Politecnio die Bari - Catalogo di prodotti della Ricerca
Not a member yet
36616 research outputs found
Sort by
Using Attention for Improving Defect Detection in Existing RC Bridges
Get PDFIn the constantly growing need for sustainable mobility and transportation, on-site inspections of existing reinforced concrete (RC) bridges are critical in ensuring the safety of such infrastructures. However, surveying RC bridges presents several challenges, such as the high costs and effort required by the surveyors, the subjectivity in assessing identified defects, and the possible lapses of attention when inspections are systematically repeated on different bridges. Hence, traditional methods of on-site inspection can be enhanced by leveraging digital innovations and by developing new instruments that support road management companies in ensuring the safety of the existing infrastructure. Among the new technologies, deep learning-based object detection systems provide promising and effective solutions. As such, this research proposes a new, simple, intuitive and efficient tool to support engineers and surveyors in assessing the health state of existing RC bridges. To this end, domain experts gathered and labelled a dataset of real images containing typical defects found in existing RC bridges. Consequently, an improved version of YOLO11, embedding attention mechanisms to allow the network to focus on the most relevant details in each image, was trained, tested, and validated on the provided dataset, showing an overall improvement of quantitative metrics such as precision and recall, while retaining enough computational efficiency to allow real-time implementation on constrained devices. Visual explanations achieved via the Eigen-CAM algorithm were also exploited to evaluate the reliability of the predictions. The model was finally embedded in an end-to-end tool offering a graphical user interface (GUI) to allow an effective interaction between the domain expert and the machine. Overall, the proposal revealed its potential to improve the effectiveness of the survey, lowering the burden on surveyors and engineers and providing a reliable method to improve the overall security in large RC bridges portfolios
First Joint Oscillation Analysis of Super-Kamiokande Atmospheric and T2K Accelerator Neutrino Data
No full textThe Super-Kamiokande and T2K Collaborations present a joint measurement of neutrino oscillation parameters from their atmospheric and beam neutrino data. It uses a common interaction model for events overlapping in neutrino energy and correlated detector systematic uncertainties between the two datasets, which are found to be compatible. Using 3244.4 days of atmospheric data and a beam exposure of 19.7 (16.3) × 10^20 protons on target in (anti)neutrino mode, the analysis finds a 1.9σ exclusion of CP conservation (defined as J_CP = 0) and a 1.2σ exclusion of the inverted mass ordering
Low-thrust propulsion technologies for space mission applications
Get PDFIn this thesis, various aspects of physical modeling are developed to suggest the possibility that alternative low-thrust propulsion systems could be concretely employed in mission scenarios made attractive by recent developments in astrobiology. The systems considered are a solar sail using the additional contribute of thermal desorptiona and the Direct Fusion Drive (DFD) engine. The specific aspects developed in this thesis are the use of evaporable materials to increase the thrust in solar sails, the modeling of the transport of charged particles in magnetized plasmas such as those of interest for compact fusion reactors, and the simulation of specific mission scenarios. The methods used for this purpose are based on the kinetic theory of gases and plasma and on flight dynamics. The results obtained, in addition to the first integrated exposure of these two propulsion systems within specific scenarios, are: the study of the effect of evaporable material, the development of a high-performance computing program for the simulation of the dynamics of charged particles in magnetized plasmas, the detailed simulation of a scenario for reaching promising targets for deep space exploration
A review of laser-spectroscopy-based gas sensing techniques for trace formaldehyde detection
No full textFormaldehyde (H2CO) is a colorless gas with a strong irritating odor, widely used in furniture manufacturing and house decoration. Already at concentration in the few ppm range, H2CO represents great harm to human health, therefore, accurate measurement of formaldehyde concentration is of great significance for human safety. In this review, the laser-based spectroscopic techniques for formaldehyde gas detection were investigated, such as cavity ring-down spectroscopy (CRDS), cavity-enhanced absorption spectroscopy (CEAS), integrated cavity output spectroscopy (ICOS), tunable diode laser absorption spectroscopy (TDLAS), multi-pass cell absorption spectroscopy (MC), differential optical absorption spectroscopy (DOAS), non-dispersive absorption spectroscopy (NDAS), and photoacoustic spectroscopy (PAS). Among these techniques, the lowest detection limit achieved with an infrared laser source resulted in 28 ppt with a signal integration time of 40 s, and 210 ppt with an integration time of 30 s when using an ultraviolet light source
Frequency Matters: On the Impact of Carrier Frequency on Privacy in Radio Fingerprinting
Get PDFRadio Frequency Fingerprinting (RFF) relies on unique inherent imperfections in radios’ hardware to authenticate devices based on Radio Frequency emissions. In this work, we consider that fingerprints collected for multi-channel transmitters on certain frequencies get partially leaked to an adversary willing to track them, without information about the frequency used for training. In this scenario, we evaluate the performance of various state-of-the-art Convolutional Neural Networks for image-based RFF when the testing and training frequencies do not match. We demonstrate that RFF performances degrade significantly when training and testing frequencies differ, down to a random guess when they are sufficiently apart
Development of advanced technologies for enhancing obstacle-negotiation capabilities in mobile robots
Get PDFObstacle negotiation is one of the major challenges for mobile robots, particularly
in environments where they must operate on uneven terrain or encounter physical
barriers such as stairs, debris, or rough ground. Practical applications for such
robots range from humanitarian assistance to logistics and inspection of hazardous
or hard-to-reach environments. Although various technological approaches have
been developed and successfully implemented, there is still a need to enhance
robots’ ability to adapt to a wide range of complex obstacles, ensuring greater
operational efficiency and safety.
