Percorrer por data de Publicação, começado por "2025-02-14"
A mostrar 1 - 3 de 3
Resultados por página
Opções de ordenação
- Generation of Synthetic Data Using AIPublication . MELO, BRUNO OLIVEIRA; Santos, Veríssimo Manuel Brandão LimaThe process of accurately monitoring tire pressure in a vehicle is crucial for safety, tire longevity, and fuel efficiency, is present in modern cars due to regulatory changes. Indirect Tire Pressure Monitoring System (iTPMS), one of the systems that monitors tire pressure, is used by accessing already existent sensor data in the vehicle to determine if there is a Tire Pressure Loss (TPL). Despite it being cost-effective and easier to integrate and maintain, it relies on data acquired in a controlled environment to fine-tune the final production vehicle Tire Pressure Monitoring System (TPMS). This work explores the generation of synthetic data for these tests using Artificial Intelligence (AI) techniques, specifically focusing on Machine Learning (ML) models to simulate realistic sensor data to be used for iTPMS development and therefore increase testing capabilities while significantly reducing costs involved in current development, providing a scalable solution for dataset expansion. The generation process utilizes regression techniques to model complex relations between different vehicle parameters and sensors, including data preprocessing, model training, and a validation phase, ensuring fidelity to real-world scenarios. Additionally, a Graphical User Interface (GUI) solution is provided, bridging the gap between the ease of use, and users without technical knowledge of ML solutions, providing easy-to-read interfaces and straightforward variable entry for synthetic data generation. Both the models and GUI solution are created using MATLAB.
- IoT Based Automated Moving System for Weaving InspectionPublication . CARVALHO, JOÃO FRANCISCO FERREIRA MOREIRA DE; Figueiredo, Lino Manuel BaptistaThis dissertation explores the significant impact of the textile industry on waste production, focusing on Smartex’s commitment to eco-friendly practices. Smartex, primarily active in the knitting market, aims to expand its sustainable solutions to the weaving sector. The crux of this project was to develop an automated moving system for weaving inspection. This system integrates innovative technologies such as linear motion solutions, Internet of Things (IoT), and communication protocols, highlighting a synergy between mechanical engineering and electrical and computer engineering principles. The hardware framework of the system comprises dual lead screws driven by stepper motors, managed via a Raspberry Pi 4 and an HR8825 motor driver. This setup is controlled by software that operates within the MQTT network. This network is distinctive for its dynamic election broker system, which enhances the automation’s reliability by providing network redundancy crucial for industrial applications. While the MQTT automated solution demonstrated success, the mechanical aspect faced challenges, signaling potential future precision issues due to slippage. This outcome highlights a disparity between the project’s complexity and Smartex’s philosophy of creating simplified, easy-to-install products. The project, therefore, serves as a conduit for applying academic and industry knowledge, underscoring the value of such collaboration for personal and professional development.
- Indoors Radar Detection: Performance Analysis and Development of Tracking AlgorithmsPublication . MELO, MARTIM COELHO DE; Dias, André Miguel PinheiroThe application of radar technology within indoor environments presents a challenge owing to the complex and dynamic nature of such spaces. Accurate tracking and classification of objects in indoor scenarios require specialized solutions to surmount the myriad of challenges posed by factors like clutter, multipath propagation, and interference. This work presents a comprehensive analysis of an indoor radar-based object tracking system, aiming to address these challenges. The proposed system leverages radar sensors placed strategically within the environment and employs Kalman filter variants, including the Unscented Kalman Filter and the Standard Kalman Filter, for tracking. Through a rigorous evaluation process encompassing various parameter configurations and testing across multiple scenes and trajectories, the system’s performance is assessed. The evaluation highlights the impact of parameters like process noise, minimum cluster points, and epsilon value on tracking accuracy. The results showcase that the system achieves an average error of 0.182 m across diverse indoor scenarios and movement patterns.