Ming C. Leu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Ming C. Leu has a total of 38 co-authors. Here are the top co-authors by number of shared publications:
- Haodong Chen
Seattle Children’s Hospital
4 shared publications - Niloofar Zendehdel
Missouri University of Science and Technology
3 shared publications - Zhaozheng Yin
Stony Brook University
3 shared publications - Adam Bratten
Department of Materials Science and Engineering Missouri University of Science and Technology Rolla Missouri USA
2 shared publications - Gregory E. Hilmas
Materials Science and Engineering Missouri University of Science and Technology Rolla Missouri USA
2 shared publications - Haiming Wen
Department of Materials Science and Engineering, Missouri University of Science and Technology, Rolla, USA
2 shared publications
Latest Publications
- 14 September 2026
- Published by Emerald Publishing in Rapid Prototyping Journal
- p. 1-23
In-situ thermography and material characterization of scanning strategy effects in laser foil printing of Zr-based metallic glass
Purpose: This paper aims to investigate the thermal behavior and amorphous stability of zirconium-based metallic glass (Zr65.7Cu15.6Ni11.7Al3.7Ti3.3) fabricated using the laser foil printing (LFP) technique. This study focuses on establishing the relationship between laser scanning strategies and melt pool cooling rates, emphasizing their influence on amorphous retention and crystallization behavior during multilayer deposition on Zr702 substrates. Design/methodology/approach: In-situ thermography using a mid-wavelength infrared camera was applied to monitor the real-time temperature evolution of the melt pool and heat-affected zone. Finite element simulations were performed to analyze melt pool temperature gradients and transient heat transfer. Continuous and discrete welding strategies were compared to evaluate their effects on cooling rate and amorphous stability. Postprocess characterization, including X-ray diffraction, energy-dispersive spectroscopy and microhardness testing, was conducted to assess structural, compositional and mechanical changes induced by different thermal conditions. Findings: Continuous welding produced larger and deeper melt pools due to sustained energy input, resulting in significant heat accumulation, and lower cooling rates that led to partial crystallization. In contrast, discrete welding achieved localized and intermittent energy input, generating higher and more stable cooling rates that effectively suppressed heat buildup. Consequently, fully amorphous structures were retained across multiple layers. Compositional analysis revealed a gradual decrease in zirconium content and enrichment of nickel and copper with increasing build height, while microhardness testing confirmed higher hardness values in samples fabricated using discrete welding. Originality/value: This study provides new insights into the correlation between laser scanning strategy and thermal dynamics in the LFP processing of metallic glass. The findings demonstrate that discrete welding offers superior thermal control, enhanced amorphous retention and improved mechanical performance. The established process–structure–property relationship offers practical guidelines for advancing additive manufacturing of bulk metallic glasses with high thermal stability and reliability.
- 14 September 2026
- Published by Emerald Publishing in Rapid Prototyping Journal
- p. 1-23
In-situ thermography and material characterization of scanning strategy effects in laser foil printing of Zr-based metallic glass
Purpose: This paper aims to investigate the thermal behavior and amorphous stability of zirconium-based metallic glass (Zr65.7Cu15.6Ni11.7Al3.7Ti3.3) fabricated using the laser foil printing (LFP) technique. This study focuses on establishing the relationship between laser scanning strategies and melt pool cooling rates, emphasizing their influence on amorphous retention and crystallization behavior during multilayer deposition on Zr702 substrates. Design/methodology/approach: In-situ thermography using a mid-wavelength infrared camera was applied to monitor the real-time temperature evolution of the melt pool and heat-affected zone. Finite element simulations were performed to analyze melt pool temperature gradients and transient heat transfer. Continuous and discrete welding strategies were compared to evaluate their effects on cooling rate and amorphous stability. Postprocess characterization, including X-ray diffraction, energy-dispersive spectroscopy and microhardness testing, was conducted to assess structural, compositional and mechanical changes induced by different thermal conditions. Findings: Continuous welding produced larger and deeper melt pools due to sustained energy input, resulting in significant heat accumulation, and lower cooling rates that led to partial crystallization. In contrast, discrete welding achieved localized and intermittent energy input, generating higher and more stable cooling rates that effectively suppressed heat buildup. Consequently, fully amorphous structures were retained across multiple layers. Compositional analysis revealed a gradual decrease in zirconium content and enrichment of nickel and copper with increasing build height, while microhardness testing confirmed higher hardness values in samples fabricated using discrete welding. Originality/value: This study provides new insights into the correlation between laser scanning strategy and thermal dynamics in the LFP processing of metallic glass. The findings demonstrate that discrete welding offers superior thermal control, enhanced amorphous retention and improved mechanical performance. The established process–structure–property relationship offers practical guidelines for advancing additive manufacturing of bulk metallic glasses with high thermal stability and reliability.
