Abstract

Title:  Photoelastic Insight into Stress Distribution on Curved Surfaces with Singularities 

Speaker:   RAMIT CHUGH (Chhattisgarh Swami Vivekanand Technical University, Bhilai)

Abstract: This abstract presents a comprehensive investigation into stress analysis in curved surfaces with singularities employing the innovative technique of photoelasticity. The study delves into the complex interplay of stresses on curved surfaces, which are often encountered in various engineering applications. The utilization of photoelasticity as a powerful experimental method enables a direct and visually insightful examination of stress distribution in such intricate geometries.  The research focuses on identifying and understanding singularities in curved surfaces, where stress concentrations may lead to critical failure points. By employing advanced photoelastic models and experimental setups, the study aims to characterize the stress patterns associated with these singularities and elucidate their impact on structural integrity. The experimental data obtained through the photoelastic technique will be analyzed using sophisticated image processing and numerical methods to extract quantitative information about stress magnitudes, orientations, and distribution across the curved surfaces.  Furthermore, the abstract highlights the significance of this research in enhancing the design and analysis of engineering structures subjected to complex loading conditions. The outcomes of this study will contribute valuable insights into mitigating stress-related issues in curved surfaces, ultimately leading to the development of more resilient and efficient structures. This work not only advances the understanding of stress analysis in curved surfaces but also underscores the practical applications of photoelasticity as a robust tool for investigating stress complexities in real-world engineering scenarios.

Title:  Fault Diagnosis in Rolling Element Bearings using Fast Fourier Transform Analysis and Machine Learning Algorithms

Speaker:   Devendra Sahu (Chhattisgarh Swami Vivekanand Technical University, Bhilai)

Abstract: This study explores two distinct approaches for fault diagnosis in rolling element bearings, namely Fast Fourier Transform (FFT) analysis and machine learning techniques. Rolling element bearings are critical components in industrial machinery, and the timely identification of faults is essential for preventing unexpected failures and optimizing maintenance strategies. The FFT analysis is employed as a signal processing technique to transform vibration signals from bearings into the frequency domain. The resulting frequency spectra, amplitudes, and associated features serve as input for traditional diagnostic methods. Concurrently, machine learning techniques, including Support Vector Machines, Random Forests, and Neural Networks, are utilized to develop fault diagnosis models using raw vibration data.   The research involves a comparative analysis of the effectiveness of FFT and machine learning in identifying various types and severities of bearing faults. Experimental data from simulated and real-world fault scenarios are used to assess the performance of each approach. Metrics such as accuracy, precision, recall, and F1 score are employed to quantify and compare the diagnostic capabilities of both methods.  Results from the comparative analysis provide insights into the strengths and limitations of FFT and machine learning techniques for rolling element bearing fault diagnosis. This research contributes valuable information to the field of condition monitoring and predictive maintenance, aiding in the selection of appropriate methodologies based on specific application requirements. The findings aim to guide practitioners in choosing the most suitable approach for enhancing machinery reliability and operational efficiency in industrial settings.

Title:  QuITO: a direct trajectory optimization algorithm for constrained optimal control problems

Speaker:  Siddhartha Ganguly (IIT Bombay)

Abstract: One of the key challenges in the field of optimal control and dynamic optimization is to design efficient and tractable numerical algorithms for producing optimal state-action trajectories under path constraints — constraints on the state-action trajectory at each instant of time. Such problems can be tackled broadly by Indirect and Direct methods. Indirect methods are based on the well-known Pontryagins Maximum Principle (PMP) which uses tools from the calculus of variations and functional analysis, while direct methods see the optimal control problem through the glasses of optimization. In this talk, I will introduce a novel direct trajectory optimization algorithm — QuITO (Quasi-Interpolation based Trajectory Optimization) that employs tools from approximation theory and signal processing to furnish constrained state-action trajectories with tight and uniform guarantees of convergence for a wide class of OCPs. 

Title:  Attitude Control of Quadcopter by Koopman Framework

Speaker:   Akumalla Ravi Kiran (IIT Mandi )

Abstract: The work investigates on designing an experimental framework for forest fire fighting with implementation of robotic technology namely swarm of drones. Considering the requirements in the application, the existing techniques of time varying formation of the swarm, fault identification, isolation and reformation of the swarm is given more importance. To enhance the reliability of real time application, the above-mentioned techniques are being solved by a data driven approach namely Koopman framework where a global linear representation of a nonlinear system can be obtained purely from the time domain data. The current presentation mainly focuses about controlling the attitude of a quadcopter using Koopman Framework.

Title:  Advancements in formation control for multi-agent systems

Speaker:   Ankush Thakur (IIT MANDI )

Abstract: This work investigates cooperative time-varying formation tracking (TVFT) control for a class of Lipschitz nonlinear multi-agent systems subjected to actuator failures, ensuring runtime formation switching and collision avoidance. 

