Politecnico di Torino

In safety-critical applications, microcontrollers have to be tested to satisfy strict quality and performances constraints. It has been demonstrated that on-chip ring oscillators can be used as speed monitors to reliably predict the performances. However, any machine-learning model is likely to be inaccurate if trained on an inadequate dataset, and labeling data for training is quite a costly process. In this paper, we present a methodology based on active learning to select the best samples to be included in the training set, significantly reducing the time and cost required. Moreover, since different speed measurements are available, we designed a multi-label technique to take advantage of their correlations. Experimental results demonstrate that the approach halves the training-set size, with respect to a random-labelling, while it increases the predictive accuracy, with respect to standard single-label machine-learning models.

p>Failures of structures occur in all parts of the world as the result of design errors, construction defects, abuse or misuse, ageing and deterioration of the structure, lack of maintenance, as well as environmental effects such as wind, flood, snow, earthquake and, of course, human errors. They can result in catastrophic human costs as well as heavy financial losses to all involved, including local economic growth deceleration, expensive delays and repairs, as well as other repercussions, such as legal actions to responsible parties. ‘Welcome’ effects of these unfortunate events include a better understanding of the origins and causes of structural failures, their corresponding lessons learnt, and a more effective mitigation of their occurrence through changes in codes, standards, guidelines, and practice. In several countries the investigation process of the causes of failures, responsibilities, and resolution of the consequent claims have created an active, demanding, and specialised field of professional practice – often referred to as Forensic Structural Engineering – with well-defined technical and legal procedures. This bulletin is the result of the work lead by the Task Group 5.1 ‘Forensic Structural Engineering’. It provides understanding of the origins, causes, and consequences of failures, their forensic investigations, and the lessons learnt from them. The aim of the bulletin is not only to describe different examples but, mainly, to use emblematic case studies to show procedures that can be used when dealing with structural failures. In addition to obtaining a deeper insight into the technical causes for structural failure, the reader would be duly informed about the different countries’ legal issues related to the investigation process. The bulletin is aimed at young, mid-career and experienced structural engineers who want to acquire a better understanding of failure mechanisms towards improving their design, inspection, construction, administrative, and other project-related practices to avoid pitfalls that may lead to failures. It also aims at those wanting to acquire a working knowledge of the challenging professional practice of forensic structural engineering.</p

In an increasingly complex world, the design disciplines often remain tied to rigid modes of practice. Among these is the growing weakness of the collaboration between academic and professional environments, which instead the scientific literature indicates as a potential ground for innovation due to their complementary nature and the different types of expertise they nurture (Amin and Roberts in Res Policy 37:353–369, 2008). How could we reframe and rethink architectural design to strengthen the relationship between academia and professional practice? How could this interaction innovate the practice and the architecture discipline in a global perspective—in the pursuit of the values expressed by SDGs 8 (“decent work and economic growth”) and 11 (“sustainable cities and communities”)? In specific geographical and regulatory contexts, professional and academic institutions find wide spaces for collaboration (University Design Institutes in China, PennPraxis at the University of Pennsylvania, etc.). Against this background, the project Polito Studio aims to reframe existing institutional structures and define a platform for collaboration-in-practice projects between the academic and professional spheres. Such a platform is constructed through successive forms of “putting into practice” (Barbera and Parisi, Innovatori sociali, Il Mulino, Bologna, 2019) that allow to define, in the running, objectives, tools, and actions. The project intends to create a horizontal platform where a newly formed group of practitioners work together at a transnational scale, “perforating” international markets most commonly dominated by large global firms, through professional expertise combined with scientific research, and modes of labour and care that are more easily found at the small scale of practice.

The paper proposes a novel approach to enhance the resilience of mutual collaborative activity between humans and robots in industrial assembly tasks. The approach exploits Adversarial Reinforcement Learning (ARL) to enable a robot to learn an assembly policy that is robust against human mistakes. The adversary can represent various sources of uncertainty or disturbance in the environment. By learning from adversarial feedback, the agent can improve its performance and adaptability in challenging scenarios. The paper applies ARL to the execution of the assembly task sequence. The robot acts as one agent and learns how to assist the human partner during the assembly. The agent simulating the human partner acts as the adversary and deliberately introduces mistakes during the assembly process. The robot also learns how to cope with different levels of human competence and cooperation by adjusting its own behaviour accordingly. The paper evaluates the proposed approach through experiments reproducing complex assembly sequences and compares it with baseline methods that use conventional optimization algorithms. The results show that ARL does not outperforms conventional optimization algorithms in terms of task completion time but guarantee robustness against human mistakes. The paper also discusses the implications for human-robot collaboration and suggests future directions for research.