- Open Access
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 163902-163920
EEG-Based Classification of Human Motor Intention Using a Hybrid EEGNet–Transformer Model
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 163902-163920
Open AccessEEG-Based Classification of Human Motor Intention Using a Hybrid EEGNet–Transformer Model
- Open Access
- 10 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 101852-101868
Hands-Free UAV Control: Real-Time Eye Movement Detection Using EOG and LSTM Networks
Industry 4.0 has created a growing need for effective human-robot collaboration (HRC). As robots and humans work more closely together, efficient communication becomes essential for coordinating their actions seamlessly. While speech may seem like the obvious choice for communication, noisy factory environments can render it impractical. Additionally, workers often have their hands occupied with assembly tasks, making hand-controlled interfaces less practical for controlling robots. To address these challenges, this paper presents a novel, hands-free method for robot control using electrooculography (EOG) signals–specifically, eye movements and blinks–with unmanned aerial vehicles (UAVs) used as the demonstration platform. We developed a real-time system that captures EOG signals through a graphical user interface (GUI), allowing users to direct their gaze toward on-screen commands and confirm selections with blinks. An LSTM (Long Short-Term Memory) model was trained to detect gaze coordinates and blink events from the EOG data. For performance benchmarking, we implemented a vision-based video-oculography (VOG) system as a camera-based baseline and compared both approaches under various conditions, including low light, wind disturbances, and deployment on embedded hardware as a resource-constrained device. Our comparative analysis showed the EOG-based method achieved 97.6% accuracy in command detection under normal conditions and maintained strong performance across challenging environments where the vision-based VOG system experienced significant degradation. The proposed EOG method also operated in real time at 47.97 FPS on a high-performance workstation and 38.87 FPS on the resource-constrained NVIDIA Jetson Orin Nano. On this platform, it ran 13.9× faster and used 319× less physical memory than the VOG-based alternative. These results underscore the advantages of EOG-based control in delivering reliable, real-time performance on both high performance workstations and resource-constrained platforms, outperforming vision-based alternatives in efficiency and adaptability.
- 10 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 101852-101868
Open AccessHands-Free UAV Control: Real-Time Eye Movement Detection Using EOG and LSTM Networks
Industry 4.0 has created a growing need for effective human-robot collaboration (HRC). As robots and humans work more closely together, efficient communication becomes essential for coordinating their actions seamlessly. While speech may seem like the obvious choice for communication, noisy factory environments can render it impractical. Additionally, workers often have their hands occupied with assembly tasks, making hand-controlled interfaces less practical for controlling robots. To address these challenges, this paper presents a novel, hands-free method for robot control using electrooculography (EOG) signals–specifically, eye movements and blinks–with unmanned aerial vehicles (UAVs) used as the demonstration platform. We developed a real-time system that captures EOG signals through a graphical user interface (GUI), allowing users to direct their gaze toward on-screen commands and confirm selections with blinks. An LSTM (Long Short-Term Memory) model was trained to detect gaze coordinates and blink events from the EOG data. For performance benchmarking, we implemented a vision-based video-oculography (VOG) system as a camera-based baseline and compared both approaches under various conditions, including low light, wind disturbances, and deployment on embedded hardware as a resource-constrained device. Our comparative analysis showed the EOG-based method achieved 97.6% accuracy in command detection under normal conditions and maintained strong performance across challenging environments where the vision-based VOG system experienced significant degradation. The proposed EOG method also operated in real time at 47.97 FPS on a high-performance workstation and 38.87 FPS on the resource-constrained NVIDIA Jetson Orin Nano. On this platform, it ran 13.9× faster and used 319× less physical memory than the VOG-based alternative. These results underscore the advantages of EOG-based control in delivering reliable, real-time performance on both high performance workstations and resource-constrained platforms, outperforming vision-based alternatives in efficiency and adaptability.
