Project Portfolio

AI-Driven Digital Twins for Structural Health Monitoring with Statistical Modal Analysis and Scientific Machine Learning

• Duration: 2+ years
• Summary: Developed proprietary software modules for digital twins and structural health monitoring at CAEmate, contributing to components integrated into WeStatiX. My work focused on helping large civil structures, such as bridges and viaducts, be monitored continuously by developing computation code in my data science team. I built tools that transform raw vibration and environmental measurements into clear engineering indicators that can show how a structure is behaving over time. This included automated vibration analysis, signal processing, grouping similar structural responses, detecting unusual behavior, checking data quality, and reducing the need for manual tuning. I also worked on machine-learning models that approximate expensive physics-based simulations, making it faster to estimate loads, material properties, and calibration parameters from measured data. These workflows connected sensor measurements with finite-element models, validation reports, and repeatable monitoring pipelines, making advanced structural analysis more practical for large infrastructure assets.
• Technologies: Python, PyTorch, scikit-learn, XGBoost, NumPy, Pandas, SciPy, Matplotlib, Seaborn, DuckDB, Parquet/FastParquet, JSON/YAML, MariaDB (MySQL), Git, Linux shell.
• Fields: structural health monitoring, digital twins, vibration analysis, machine learning, signal processing, time-series analysis, numerical optimization, civil engineering, engineering physics.

PhD Dissertation: Hybrid Machine Learning and Numerical Analysis of Cartilage Biomechanics

• Duration: 4 years and 6 months.
• Summary: This research introduces data-efficient AI algorithms, such as hybrid graph neural networks, for simulating tissue biophysics, focusing on cartilage biomechanics. It also presents a novel approach to generating high-fidelity training data using advanced finite element methods and optimization algorithms. The proposed methods efficiently scale across different physical fidelities and scales. This dissertation is written in English, based on my published papers and code, under the supervision of Prof. Bruno Carpentieri and Prof. Gerhard A. Holzapfel.
• Technologies: Python (TensorFlow, Keras, Matplotlib, NumPy, Sklearn, Pandas, SciPy, and Abaqus scripts) and Fortran (Abaqus subroutines).
• Fields: Data science (machine learning), applied mathematics (numerical analysis and optimization), and engineering physics (biomechanics).

MSc Thesis: Computational and Biomechanical Investigation into the Degeneration of the Main Articular Cartilage Constituents in Osteoarthritis

• Duration: 2 years and 6 months.
• Summary: This research presents a novel and accurate computational model using finite element analysis to simulate osteoarthritis, i.e., the degeneration of joint substructures, particularly articular cartilage. The study highlights the critical role of subchondral bone changes in fluid movement through the joint, affecting both chemical and solid components. This thesis is primarily in Persian (nearly the same as my first journal paper in English), supervised by Prof. Mohammad Haghpanahi and Prof. Mohammad Razi.
• Technologies: Python (Abaqus scripts) and Fortran (Abaqus subroutines).
• Fields: Applied mathematics (numerical analysis) and engineering physics (biomechanics).