Vanishing viscosity approximations are considered here for discontinuous sweeping processes with non-convex constraints. It is shown that they are well-posed for sufficiently small viscosity parameters, and that their solutions converge pointwise, as the viscosity parameter tends to zero, to the left-continuous solution to the sweeping process in the Kurzweil integral setting. The convergence is uniform if the input is continuous.

In this paper, multiscale modeling of polymer composites consisting of Graphene Foam (GF) and Polydimethylsiloxane (PDMS) is conducted, and their Thermal Conductivity (TC) is investigated through the use of nano-to-microscale analyses. The TC of the PDMS matrix and GF is calculated using the Molecular Dynamics (MD) method. The effective properties of the composites are computed utilizing the Mechanics of Structure Genome (MSG) coupled with Carrera Unified Formulation (CUF) as a novel micromechanical method, which allows an accurate description of the problem resulting in a high-fidelity analysis. Due to the unique interconnected structure of GF, the TC of GF/PDMS composite reaches 0.406 Wm-1 K-1 for GF with 63% porosity, which is about 69% ± 2% higher than that of neat PDMS.

Mechanical tests on bone plates are mandatory for regulatory purposes and, typically, the ASTM F382 standard is used, which involves a four-point bending test setup to evaluate the cyclic bending fatigue performance of the bone plate. These test campaigns require a considerable financial outlay and long execution times; therefore, an accurate prediction of experimental outcomes can reduce test runtime with beneficial cost cuts for manufacturers. Hence, an analytical framework is here proposed for the direct estimation of the maximum bending moment of a bone plate under fatigue loading, to guide the identification of the runout load for regulatory testing. Eleven bone plates awaiting certification were subjected to a comprehensive testing campaign following ASTM F382 protocols to evaluate their static and fatigue bending properties. An analytical prediction of the maximum bending moment was subsequently implemented based on ultimate strength and plate geometry. The experimental loads obtained from fatigue testing were then used to verify the prediction accuracy of the analytical approach. Results showed promising predictive ability, with R ² coefficients above 0.95 in the runout condition, with potential impact in reducing the experimental tests needed for the CE marking of bone plates.

The paper presents the results of an experimental investigation assessing the influence of the openings typology and position on the cyclic response of masonry-infilled reinforced concrete frames. The experimental program consisted of seven 2/3 scale square frame specimens infilled with hollow clay bricks, which were subjected to quasi-static lateral cyclic tests up to large drifts. A reference bare frame was also tested for comparison. The specimens included solid infill walls, infills with door openings, and infills with window central and eccentric openings. The experimental responses of the specimens were analyzed in terms of strength, stiffness, energy dissipation capacity, and equivalent damping, and were compared to those of the reference bare frame and fully infilled frame. Results revealed a significant modification of the resisting mechanisms as a function of the typology and position of the openings with respect to the case of a fully infilled frame, although the lateral resisting capacity was not substantially modified. On the other hand, it was observed that the achievement of the limit state thresholds occurred at substantially different interstorey drifts, indicating different damage metrics. Simple regression analyses, based on the experimental results of this study and from the literature, were finally conducted to interpret the modification of the peak force and of the corresponding drift as a function of synthetic parameters defining the opening arrangement.

Inclusive writing is compulsory in formal communications. However, employees in private organizations, universities, and ministries often lack inclusive writing skills. For example, despite Italian grammar having masculine and feminine declensions of words, many official documents have a disrespectful prevalence of the masculine form. To promote inclusive writing practices, we present Inclusively, a language support tool that leverages natural language processing techniques to automatically identify instances of non-inclusive language and suggest more inclusive alternatives. The tool can be used as a text proofreader and, at the same time, fosters self-learning of inclusive writing forms. The recorded demo of the tool, available at https://youtu.be/3uiW_ti8wmY, shows how end-users can interact with Inclusively to feed new data, visualize the non-inclusive pieces of text, explore the list of alternative forms, and provide feedback or human annotations for system fine-tuning.