This thesis focuses on the analysis and development of innovative technologies
to improve robots’ ability to overcome different types of obstacles. The primary
goal is to enhance the robots’ capability to operate in challenging environments
and ensure smooth and safe movement even in the presence of physical barriers,
thus contributing to the advancement of mobile robotics.
The first part of the thesis presents a systematic review of the scientific and
engineering literature on stair-climbing mechanisms is given. It provides concise
descriptions of the mechanisms and operating methods, highlighting the advan
tages and disadvantages of various climbing platforms. To quantitatively assess
system performance, several metrics are introduced. Using these metrics, it be
comes possible to compare vehicles with different locomotion modes and charac
teristics, offering researchers and practitioners valuable insights into stair-climbing
vehicles and enabling them to select the most suitable platform for transporting
people and heavy loads up staircases.
The second part of the thesis aims to propose a rigorous analysis approach to
study what happens when different kind of rubber belts or tires are in contact
with a corner edge and what forces are exchanged between these two elements. A
general introduction is given by mainly focusing on the scientific literature lack
of a comprehensive wheel-obstacle contact model for the step-climbing problem.
Then the importance of considering tire deformation has been emphasised and a
novel approach to wheel-obstacle contact mechanics is given. A description of the
test bench specifically developed for this work is provided along the experimental
results for two cases of flat belt and tire patch.
The third part of the thesis presents experimental results on the behavior of a
conventional pneumatic tire clearing a step-obstacle, alongside an analytical model
developed to analyze the interaction between a deformable tire and the corner edge
of a step-obstacle.
Finally, the ”XXbot” concept is developed. The thesis proposes a specialized
model that predicts how the system will move based on the terrain profile. Stair
climbing simulations were then performed using multibody simulation software
MSC-Adams, and the results are presented to demonstrate the effectiveness of the
proposed vehicle. The findings indicate that the robot can be adapted for various
applications, such as stair-climbing wheelchair platforms
Analysis of Gamma-Ray Burst Closure Relationship in Multiple Wavelengths
Get PDFGamma-ray bursts (GRBs) are intense pulses of high-energy emission associated with the death of massive stars or compact objects' coalescence. Their multiwavelength observations help verify the reliability of the standard fireball model. We analyze 14 GRBs observed contemporaneously in gamma rays by the Fermi Large Area Telescope, in X-rays by the Swift Telescope, and in the optical bands by Swift and many ground-based telescopes. We study the correlation between the spectral and temporal indices using closure relations according to the synchrotron forward-shock model in a stratified medium (n proportional to r(-k)) with k ranging from 0 to 2.5. We find that the model without energy injection is preferred over the one with energy injection in all the investigated wavelengths. In gamma rays, we only explored the nu > max{nu(c), nu(m)} (slow cooling, SC/fast cooling, FC) cooling condition (where nu(c) and nu(m) are the cooling and characteristic frequencies, namely the frequencies at the spectral break). In the X-ray and optical bands, we explored all the cooling conditions, including nu(m) < nu < nu(c) (SC), nu(c) < nu < nu(m) (FC), and SC/FC, and found a clear preference for SC for X-rays and SC/FC for optical. Within these cooling conditions, X-rays exhibit the highest rate of occurrence for the density profile with k = 0, while the optical band has the highest occurrence for k = 2.5 when considering no energy injection. Although we can pinpoint a definite environment for some GRBs, we find degeneracies in other GRBs
Conversational User Interfaces and Agents
No full textConversational agents (CAs), such as chatbots or virtual assistants, represent Artificial Intelligence (AI) systems designed to facilitate human–machine interaction through natural language. These systems are revolutionizing communication by improving efficiency, effectiveness, and user-friendliness. CAs find applications across various domains, including customer service, health care, and education. This chapter thoroughly explores CAs, focusing on their underlying structure, the problems they aim to address, and the current challenges documented in the existing literature. Furthermore, we delve into the architectures of CAs, shedding light on the key distinctions between modular and end-to-end implementations. To achieve this objective, we introduce a comprehensive taxonomy categorizing the tasks CAs are designed to tackle, encompassing information retrieval, question answering, and chitchat. We scrutinize the advantages and disadvantages of potential models for each of these tasks, besides comprehensively investigating the recent developed methodologies and incorporating a detailed analysis of rule-based and data-driven strategies. Throughout this examination, we emphasize the strengths, limitations, and potential future directions, including the imperative need to develop ethical and reliable conversational agents
