Welcome to my portfolio!

Graduate Projects

Master’s Thesis Mid-Term Report: “States estimation of lithum ion battery using data driven methods”

Key Skills: Matlab, Python (PyTorch, scipy, sklearn), Battery SOH & RUL estimation

This shows the progress report on the Master Thesis following the battery cycling experiments and AI based model especially IPEformer which is the lightweight state-of-art time series transformer (TST) for battery SOH and RUL estimation.

It also shows the battery cycling experiments on 2200 mAh cyclindrical cells conduted in lab.
There are four other dataset used for model validation on public dataset as well.
67 Handcrafted feature extraction for all five datasets using matlab (Check out the Slide pdf for more details).

PHM Conference Paper: “Battery Prognostics and Health Management Using CNN-BiGRU with Temporal Attention on CS2 Cells”

Key Skills: Python (PyTorch, scipy, sklearn), Battery PHM, CNN-BiGRU-Temporal Attention, BMS

Abstract: Lithium-ion batteries (LiBs) are reliable and efficient energy sources; they are extensively employed worldwide in modern energy storage systems. The State of Health (SOH) and Remaining Useful Life (RUL) are considered some of the critical parameters for monitoring battery performance and safety. SOH and RUL reflect the battery’s overall condition, functionality, and remaining lifespan. Their accurate estimation is essential for ensuring safe operation, effective energy management, and extension of battery’s service life. However, to meet the demands of real-world applications, the accuracy of SOH and RUL estimation must be further improved. This research work proposes a Deep Learning (DL) based method that combines Convolutional Neural Network (CNN), Bidirectional Gated Recurrent Unit (BiGRU), and temporal attention for the SOH and RUL estimation of LiBs. The proposed architecture was tested through leave-one out cross-validation (LOOCV). It achieved a Mean Absolute Error (MAE) of 0.0072, a Root Mean Squared Error (RMSE) of 0.0121, and a coefficient of determination (R2) of 0.9965 on group I of the Center for Advanced Life Cycle Engineering (CALCE) CS2-type LiBs dataset. Compared prediction results with other DL based models, including CNN, Long Short-Term Memory (LSTM), CNN-Bidirectional LSTM (CNN-BiLSTM), and BiLSTM-Gated Recurrent Unit (BiLSTM-GRU), showing that our proposed method performed better in term of SOH and RUL estimation and also providing better solution for battery management system (BMS).

Master’s Thesis Prposal: “State parameter estimation of Li-ion battery using Deep Neural Networks and Equivalent circuit model”

Key Skills: Python (PyTorch, scikit-learn, fastai, PyBamm), `Matlab and Simulink, Battery modelling, Battery testing, BMS

With the comprehensive review of Battery Management System(BMS) and various battery modeling techniques, it is evident that the BMS plays a crucial role in enhancing battery performance, extending lifespan, and ensuring protection against damage.
Among the different modeling approaches, the Equivalent Circuit Model (ECM) is the most widely used due to its simplicity and effectiveness. On the other hand, the Electrochemical Model provides highly accurate state predictions by incorporating detailed internal cell parameters, but it requires significant computational resources and is challenging to implement.

The Data-Driven Modeling Approach, often referred to as a “black-box model,” has shown promising results. It is relatively easy to set up but heavily relies on high-quality experimental training data. Therefore, extensive testing, such as battery cycling and Electrochemical Impedance Spectroscopy (EIS), is necessary to develop reliable models for commercialLithium-ion Batteries. Developing an accurate Data-Driven Battery Model using python(PyBamm, PyTorch, fastai) and implementing it in the MATLAB/Simulink environment for Electric Vehicle (EV) simulations, considering various driving profiles and ambient temperature conditions.