- Open Access
- 25 April 2025
- Published by Elsevier in CIRP Annals
- Vol. 74 (1) , 275-279
Study on field emission characteristics of carbon nanotube arrays patterned via laser welding of dissimilar materials
- 25 April 2025
- Published by Elsevier in CIRP Annals
- Vol. 74 (1) , 275-279
Open AccessStudy on field emission characteristics of carbon nanotube arrays patterned via laser welding of dissimilar materials
- 22 April 2025
- Published by Elsevier in Particuology
- Vol. 102, 78-85
Numerical study and experimental validation of copper powder plasma spheroidization process
- 22 April 2025
- Published by Elsevier in Particuology
- Vol. 102, 78-85
Numerical study and experimental validation of copper powder plasma spheroidization process
- Top Cited PapersOpen Access
- 26 January 2025
- Published by Elsevier in Engineering Applications of Artificial Intelligence
- Vol. 144, 110076
A gaze-driven manufacturing assembly assistant system with integrated step recognition, repetition analysis, and real-time feedback
- 26 January 2025
- Published by Elsevier in Engineering Applications of Artificial Intelligence
- Vol. 144, 110076
Top Cited PapersOpen AccessA gaze-driven manufacturing assembly assistant system with integrated step recognition, repetition analysis, and real-time feedback
- Open Access
- 4 August 2023
- Published by Elsevier in Additive Manufacturing
- Vol. 74, 103726
In-situ lock-in thermographic measurement of powder layer thermal diffusivity and thickness in laser powder bed fusion
- 4 August 2023
- Published by Elsevier in Additive Manufacturing
- Vol. 74, 103726
Open AccessIn-situ lock-in thermographic measurement of powder layer thermal diffusivity and thickness in laser powder bed fusion
- Open Access
- 22 May 2023
- Published by Springer Nature in Research Square
Fine-Grained Activity Classification in Assembly Based on Multi-Visual Modalities
Assembly activity recognition and prediction help to improve productivity, quality control, and safety measures in smart factories. This study aims to sense, recognize, and predict a worker's continuous fine-grained assembly activities in a manufacturing platform. We propose a two-stage network for workers' fine-grained activity classification by leveraging scene-level and temporal-level activity features. The first stage is a feature awareness block that extracts scene-level features from multi-visual modalities, including red-green-blue (RGB) and hand skeleton frames. We use the transfer learning method in the first stage and compare three different pre-trained feature extraction models. Then, we transmit the feature information from the first stage to the second stage to learn the temporal-level features of activities. The second stage consists of the Recurrent Neural Network (RNN) layers and a final classifier. We compare the performance of two different RNNs in the second stage, including the Long Short-Term Memory (LSTM) and the Gated Recurrent Unit (GRU). The partial video observation method is used in the prediction of fine-grained activities. In the experiments using the trimmed activity videos, our model achieves an accuracy of > 99% on our dataset and > 98% on the public dataset UCF 101, outperforming the state-of-the-art models. The prediction model achieves an accuracy of > 97% in predicting activity labels using 50% of the onset activity video information. In the experiments using an untrimmed video with continuous assembly activities, we combine our recognition and prediction models and achieve an accuracy of > 91% in real time, surpassing the state-of-the-art models for the recognition of continuous assembly activities.
