C++ based Monte Carlo Grain Boundary Simulation Model
At JSW, I undertook the development of a sophisticated Monte Carlo simulation model aimed at predicting grain boundary movements during the annealing process of steel samples. This process, characterized by its inherent uncertainty, required a robust framework to simulate the complex interactions occurring at elevated temperatures. I utilized the Monte-Carlo Potts model framework, integrating critical inputs such as the crystallographic texture and Kernel Average Misorientation of the initial material. These factors significantly influence the final grain structure after annealing. The model was designed to simulate boundary movements with high precision, allowing us to optimize the annealing process by adjusting variables like temperature and soaking time.
The results were remarkable: by applying this model, we managed to reduce the net annealing time by 25%, which translated into substantial cost savings—over 12 crores in projected impact. The success of this project not only enhanced our operational efficiency but also provided valuable insights that were recognized at the IIM-ATM 2023 conference, where I had the opportunity to present our findings. This work demonstrated the power of combining advanced simulation techniques with material science, driving innovation and efficiency in industrial processes.
You can access the entire poster here.
Moldboard Retention Screw system for lowering the maintenance time
The project focused on developing a new retention system for the moldboard cutting edge, which has been conceptualized and modeled as a cam-based mechanism to replace the existing nut-bolt assembly. Through the assessment of research papers on retention systems in various industries, a novel numerical method has been established to develop a cam profile capable of reaching a maximum preloading of 176kN. Additionally, a mathematical model has been formulated to incorporate non-linear profiled Belleville springs, enhancing the system's performance and wear resistance. These advancements have resulted in a significant reduction in service and assembly time by more than two-fold, emphasizing the project's success in meeting its objectives of efficiency and cost reduction.
Quantum Phononics
The aim of my dual degree project was to design an electromagnetic mechanical resonator utilizing Carbon Nanotubes, engineered to generate antibunched phonons, the quantized units of sound. A central challenge of this research was to elevate the operational temperature limit, within which phonon antibunching remains effective, from well below 1 Kelvin to above 1 Kelvin. The system I proposed consisted of an array of Carbon Nanotubes, which were excited by an external electromagnetic force. Through quantum simulations, I identified a robust antibunching region in a system comprising three nanotubes. Subsequently, I applied non-local Finite Element Method (FEM) analysis to these nanotubes and performed a comprehensive parametric study on critical factors such as nanotube diameter, length, and support structures. My findings demonstrated that the operational temperature could be increased to 1.368 Kelvin while preserving the strong antibunching regime, marking a significant advancement in the field.
You can access a detailed report here.
YOLO v3 based Image Detection Model for Production Drawings
In addition, I developed a cutting-edge Deep Learning model in Python based on the YOLOv3 architecture for precise detection of arrows in CAD drawings. Through comprehensive training and validation, this model achieved high accuracy in its predictions, validating the effectiveness of both the model and the Data Preparation tool. These projects highlight my expertise in software development, machine learning, and innovative problem-solving.
During my internship at AmadaSoft India, I contributed to the development of an advanced Data Preparation tool in C++ designed specifically for training Convolutional Neural Networks (CNNs). This tool, engineered using OpenCV and Windows Forms, facilitated streamlined image processing and featured a user-friendly graphical interface. By implementing sophisticated techniques, such as Edge Detection, I successfully halved the data preparation time, demonstrating my capability to optimize workflows and improve operational efficiency.
C# based GUI for Patent Summary storage and sharing
I developed a C#-based patent summary system designed to streamline access to competitor patents, enhancing the speed and efficiency of patent research. This system utilized cloud infrastructure to facilitate seamless knowledge sharing and ensure that users could quickly retrieve and analyze relevant patent information. By integrating cloud services, the solution not only improved accessibility but also supported collaborative efforts in patent analysis, thereby providing valuable insights and fostering a more agile approach to competitive intelligence.
You can access my GitHub repository here.