Lithium ion battery State of Charge estimation using Bidirectional Long Short Term Memory(BiLSTM)

Key Skills: Python (numpy, pandas, matplotlib, tensorflow, seaborn), time series forecasting

This project uses dataset of 3Ah LG HG2 battery cell tested in various test conditions and temperature. Sequential model from tensorflow with input layer, bidirectionaLSTM layer and three forward layers with selu activation function and an output layer with linear activation function. The model is trained in google colaboratory using T4 GPU and it took about 18 minutes to train on full data. The Huber losss is used for minimization with Adam optimizer from tensorflow.keras library which achieved loss: 2.2698e-04 , mae: 0.0176, mape: 8.3553, rmse: 0.0213 on traning data and loss: 6.8946e-04, mae: 0.0281, mape: 11.0483, rmse: 0.0368 on testing data.

Neural networks and linear regression from scratch in python using numpy array

Key Skills: Python (numpy, pandas, matplotlib, scikit-learn)

A machine learning project showcasing Linear Regression (LR) and Neural Net Regressionbuilt from scratch. Achieved high accuracy on the California housing dataset< (sklearn), with LR: MAE 0.5539, RMSE 0.7404 and Neural Net: MAE 0.5365, RMSE 0.7214. Comparing with Sklearn’s LinearRegression : MAE 0.5332, RMSE 0.7456,

Stock Price Visualization Web App

Key Skills: Python (yfinance, Streamlit, pandas), Financial Data Analysis, Time-Series Visualization, API Integration

A web-based stock price visualization tool that retrieves and displays historical stock data using yFinance and Streamlit. The app allows users to analyze closing prices and trading volumes of stocks, with a default focus on Google (GOOGL).

The project features an interactive interface where users can enter stock tickers, select date ranges, and generate real-time time-series charts. It leverages yfinance for retrieving financial data, pandas for data manipulation, and Streamlit for building an interactive UI.

Undergratuate Projects

Bachelor’s Thesis: “Numerical simulation of supersonic flow around civil aircraft and sonic boom prediction”

Key Skills: CFD Softwares (such as Star ccm+, Ansys fluent),Solidworks, bBoom, MATLAB, Python, Tecplot

This paper addresses the challenge of excessive sonic boom intensity in the design of next-generation supersonic passenger aircraft. It explores sonic boom prediction through four key areas:

Reviewing and selecting appropriate prediction methods for conceptual and preliminary aircraft design, emphasizing the importance of minimizing sonic booms for future viability;
Developing a near-field sonic boom calculation method using CFD software, refining grid generation, and comparing mesh results;
Utilizing the bBoom program for far-field sonic boom propagation, analyzing the impact of atmospheric conditions and parameters, and comparing the JWB and DWB models, with JWB producing weaker shocks; and
Identifying challenges in accurately computing aft portions of the sonic boom signature, particularly in wake and plume zones, highlighting the need for improved meshing techniques and future research on faster and more precise near-field solutions.

Course Project: “Remote Control Small Aircraft Design”

Key Skill: Solidworks, RC-phoenix, Laser cutting, 3D-printing

Focused on small lightweight air frames using laser cutting and 3D printing, optimized aerodynamics, efficient power systems, and control surfaces for stability, maneuverability, and efficient flight performance, Supported by test flights.

Course Project: ” Conceptual Design of Business Jet”

Key Skill: Solidworks, Star ccm+, Flight dynamics, aerodynamic design

Optimizing jet performance through flight dynamics calculation using mathematica, aerodynamic design using STAR-CCM+, efficient propulsion, and passenger comfort. Integrated sustainable technologies to reduce environmental impact. CAD design using Solidworks

Electronics Lab Project: “Potable Radio Design”

Key Skill: PCB design, C++, soldering and wiring electronics

Responsible for designing a radio PCB, incorporating RF circuit design, antenna interface, signal processing, power supply, optimized trace routing, and tests to ensure efficient signal transmission and minimal interference.

Summer Camp Project: “Small Sounding Rocket Design”

Key Skill: solidworks, teamwork, group project, Laser cutting, 3D-printing

Responsible for CAD design manufacturing and presentation of a small sounding rocket with lightweight air-frame for strength, using solid thrusters, provided payload space for sensors, ensured stability with fins/canards, and included recovery systems such as parachute.