The deployment of Deep Learning (DL) models is still precluded in those contexts where the amount of supervised data is limited. To answer this issue, active learning strategies aim at minimizing the amount of labelled data required to train a DL model. Most active strategies are based on uncertain sample selection, and even often restricted to samples lying close to the decision boundary. These techniques are theoretically sound, but an understanding of the selected samples based on their content is not straightforward, further driving non-experts to consider DL as a black-box. For the first time, here we propose to take into consideration common domain-knowledge and enable non-expert users to train a model with fewer samples. In our Knowledge-driven Active Learning (KAL) framework, rule-based knowledge is converted into logic constraints and their violation is checked as a natural guide for sample selection. We show that even simple relationships among data and output classes offer a way to spot predictions for which the model need supervision. We empirically show that KAL (i) outperforms many active learning strategies, particularly in those contexts where domain knowledge is rich, (ii) it discovers data distribution lying far from the initial training data, (iii) it ensures domain experts that the provided knowledge is acquired by the model, (iv) it is suitable for regression and object recognition tasks unlike uncertainty-based strategies, and (v) its computational demand is low.

Binary neural networks (BNNs) are an attractive solution for developing and deploying deep neural network (DNN)-based applications in resource constrained devices. Despite their success, BNNs still suffer from a fixed and limited compression factor that may be explained by the fact that existing pruning methods for full-precision DNNs cannot be directly applied to BNNs. In fact, weight pruning of BNNs leads to performance degradation, which suggests that the standard binarization domain of BNNs is not well adapted for the task. This work proposes a novel more general binary domain that extends the standard binary one that is more robust to pruning techniques, thus guaranteeing improved compression and avoiding severe performance losses. We demonstrate a closed-form solution for quantizing the weights of a full-precision network into the proposed binary domain. Finally, we show the flexibility of our method, which can be combined with other pruning strategies. Experiments over CIFAR-10 and CIFAR-100 demonstrate that the novel approach is able to generate efficient sparse networks with reduced memory usage and run-time latency, while maintaining performance.

This paper presents a task allocation strategy for a multi-robot system with a human supervisor. The multi-robot system consists of a team of heterogeneous robots with different capabilities that operate in a dynamic scenario that can change in the robots’ capabilities or in the operational requirements. The human supervisor can intervene in the operation scenario by approving the final plan before its execution or forcing a robot to execute a specific task. The proposed task allocation strategy leverages an auction-based method in combination with a sampling-based multi-goal motion planning. The latter is used to evaluate the costs of execution of tasks based on realistic features of paths. The proposed architecture enables the allocation of tasks accounting for priorities and precedence constraints, as well as the quick re-allocation of tasks after a dynamic perturbation occurs –a crucial feature when the human supervisor preempts the outcome of the algorithm and makes manual adjustments. An extensive simulation campaign in a rescue scenario validates our approach in dynamic scenarios comprising a sensor failure of a robot, a total failure of a robot, and a human-driven re-allocation. We highlight the benefits of the proposed multi-goal strategy by comparing it with single-goal motion planning strategies at the state of the art. Finally, we provide evidence for the system efficiency by demonstrating the powerful synergistic combination of the auction-based allocation and the multi-goal motion planning approach.

The article recalls the history of Geografia Democratica, a collective of scholars that during the second half of the 1970s sought to dismantle the old deterministic approach and promote a critical and radical turn in Italian academic geography. The aim is to contribute to the ongoing debate upon ‘other geographical traditions’ beyond the Anglo-American hegemony, to highlight the pluriversal roots of contemporary critical geographies and the influence that the transnational circulation of knowledge had in their unfolding, in light of recent quests for a more global geographical imagination. To do so, the article first engages with Geografia Democratica as a ‘rupture experience’ in the mainstream of Italian geography, and then discusses how it intersected or not similar turns occurred elsewhere, focusing on the mostly implicit dialogue between Italian and Anglo-American critical/radical geographies of the time. By following the controversial story of the collective during and after its short existence, we question its legacy for today’s geographical scholarship, and reflect more generally upon the significance of reviving other critical and radical traditions. To highlight the plurality of our disciplinary past, we suggest, is crucial not only to fill the ‘asymmetric ignorance’ between various traditions, but also to nurture and reposition the present ‘worlding’ practices of non-Anglophone critical geographers.

