Smart Manufacturing lab
Smart Manufacturing Lab, a place where advanced technology and innovation come together. We are dedicated to pushing the boundaries of manufacturing through cutting-edge research and development. Our team of experts is committed to discovering new solutions and improving existing processes to create more efficient and effective manufacturing systems. The lab covers areas such as Big Data Analytics, IIoT, Mechatronics, Cyber Security, Robotics, Computer Vision, Connected Manufacturing, Cloud Computing, Digital Twins, etc. We offer Virtual and Remote Labs along with our courses. Virtual Reality Lab is a work in progress. The virtual lab runs with the software tools, and the remote lab is administered through the hardware connected via remote desktop connections.
SUBJECT EXPERTS
Dr. Nithin Tom Mathew
nithin.mathew@pilani.bits-pilani.ac.in
Prof. SS Kendre
kendres.shankarrao@wilp.bits-pilani.ac.in
Arunkarthik Thangaraj
arun.karthik@wilp.bits-pilani.ac.in
Aditya Goel
aditya.goel@wilp.bits-pilani.ac.in
Subrata Karmakar
subratakarmakar@wilp.bits-pilani.ac.in
Sivaraman Shankaran
sivaram.sankaran@wilp.bits-pilani.ac.in
Vennish Muthu
vennishmuthu@wilp.bits-pilani.ac.in
Anagha Bhat
anagha.bhat@wilp.bits-pilani.ac.in
PROJECT DETAILS
Application of Embedded Systems in Industry 4.0 - Smart Inventory Management System
The project focuses on efficient stock management, real-time inventory monitoring, and automation for accuracy and precision. RFID-based systems integrate sensors to track inventory levels dynamically and send the data to a cloud platform for real-time visualization.
Hardware and Software Requirements:
Microcontroller: ESP32, Temperature and Humidity Sensor: DHT-11 Load Cell Module with HX711 Amplifier, RFID RC522 Tag Reader.
Arduino IDE for microcontroller programming, Thingspeak for IoT data visualization, Python for middleware scripting.
Learning Outcomes:
Microcontroller programming and hardware integration proficiency
Real-time data processing logic development
RFID technology implementation understanding
IoT concepts and error handling
Wear and Tear Analysis of Micro Tools using Digital Image Processing
This project involves using Digital Image Processing and Machine Learning techniques to analyze images of microtools, estimate their wear and tear, and predict their lifespan. A dataset representing various wear states is created, and machine learning models are developed to achieve predictions.
Hardware and Software Requirements:
Imaging equipment for capturing microtool conditions, TensorFlow, Neural Networks, and Deep Learning frameworks, Python for preprocessing and model training
Learning Outcomes:
Proficiency in machine learning and image processing techniques
Understanding wear and tear prediction in industrial tools
Practical knowledge of predictive maintenance
AI Powered QnA Chatbot using Mistral-7B
This project involves developing a chatbot system capable of understanding and answering user queries based on Standard Operating Procedures (SOPs). By leveraging transformer models and Large Language Models (LLMs), the system can provide precise and relevant answers to queries about specific SOPs. It processes input text, matches it with SOP content, and returns accurate, context-aware responses.A vector database is generated by processing the document. The Mistral model generates responses for the prompts accordingly. • As everyone knows, performance of bot depends upon the computational power available- the higher the GPU/CPU the faster will the document be processed and generation of responses.
Hardware and Software Requirements:
i-9 processors with 32 GB GPU and 120 GB RAM generally in professional environments. Python (3.8+), Transformers library
Streamlit for interactive web interface
Database management with MySQL or MongoDB
Pre-trained LLMs for query processing
Learning Outcomes:
Proficiency in implementing transformer models for natural language understanding
Development of chatbot systems with real-time response capabilities
Integration of LLMs for context-specific information retrieval
Web-based user interface creation and backend management
The project overall improved my understanding of real-world RAG application building from scratch. The use of pre trained LLMs and their applications was really interesting.
Conversational Retrieval Chain: Built a ConversationalRetrievalChain using LangChain, combining retrieval-based search with conversational capabilities.
Memory Integration: Integrated ConversationBufferMemory to retain context across multiple queries.
Question-Answering Flow: Designed pipeline to handle user queries and display generated answers dynamically.
Chunk Splitting: Leverage Text Splitter to divide large documents into manageable text chunks for efficient processing.
Vector Store: Implemented FAISS for vectorizing and storing document chunks, enabling fast response generation.
Multi-Document Support: Built a system to handle multiple document uploads and integrate their data for chatbot interactions.
Scalable Design: Incorporated modular components (LLM, embeddings, vector store) to adapt to various use cases.
Model Configuration: Gained hands-on experience configuring LlamaCpp and also adjusting hyperparameters like n_ctx, top_p etc.
Integration of External Libraries: Learned to integrate third-party libraries such as HuggingFace, FAISS, and LangChain effectively.
Building a Digital Twin for a Water Bottling Plant
The project involves creating machine learning models for bottling plant components (heater, cooler, mixer) and integrating these models in MATLAB/Simulink for real-time simulations. Data exchange is facilitated using a MySQL database and a Python middleware for dynamic parameter updates.
Hardware and Software Requirements:
Components for testing water bottling operations, Python (Numpy, Pandas, Scikit Learn), MATLAB and Simulink, MySQL and JDBC Connector.
Learning Outcomes:
Effective modeling of hardware appliances using simulation tools
Real-time integration of machine learning models with industrial processes
Data handling and communication between platforms
Dynamic Intelligent Questionnaire System
A system that dynamically adjusts the difficulty of assessment questions based on user performance. It uses semantic similarity for evaluating responses and tracks user progress for detailed insights.
Hardware and Software Requirements:
Python (3.8+), Streamlit for web development, SentenceTransformer model for semantic similarity, MySQL for database management.
Learning Outcomes:
Proficiency in adaptive algorithms and semantic similarity techniques
Backend integration and database management
Web development skills with Streamlit
STUDENT DETAILS
Sayantan Patra : f20202297@hyderabad.bits-pilani.ac.in
Venkata Anirudh Iragavarapu : f20201783@pilani.bits-pilani.ac.in
Varshith Srinivasa Peddada : f20211660@hyderabad.bits-pilani.ac.in
Satyam Sharan : f20212985@goa.bits-pilani.ac.in
Anirudh L Reddy : f20212581@goa.bits-pilani.ac.in