17], [18], [19], [20], [21], [22], [23], and [24] with applications spreading rapidly from computer vision and machine learning communities to domain-specific areas such as medical [25] [26], [27], [28], [29], and remote sensing [30], [31], [32] [33], [34], [35], [36], [37], [38], [39], [40], and [41]; industrial process [42], [43], [44], [45], [46], [47], and [48]; and agriculture [49], [50], [51], [52].…”

Section: B Gan (Generative Adversarial Network) To Producementioning

confidence: 99%

Oil Palm USB (Unstripped Bunch) Detector Trained on Synthetic Images Generated by PGGAN

Aji,

Bin Ghazali,

Akbar

2023

Journal of Robotics and Control

View full text Add to dashboard Cite

Identifying Unstriped Bunches (USB) is a pivotal challenge in palm oil production, contributing to reduced mill efficiency. Existing manual detection methods are proven time-consuming and prone to inaccuracies. Therefore, we propose an innovative solution harnessing computer vision technology. Specifically, we leverage the Faster R-CNN (Region-based Convolution Neural Network), a robust object detection algorithm, and complement it with Progressive Growing Generative Adversarial Networks (PGGAN) for synthetic image generation. Nevertheless, a scarcity of authentic USB images may hinder the application of Faster R-CNN. Herein, PGGAN is assumed to be pivotal in generating synthetic images of Empty Fruit Bunches (EFB) and USB. Our approach pairs synthetic images with authentic ones to train the Faster R-CNN. The VGG16 feature generator serves as the architectural backbone, fostering enhanced learning. According to our experimental results, USB detectors that were trained solely with authentic images resulted in an accuracy of 77.1%, which highlights the potential of this methodology. However, employing solely synthetic images leads to a slightly reduced accuracy of 75.3%. Strikingly, the fusion of authentic and synthetic images in a balanced ratio of 1:1 fuels a remarkable accuracy surge to 87.9%, signifying a 10.1% improvement. This innovative amalgamation underscores the potential of synthetic data augmentation in refining detection systems. By amalgamating authentic and synthetic data, we unlock a novel dimension of accuracy in USB detection, which was previously unattainable. This contribution holds significant implications for the industry, ensuring further exploration into advanced data synthesis techniques and refining detection models.

show abstract

Generative Adversarial Network-Based Fault Detection in Semiconductor Equipment with Class-Imbalanced Data

Cited by 7 publications

References 40 publications

Impact of Pulse Parameters of a DC Power Generator on the Microstructural and Mechanical Properties of Sputtered AlN Film with In-Situ OES Data Analysis

Impact of Pulse Parameters of a DC Power Generator on the Microstructural and Mechanical Properties of Sputtered AlN Film with In-Situ OES Data Analysis

Explainable artificial intelligence-based evidential inferencing on process faults in plasma etching

Oil Palm USB (Unstripped Bunch) Detector Trained on Synthetic Images Generated by PGGAN

Contact Info

Product

Resources

About