Improved Monitoring of Wind Speed Using 3D Printing and Data‐Driven Deep Learning Model for Wind Power Systems DOI Creative Commons
Sanghun Shin, Sangyeun Park, Hongyun So

et al.

International Journal of Energy Research, Journal Year: 2024, Volume and Issue: 2024(1)

Published: Jan. 1, 2024

This study presents a novel method for airflow rate (i.e., wind speed) sensing using three‐dimensional (3D) printing‐assisted flow sensor and deep neural network (DNN). The 3D printing of thermoplastic polyurethane can realize multisensing devices different values. Herein, the 3D‐printed with an actuating membrane is used to simultaneously measure two electrical parameters capacitance resistance) depending on rate. Subsequently, data‐driven DNN model introduced trained 6,965 experimental data points, including input (resistance capacitance) output (airflow rate) without external interferences during measurements. mean absolute error (MAE), squared (MSE), root logarithmic (RMSLE) measured predicted values by multiple inputs are 0.59, 0.7, 0.18 continuous test dataset interference 1.16, 3.95, 0.73 interference, respectively. Compared prediction results single‐input cases, average MAE, MSE, RMSLE significantly decrease 70.37%, 88.74%, 72.26% datasets 51.91%, 53.01%, 12.20% suggest cost‐effective accurate technology speed monitoring in power systems.

Language: Английский

Temperature-induced degradation of GaN HEMT: An in situ heating study DOI
Md Abu Jafar Rasel, Di Zhang, Aiping Chen

et al.

Journal of Vacuum Science & Technology B Nanotechnology and Microelectronics Materials Processing Measurement and Phenomena, Journal Year: 2024, Volume and Issue: 42(3)

Published: April 26, 2024

High-power electronics, such as GaN high electron mobility transistors (HEMTs), are expected to perform reliably in high-temperature conditions. This study aims gain an understanding of the microscopic origin both material and device vulnerabilities temperatures by real-time monitoring onset structural degradation under varying temperature is achieved operating HEMT devices situ inside a transmission microscope (TEM). Electron-transparent specimens prepared from bulk heated up 800 °C. High-resolution TEM (HRTEM), scanning (STEM), energy-dispersive x-ray spectroscopy (EDS), geometric phase analysis (GPA) performed evaluate crystal quality, diffusion, strain propagation sample before after heating. Gate contact area reduction visible 470 °C accompanied Ni/Au intermixing near gate/AlGaN interface. Elevated induce significant out-of-plane lattice expansion at SiNx/GaN/AlGaN interface, revealed geometry-phase GPA maps, while in-plane strains remain relatively consistent. Exposure exceeding 500 leads almost two orders magnitude increase leakage current this study, which complements results our experiment. The findings offer visual insights into identifying initial location highlight impact on device’s structure, electrical properties, degradation.

Language: Английский

Citations

7
Recent developments of in-situ process and in-line quality monitoring in injection molding using intelligent sensors DOI
Sanghun Shin, Keuntae Baek, Jae‐Min Oh

et al.

Sensors and Actuators A Physical, Journal Year: 2025, Volume and Issue: unknown, P. 116248 - 116248

Published: Jan. 1, 2025

Language: Английский

Citations

0
Deep learning-developed multi-light source discrimination capability of stretchable capacitive photodetector DOI Creative Commons

Su Bin Choi,

Jun Sang Choi,

Hyun Sik Shin

et al.

npj Flexible Electronics, Journal Year: 2025, Volume and Issue: 9(1)

Published: May 15, 2025

Language: Английский

Citations

0
Investigation of Laser Ablation Quality Based on Data Science and Machine Learning XGBoost Classifier DOI Creative Commons

Chien-Chung Tsai,

Tung-Hon Yiu

Applied Sciences, Journal Year: 2023, Volume and Issue: 14(1), P. 326 - 326

Published: Dec. 29, 2023

This work proposes a matching data science approach for the laser ablation quality, reb, study of Si3N4 film based on supervised machine learning classifiers in CMOS-MEMS process. The demonstrates that there exists an energy threshold, Eth, ablation. If surpasses this increasing interval time will not contribute significantly to recovery pulse energy. Thus, reb enhancement is limited. When greater than 0.258 mJ, critical value at which relatively low each level, respectively. In addition, variation Δreb, independent invariant point between 0.32 mJ and 0.36 mJ. Energy exhibit Pearson correlation 0.82 0.53 with To maintain Δreb below 0.15, green operating energies 0.258–0.378 can adopt baseline initial multiplied by 1/∜2. Additionally, 0.288–0.378 during ablation, be kept 0.1. With forced partition methods, namely, k-means method percentile method, XGBoost (v 2.0.3) classifier maintains competitive accuracy across test sizes 0.20–0.40, outperforming algorithms Random Forest Logistic Regression, highest 0.78 size 0.20.

Language: Английский

Citations

6
A Job Recommendation Model Based on a Two-Layer Attention Mechanism DOI Open Access
Yu M,

Shaojie Lin,

Yuxuan Cheng

et al.

Electronics, Journal Year: 2024, Volume and Issue: 13(3), P. 485 - 485

Published: Jan. 24, 2024

In the field of job recruitment, traditional recommendation methods only rely on users’ rating data positions for information matching. This simple strategy has problems such as low utilization multi-source heterogeneous and difficulty in mining relevant between recruiters applicants. Therefore, this paper proposes a recurrent neural network model based two-layer attention mechanism. The first improves entity representation applicants through user behavior, company-related knowledge other information. entities their combinations are then mapped to vector space using one-hot TransR methods, with mechanism is used obtain potential interests from click sequence, list generated. experimental results show that achieves better than previous models.

Language: Английский

Citations

1
Improved Monitoring of Wind Speed Using 3D Printing and Data‐Driven Deep Learning Model for Wind Power Systems DOI Creative Commons
Sanghun Shin, Sangyeun Park, Hongyun So

et al.

International Journal of Energy Research, Journal Year: 2024, Volume and Issue: 2024(1)

Published: Jan. 1, 2024

This study presents a novel method for airflow rate (i.e., wind speed) sensing using three‐dimensional (3D) printing‐assisted flow sensor and deep neural network (DNN). The 3D printing of thermoplastic polyurethane can realize multisensing devices different values. Herein, the 3D‐printed with an actuating membrane is used to simultaneously measure two electrical parameters capacitance resistance) depending on rate. Subsequently, data‐driven DNN model introduced trained 6,965 experimental data points, including input (resistance capacitance) output (airflow rate) without external interferences during measurements. mean absolute error (MAE), squared (MSE), root logarithmic (RMSLE) measured predicted values by multiple inputs are 0.59, 0.7, 0.18 continuous test dataset interference 1.16, 3.95, 0.73 interference, respectively. Compared prediction results single‐input cases, average MAE, MSE, RMSLE significantly decrease 70.37%, 88.74%, 72.26% datasets 51.91%, 53.01%, 12.20% suggest cost‐effective accurate technology speed monitoring in power systems.

Language: Английский

Citations

1