- 22 May 2023
- Published by Springer Nature in Research Square
Open AccessFine-Grained Activity Classification in Assembly Based on Multi-Visual Modalities
Assembly activity recognition and prediction help to improve productivity, quality control, and safety measures in smart factories. This study aims to sense, recognize, and predict a worker's continuous fine-grained assembly activities in a manufacturing platform. We propose a two-stage network for workers' fine-grained activity classification by leveraging scene-level and temporal-level activity features. The first stage is a feature awareness block that extracts scene-level features from multi-visual modalities, including red-green-blue (RGB) and hand skeleton frames. We use the transfer learning method in the first stage and compare three different pre-trained feature extraction models. Then, we transmit the feature information from the first stage to the second stage to learn the temporal-level features of activities. The second stage consists of the Recurrent Neural Network (RNN) layers and a final classifier. We compare the performance of two different RNNs in the second stage, including the Long Short-Term Memory (LSTM) and the Gated Recurrent Unit (GRU). The partial video observation method is used in the prediction of fine-grained activities. In the experiments using the trimmed activity videos, our model achieves an accuracy of > 99% on our dataset and > 98% on the public dataset UCF 101, outperforming the state-of-the-art models. The prediction model achieves an accuracy of > 97% in predicting activity labels using 50% of the onset activity video information. In the experiments using an untrimmed video with continuous assembly activities, we combine our recognition and prediction models and achieve an accuracy of > 91% in real time, surpassing the state-of-the-art models for the recognition of continuous assembly activities.
- 7 March 2023
- Published by Wiley in Journal of the American Ceramic Society
- Vol. 106 (7) , 4028-4037
Additive manufacturing of high mechanical strength continuous C f /SiC composites using a 3D extrusion technique and polycarbosilane‐coated carbon fibers
A novel additive manufacturing approach is herein reported for manufacturing high mechanical strength continuous carbon fiber‐reinforced silicon carbide (C f /SiC) composite materials. Continuous carbon fibers were coated with polycarbosilane (PCS) using a colloidal evaporative deposition process and then coextruded with high solid content SiC ink. The zeta potential of the SiC ink was adjusted to optimize the printing ability of the suspension. During sintering, small SiC grains and whiskers were generated in the gaps in and around the PCS‐coated carbon fibers, which led to the improved flexural strength and density of the composites. Meanwhile, the PCS coating on the surface of the carbon fibers prevented the carbon fibers from reacting with SiO gas generated by reactions between the SiC matrix and SiO 2 and sintering additives (Al 2 O 3 and Y 2 O 3 ), effectively preserving the structural integrity of the carbon fibers. Compared to the SiC specimens containing uncoated carbon fibers, the density of the specimens fabricated with coated carbon fibers was increased from 2.51 to 2.85 g/cm 3 , and the strength was increased from 190 to 232 MPa.
- 7 March 2023
- Published by Wiley in Journal of the American Ceramic Society
- Vol. 106 (7) , 4028-4037
Additive manufacturing of high mechanical strength continuous C f /SiC composites using a 3D extrusion technique and polycarbosilane‐coated carbon fibers
A novel additive manufacturing approach is herein reported for manufacturing high mechanical strength continuous carbon fiber‐reinforced silicon carbide (C f /SiC) composite materials. Continuous carbon fibers were coated with polycarbosilane (PCS) using a colloidal evaporative deposition process and then coextruded with high solid content SiC ink. The zeta potential of the SiC ink was adjusted to optimize the printing ability of the suspension. During sintering, small SiC grains and whiskers were generated in the gaps in and around the PCS‐coated carbon fibers, which led to the improved flexural strength and density of the composites. Meanwhile, the PCS coating on the surface of the carbon fibers prevented the carbon fibers from reacting with SiO gas generated by reactions between the SiC matrix and SiO 2 and sintering additives (Al 2 O 3 and Y 2 O 3 ), effectively preserving the structural integrity of the carbon fibers. Compared to the SiC specimens containing uncoated carbon fibers, the density of the specimens fabricated with coated carbon fibers was increased from 2.51 to 2.85 g/cm 3 , and the strength was increased from 190 to 232 MPa.
- 17 June 2022
- Published by Elsevier in Ceramics International
- Vol. 48 (19) , 28444-28454
Additive manufacturing of complexly shaped SiC with high density via extrusion-based technique – Effects of slurry thixotropic behavior and 3D printing parameters
- 17 June 2022
- Published by Elsevier in Ceramics International
- Vol. 48 (19) , 28444-28454
Additive manufacturing of complexly shaped SiC with high density via extrusion-based technique – Effects of slurry thixotropic behavior and 3D printing parameters