Higher Institute on Territorial Systems for Innovation
PORTO@iris (Publications Open Repository TOrino - Politecnico di Torino)Not a member yet
146173 research outputs found
Sort by
Enhancing Software Maintainability through LLM-Assisted Code Refactoring
No full textHigh code quality, particularly in terms of maintainability, is crucial for ensuring that software remains efficient and adaptable over time, while minimizing long-term maintenance costs.
As artificial intelligence continues to evolve, its application in software development offers new opportunities to improve code quality. This study investigates the use of Large Language Models (LLMs) to enhance software maintainability through code refactoring.
The results indicate that LLMs can be effectively utilized for this purpose, with effectiveness varying depending on the model and the evaluation metric used. Although the study is based on a limited set of Python projects and specific prompting strategies, it provides a meaningful step toward understanding the broader applicability of LLMs in this context
Analysis of the work-to-heat conversion beyond the necking onset in non-isothermal tensile tests
Get PDFWhen metals undergo plastic deformation at sufficiently high speed, the plastic work dissipated as heat can cause material self-heating, influencing the material's mechanical behavior. This phenomenon is particularly critical during localized deformation. The present study proposes two methods for investigating the work-to-heat conversion during the post-necking phase of tensile tests on cylindrical dog-bone samples. One method directly applies the heat equation and is developed for adiabatic conditions only; the other method requires iterative thermal-structural finite element simulations but is also applicable to non-adiabatic conditions. Both methods involve recording the tests with optical and infrared cameras with adequate temporal and spatial resolution to fully exploit the heterogeneous fields and to improve the reliability of the results. A key distinction from other studies is the avoidance of using Digital Image Correlation technique, opting instead for a method based on necking silhouettes, which reduces experimental complexity and avoids issues related to speckle damage. Both the direct and iterative approaches, after being successfully tested against numerical benchmarks, were applied to an experimental case study. The material tested was 17–4PH martensitic stainless steel, known for its moderate strain rate sensitivity and substantial self-heating due to low thermal conductivity. Tests were conducted at room temperature and at nominal strain rates of 1, 10, and 1000 s−1. The results demonstrated the practical applicability of the proposed methods, suggesting their potential for investigating the work-to-heat conversion up to large strains by exploiting the post-necking phase of tensile tests with limited experimental complexity
Divergence-Aware Training with Automatic Subgroup Mitigation for Breast Tumor Segmentation
No full textDeep learning models for breast tumor segmentation in DCE-MRI may exhibit disparities in performance across demographic and clinical subgroups, raising concerns about fairness and clinical trustworthiness. In this work, we propose a subgroup-aware in-processing mitigation strategy that integrates divergence-based regularization directly into the training loop. By leveraging interpretable metadata (e.g., age, menopausal status, breast density), we identify subgroups where the model underperforms and assign higher loss weights to these samples in proportion to their divergence from average performance. Our method enables the model to focus training on underrepresented or harder-to-segment subpopulations without requiring external data or post-processing correction. We evaluate our approach on the MAMA-MIA 2025 challenge dataset, demonstrating improvements in both overall segmentation quality and fairness score. Our results highlight the potential of in-processing mitigation as an effective and practical pathway toward equitable medical image segmentation
Enabling Integrity Measurement for Secure Applications in the Enarx Framework
Get PDFThe Cloud Computing paradigm has significantly spread thanks to the high-speed Internet connection, standardization of digital technology, and the wide adoption of mobile devices. As a result, several privacy-enhancing technologies have been developed, among which Confidential Computing aims to protect data in use. Among the various solutions proposed for Confidential Computing, the Trusted Execution Environments (TEE) is becoming increasingly adopted, even in industrial scenarios, providing a shielded area where data and code can be processed and stored. However, heterogeneous TEE technologies are now available, making trusted application development difficult for developers. To overcome the problem of developing and deploying applications caused by the deep differences between the currently available TEE technologies, the project Enarx has been proposed. Enarx permits application development for various TEE instances in the public cloud, being CPU-architecture independent and guaranteeing the security of applications from cloud providers. The Enarx logic loads an application attesting the hardware and the Enarx components but misses the integrity verification of the user-developed application. Therefore, the primary objective of our work is to propose an extension where Enarx can verify the user application's trustworthiness deployed in underneath the TEE. The next objective is to integrate the extended Enarx framework with the Trust Monitor system, a centralized monitoring and reporting solution to assess the trustworthiness of a heterogeneous critical infrastructure, like the cloud environment. A validation phase has been conducted, proving the solution fulfils the defined goals in terms of functionalities and performance
A vehicle dynamics-oriented estimator for soft soil/tyre contact parameters from experimental testing
No full textModelling the interaction between tyres and unconsolidated soft surfaces has assumed a crucial role in predicting off-road vehicle performance in different machine areas such as planetary exploration and agriculture. The direct measurement of the soft soil/tyre contact parameters is a challenging task, addressed by expensive experimental campaigns and specific tools such as sensor-equipped wheels. In this paper an alternative cost-effective approach is proposed to estimate the contact parameters for semi-empirical formulations. The method relies on the experimental measurement typically available on the CAN bus of passenger vehicles. Specifically, the algorithm is tested with data gathered during acceleration manoeuvres performed on two different soft surfaces, i.e., snow and sand. The experimental signals are used to feed a 5 Degree Of Freedom (DOF) virtual vehicle equipped with a custom semi-empirical soil contact model. An optimisation problem with the target of minimising the differences between experimental and numerical traction performance is designed for the estimation of the sinkage module, cohesion, friction angle, elastic recovery and the multi-pass factor. Finally, the estimated parameters are validated using different experimental signals and data from literature, demonstrating the robustness of the methodology