Advanced acoustic emission methods for damage mechanisms monitoring in aerospace materials
Get PDFThis thesis develops advanced Acoustic Emission (AE) techniques for monitoring and characterizing damage in aerospace materials. It focuses on AlSi10Mg produced by Selective Laser Melting (SLM) and on Carbon Fiber Reinforced Plastic (CFRP) composites. These materials are crucial for aerospace due to their unique mechanical properties. AlSi10Mg offers high strength and fatigue resistance. It is used in critical aerospace parts. CFRP composites provide stiffness strength and corrosion resistance, making them suitable for demanding aerospace environments. Monitoring damage in these materials is essential to ensure safety and structural integrity over time.
This research aims to improve damage monitoring by combining traditional AE methods with deep learning frameworks. Structural Health Monitoring (SHM) is vital in aerospace. It allows early detection of material degradation and reduces the risk of in-service damages. Traditional SHM methods can be limited in accuracy for complex materials such as AlSi10Mg and CFRP composites. This thesis introduces a robust approach that uses advanced methods to address these challenges.
Tensile tests were conducted on AlSi10Mg specimens built in different orientations. AE signals were recorded to examine their mechanical behavior. Continuous Wavelet Transform (CWT) was used to analyze these signals. This allowed differentiation between elastic and plastic deformation. Convolutional Neural Networks (CNNs) were then used to classify AE signals. Several CNN architectures, including AlexNet and SqueezeNet, were tested to improve classification accuracy. A novel approach was also introduced. It combines a Fuzzy Artificial Bee Colony (FABC) algorithm with CNN and CWT-scalogram analysis. This method includes data augmentation to improve robustness and prevent overfitting.
For CFRP composites, a Deep Autoencoder (DAE) framework was developed to automate damage mode characterization during mechanical testing. The DAE reduced the complexity of AE signals and extracted essential features. These features were clustered to identify damage modes like matrix cracking, delamination, and fiber breakage. By automating damage classification, the DAE enhances SHM by providing accurate real-time damage assessments.
This thesis shows that combining traditional AE features with deep learning models improves damage source classification for aerospace materials. These methods make SHM systems more efficient and precise. They offer advanced solutions for monitoring and maintaining structural integrity in aerospace. The research contributes to safer and more reliable aerospace applications
MORIX: Machine learning-aided framework for lethality detection and MORtality inference with eXplainable artificial intelligence in MAFLD subjects
Get PDFMetabolic dysfunction-associated fatty liver disease (MAFLD) introduces new diagnostic criteria for fatty liver disease that are independent of alcohol consumption and viral hepatitis infection. Therefore, investigating how biochemical and anthropometric factors influence mortality in MAFLD subjects is of significant interest. In this work, we propose MORIX, an Artificial Intelligence-based framework capable of predicting fatal mortality outcomes in subjects with MAFLD. MORIX utilizes data from epidemiological datasets containing carefully selected anthropometric and biochemical information. This selection is achieved through Recursive Feature Elimination (RFE) using a Random Forest (RF) to train Machine Learning (ML) algorithms and provide a mortality risk (Yes/No) output. To provide physicians with a valuable tool, MORIX was trained and tested on a dataset of MAFLD subjects, comparing five different models: Random Forest (RF), eXtreme Gradient Boosting (XGB), Support Vector Machine (SVM), Multilayer Perceptron (MLP), and Light Gradient Boosting Model (LGBM) in a 5-fold cross-validation training strategy. Experimental results identified the RF as the best model, achieving a high accuracy for both mortality risks predicted. Additionally, an eXplainable Artificial Intelligence (XAI) analysis was conducted to clarify the diagnostic logic of the RF model and to assess the impact of each feature to the prediction. Moreover, a web application was developed to predict mortality risk and provide explanations of how the input features influenced the final prediction. In conclusion, the MORIX framework is easy to apply, and the required parameters are readily available in healthcare datasets, making it a practical tool for medical professionals