The present study relates to comparison between different approaches for definition of global safety factors for non-linear analysis of slender RC members with reference to new or existing structures. Firstly, a benchmark set of 40 experimental results on reinforced concrete columns is presented. After the description of the main features of the benchmark test sets the related non-linear numerical models have been realized using fiber-modelling as solution strategy. Then, appropriate assumptions concerning aleatoric and epistemic uncertainties have been performed with the aim to run probabilistic analysis of global resistance for each one of the 40 columns. The results of the probabilistic analysis are useful to define global safety factors in line to the global resistance method. Finally, the comparison between different approaches to derive global safety factors is presented and discussed.

This paper discusses objectives, methodologies and results from a campaign of non-destructive investigations on a cable-stayed bridge in Lisbon (Portugal). Among the investigations held, the campaign encompassed dynamic testing of the stay cables in order to identify their natural vibration frequencies and to estimate their tension force. A frequency-domain output-only approach was carried out to identify the natural frequencies of the cables. Subsequently, the cable tension force was estimated on the basis of the flat taut string model. In an advantageous and cost-effective way, vibration tests were carried out without the interruption of the vehicular traffic. Results of the testing campaign were finally incorporated in the bridge condition assessment and used to provide both short- and long-term maintenance prescriptions.

Among the new technologies driving the fourth industrial revolution in the construction industry, 3D Concrete Printing (3DCP) is playing a key role. The typical process is made through robotic arms or gantries equipped with nozzles, similarly to contour crafting in other industries. Despite 3DCP is appealing when applied to complex architectural shapes, the structural behaviour and geometrical size are limitations difficult to overcome. Upscaling the extrusion process to full sections, introducing a new concept of ultrafast adaptable slip-forming, is the access key to different sectors of the construction industry, as infrastructures. This paper will present the Extruded Tunnel Lining Regeneration (ETLR) technology developed by HINFRA with the scope to automatically regenerate the lining of existing damaged tunnels. “Tailored” features and issues of the aforesaid technology will be discussed in this paper, together with a design validation related to a Fibre Reinforced Concrete (FRC) tunnel lining.

In recent years it has become clear that many western countries infrastructures require scrupulous and continuous monitoring. The main purpose of Structural Health Monitoring (SHM) is the evaluation of the safety of existing structures and the guide for maintenance interventions, where necessary. One of the techniques used in structural monitoring is the measurement of angles using clinometers Nevertheless, many issues on the reliability and the correct use of measures done with clinometers have to be addressed to achieve a trustworthy SHM. In this paper the most relevant issues related to the f.e.m. modelling of a prestressed concrete bridge deck related to the use of clinometers for SHM are presented. The study presents a test-case deck that has been under continuous monitoring for many months. A brief explanation of the data-cleaning process and the interpretation of the clinometers outputs is also given, stressing out the limitations of the technology and possible outcomes.

The fatigue response of additively manufactured (AM) specimens is mainly driven by manufacturing defects, like pores and lack of fusion defects, which are mainly responsible for the large variability of fatigue data in the S–N plot. The analysis of the results of AM tests can be therefore complex: for example, the influence of a specific factor, e.g. the building direction, can be concealed by the experimental variability. Accordingly, appropriate statistical methodologies should be employed to safely and properly analyze the results of fatigue tests on AM specimens. In the present paper, a statistical methodology for the analysis of the AM fatigue test results is proposed. The approach is based on shifting the experimental failures to a reference number of cycles starting from the estimated P–S–N curves. The experimental variability of the fatigue strength at the reference number of cycles is also considered by estimating the profile likelihood function. This methodology has been validated with literature datasets and has proven its effectiveness in dealing with the experimental scatter typical of AM fatigue test results.