Co-Design as an Enabling Method in Systemic Design Methodology for Circular and Sustainable Ecosystem
No full textThis paper explores the potential role of co-design as an enabling method in systemic methodology to promote circular ecosystems within the circular economy. By analysing two ongoing projects, we reflect on the integration of co-design into systemic design to develop sustainable solutions that effectively address contemporary environmental challenges. Co-design actively involves various stakeholders in the design process, facilitating relationships and creating resilient systems that optimize resource use and reduce waste. The systemic methodology, with its holistic approach, enables the analysis and understanding of the interconnections between different elements of an ecosystem, promoting synergies between economy, society, and environment. The first project focuses on using co-design to connect the diverse professional backgrounds within the working group. This approach aims to develop a training program tailored to a varied user base with different backgrounds and needs. By involving stakeholders such as universities of design, health professions and management in the co-design process, the project ensures that the training materials are relevant, comprehensive, and accessible to all participants which arrive from different academic backgrounds. The second project utilizes co-design to facilitate industrial symbiosis and promote circular economy practices. Through co-design, different industries collaborate to identify opportunities for resource sharing and waste minimization. By engaging stakeholders from various sectors, the project creates a network of businesses that can exchange materials, energy, and by-products, thus closing the loop and reducing environmental impact. This industrial symbiosis exemplifies how co-design can drive circular economy initiatives by fostering collaboration and innovation across industries
Accurate and Efficient Prediction of Wi-Fi Link Quality Based on Machine Learning
Get PDFWireless communications are characterized by their unpredictability, posing challenges for maintaining consistent communication quality. This article presents a comprehensive analysis of various prediction models, with a focus on achieving accurate and efficient Wi-Fi link quality forecasts using machine learning techniques. Specifically, the article evaluates the performance of data-driven models based on the linear combination of exponential moving averages, which are designed for low-complexity implementations and are then suitable for hardware platforms with limited processing resources. Accuracy of the proposed approaches was assessed using experimental data from a real-world Wi-Fi testbed, considering both channel-dependent and channel-independent training data. Remarkably, channel-independent models, which allow for generalized training by equipment manufacturers, demonstrated competitive performance. Overall, this study provides insights into the practical deployment of machine learning-based prediction models for enhancing Wi-Fi dependability in industrial environments
Strain pre-extrapolation methods for shape sensing: A comparative study between modal virtual sensor expansion and smoothing element analysis
Get PDFReconstructing displacement fields from sparse strain measurements, commonly referred to as shape sensing, has become a key component in developing effective Structural Health Monitoring (SHM) systems and for enabling accurate digital twin representations of engineering structures. Among available techniques, the inverse Finite Element Method (iFEM) is widely used but typically requires a dense sensor network. To reduce this dependency, strain pre-extrapolation methods are employed. The most established approach is Smoothing Element Analysis (SEA), which performs well for simple geometries but struggles with complex built-up structures. A recently proposed alternative, the Modal Virtual Sensor Expansion (Modal VSE), leverages modal strain shapes to virtually expand the strain field and has shown promising results, though it has not yet been benchmarked against existing methods. This study provides the first direct comparison between Modal VSE and SEA for strain pre-extrapolation and subsequent iFEM-based shape sensing of a composite stiffened panel. Results demonstrate that Modal VSE achieves higher accuracy and better adaptability across the examined configurations. Its superior performance persists even when sensor signals are corrupted by noise representative of experimental conditions. These findings highlight Modal VSE as a robust and effective tool for enhancing shape sensing in complex structural domains, thereby supporting more practical implementations of iFEM-based SHM and digital twin frameworks
Processing-driven structuring of melt compounded polymer-based systems
Get PDFL'abstract è presente nell'allegato / the abstract is in the attachmen
Multibody Modeling of an Electric Kick Scooter for Vertical Dynamics
No full textThe growing emphasis on environmental sustainability and urban mobility has driven a global rise in the adoption of electric micro-vehicles. Among these, electric kick scooters have emerged as a leading choice, owing to their ease of integration into existing public transportation networks and widespread availability through shared mobility services. As a new transport mode, the scientific community has recently focused on this vehicle category to address regulations and enhance safety and comfort aspects. This study presents the development and experimental validation of a multibody model of an e-scooter and its rider, specifically tailored for ride comfort analysis. The model, implemented in MATLAB/Simscape Multibody, includes a detailed representation of a commercial electric scooter and integrates a custom tire–road contact subsystem, suitable for vertical dynamics. Special emphasis is also placed on the rider model, where a torque-based control strategy of the joints is introduced to replicate the human body’s biomechanical response to external inputs. Experimental validation is carried out through on-road tests, consisting of riding at constant speed over a bike path with speed bumps while recording acceleration data. The results highlight the significant impact of the rider on the vertical dynamics of the scooter and demonstrate the importance of incorporating a detailed tire model to accurately capture the interaction with road irregularities