Title:  Natural Language Processing in Decision Optimization: A Text Summarization Endeavor

Speaker:   Hardik Sharma (IGNOU)

Abstract:  This presentation unveils a comprehensive project centered on the integration of Natural Language Processing (NLP) to construct an intelligent text summarization system tailored to optimize decision-making processes. The project strategically addresses real-world challenges associated with information overload, illustrating the transformative potential of NLP in augmenting decision support frameworks.  Key Project Components:  Problem Statement: An introductory exploration of the complexities introduced by information overload within decision-making scenarios, emphasizing the imperative for an automated text summarization solution.  NLP Techniques Applied: A meticulous examination of the nuanced application of NLP algorithms for text analysis and comprehension, coupled with the sophisticated processes of feature extraction and language modeling to facilitate effective content summarization.  Implementation Details: A detailed overview of the technical architecture and tools constituting the project's framework, complemented by a demonstrative exhibition of the NLP-driven text summarization algorithm in action.  Project Evaluation: An analytical approach featuring metrics employed to assess the quality and effectiveness of the summarization process. This segment includes a comparative analysis with traditional methods, elucidating the impact on decision-making speed and accuracy.  Use Cases and Applications: A systematic exploration of diverse applications across industries, encompassing domains such as finance, healthcare, and news. Real-world scenarios will be presented to underscore the project's capacity to address specific decision-making challenges.  Challenges and Lessons Learned: A reflective discussion on the challenges encountered throughout the developmental phase of the project, accompanied by the insights gleaned and lessons learned in the pragmatic implementation of NLP for decision support.  Future Extensions and Open Source Collaboration: A forward-looking discussion encompassing potential enhancements and extensions to the project, concluding with an invitation for collaborative participation and contributions to the open-source initiative.

Title:  Preliminary Investigation of Vision-based UAV– 4WISD AGV System for Collaborative Missions

Speaker:   Ashok Kumar Sivarathri (IIT Mandi )

Abstract: Tracking motion between UAV (Unmanned Aerial Vehicle) and AGV (Autonomous Ground Vehicle) is one of the basic collaborative tasks of the system. It can be either UAV tracking motion of AGV or its inverse way. UAV is spatially omnidirectional, thus, trivial task to track the motion of AGV. However, AGV must be able to track UAV with the equal ease for effective navigation and overall agility of the system. However, AGV tracking the motion of UAV is given less attention in the literature considering vision-based system. Non holonomic AGVs may not be suitable for effective navigation of the system and macanum wheeled vehicles are not suitable for outdoor missions. In this work, we consider the combination of 4WISD (4 Wheeled Independent Steering and Driving) vehicle and UAV to study the impact on collaborative navigation of the system. Vision-based tracking of UAV by 4WISD ground vehicle is considered in this study to analyze the system. As 4WISD ground vehicles are expensive in the market from the research perspective, a low-cost 4WISD vehicle is designed and fabricated in the current research. ROS (Robot Operating System) based control package is developed for the ground vehicle. 4WISD vehicle is localized using low-cost monocular camera (attached to UAV) and a magnetometer sensor. Sliding mode kinematic controller has been designed for 4WISD vehicle and verified using Lyapunov criteria. Tracking experiments are performed combining UAV with both non-holonomic and 4WISD vehicles and qualitative comparison is drawn. Experimental results show flexible tracking motion of 4WISD vehicle and indicate that combination of UAV and 4WISD may improve the collaborative navigation of the system compared to non-holonomic vehicle.

Title:  Adaptive fuzzy control of 3D overhead cranes

Speaker:  Somesh Swami (Institute of Infrastructure Technology Research and Management, Ahmedabad )

Abstract: In practical scenarios, overhead gantry cranes often display dynamic characteristics akin to a double-pendulum system. Controlling an underactuated system like a crane with 6-DOF is intricate, nonlinear, and involves coupling. Given the widespread industrial use of cranes, mitigating damages caused by swinging is imperative. In addition it is difficult to obtain satisfactory control performance due to the uncertainty of the parameters of the crane system. Aiming at these problems, we propose a block-backstepping based control approach to tackle the trajectory-tracking challenge for a three-dimensional overhead crane with a double-pendulum effect. The uncertainty in parameters are adaptively estimated and inferred by utilizing the fuzzy inference rule mechanism, which results in efficient operations of the crane in real time. More importantly, stabilization of the crane controlled by the proposed algorithm is theoretically proven using the Lyapunov function. The proposed control strategy is verified using Matlab/Simulink simulation tool, which show that the presented controller provides good performance (i.e., fast transient response, position tracking, and low swing angle) even under input disturbances.