Acoustic Porosity Mapping in Laser Powder Bed Fusion
Developed a machine learning pipeline that maps airborne acoustic signals to sub-scanline porosity
in LPBF to enable more reliable, real-time quality monitoring, using synchronized acoustic and CT data.
- My contribution: Designed the full pipeline from signal preprocessing to CT-aligned porosity prediction; led model training, evaluation, and ablation analysis.
- Tech stack: Python, PyTorch, NumPy, Matplotlib, MATLAB
- Outcome/Impact: Achieved accurate sub-scanline porosity quantification (R² ≈ 0.xx on held-out builds); results released as a 2025 arXiv preprint.
Closed-loop Process Monitoring and Control for MIG Welding
Built a multi-sensor and machine learning framework for MIG welding to reduce weld defects and
stabilize penetration by enabling real-time monitoring and parameter adjustment.
- My contribution: Identified key defect types, selected acoustic/arc-light/electrical sensors, designed the data collection pipeline, and sketched the closed-loop control logic with an industry welding partner.
- Tech stack: Python, PyTorch, MATLAB, data acquisition hardware, MIG welding equipment
- Outcome/Impact: Delivered a synchronized multi-modal sensing setup and baseline models that are now being prepared for closed-loop tests to improve weld quality and process stability.
Multimodal Melt Pool Prediction and Visualization in LPBF
Designed deep learning models that use acoustic and photodiode signals to reconstruct and visualize
melt-pool behavior in laser powder bed fusion, helping engineers understand how process settings
affect pool stability.
- My contribution: Collected and preprocessed sensor data, implemented multiple PCA+deep learning pipelines, and compared models using common error metrics and visual inspection of reconstructed melt-pool images.
- Tech stack: Python, PyTorch, NumPy, Matplotlib, LPBF experimental platform
- Outcome/Impact: Identified architectures that best capture melt-pool shape and dynamics and produced visual tools that link process parameters, signals, and melt-pool behavior for later use in monitoring.
Simulation and Optimization of Mechanical Connection Process
Simulated the self-piercing riveting (SPR) process to design and select die geometries that produce
strong and reliable joints while keeping the die model simple to use in industry.
- My contribution: Proposed a parameterized die concept, defined performance metrics for comparing dies, and ran SPR simulations for each candidate die to guide selection.
- Tech stack: Simufact.forming, SolidWorks, basic scripting for data processing
- Outcome/Impact: Recommended a die design with improved performance and a clear evaluation method that can be reused for future SPR tooling studies.
Personal Mobility Device for People with Disabilities
Designed a wheelchair tailored to user needs to improve comfort and daily mobility for people with
disabilities, moving from requirements to a manufacturable CAD model.
- My contribution: Built a PUGH matrix from technical requirements, created detailed CAD models with assemblies and animations, and checked strength using simple stress analyses.
- Tech stack: CATIA, SolidWorks, basic engineering design methods (PUGH matrix, requirements analysis)
- Outcome/Impact: Delivered a complete wheelchair design package with clear visuals and design rationale that could be handed off for prototyping.
Design and Build of a Simple Vehicle
Worked in a student team to design, simulate, and build a small vehicle, turning classroom mechanics
and design theory into a working prototype.
- My contribution: Helped size and evaluate key parameters for the frame and power system, modeled the vehicle in CAD, and assisted with welding, assembly, and on-site testing.
- Tech stack: SolidWorks, basic structural simulation, workshop tools for fabrication and welding
- Outcome/Impact: Completed a functional vehicle that met project performance targets and passed basic strength checks, giving the team hands-on experience in the full design–build cycle.