Publications

Background In professional rugby, injury prevention and player availability are major challenges. Sports analytics use data from trainings and matches to address these issues. This study leveraged comprehensive daily data from a professional rugby club to predict players' readiness for training. Using this metric helped assess its effectiveness in predicting intrinsic injuries and improving injury prevention strategies. Methods Models including logistic regression, decision trees, and Long Short-Term Memory-based neural networks, were evaluated for their predictive accuracy and ability to discern patterns indicative of injury risks or readiness for physical activities. Findings The study demonstrated that long-short term memory and convolutional one-dimension models outperform traditional machine learning methods in analyzing players' physical conditions. This approach may support earlier identification of injury risks and inform workload management. Using model evaluation and interpretability techniques, including Local Interpretable Model-Agnostic Explanations (LIME) module, the study provided a framework for sports scientists, coaches, and medical staff to mitigate injury risks and optimize training sessions. Interpretation As a preliminary exploration, this study paves the way for further research into the integration of machine learning and neural networks in sports science, promising transformative impacts on injury prevention strategies in rugby.

Rugby is a sport with a high injury rate. Much has been done to make the sport safer, particularly in terms of limiting and identifying concussions. Recently, instrumented mouthguards have been developed and used to measure events that may lead to concussion. However, these instrumented mouthguards may not have an appropriate geometry regarding shock absorption and comfort. In addition, there is no specific international standard for instrumented mouthguards. This study proposed a geometric analysis of both instrumented and non-instrumented mouthguards. Ten instrumented mouthguards were analysed and compared with three non-instrumented mouthguards. They were inspected visually, with a 3D envelope scan and with a CT scan. The results showed that the mouthguards did not comply with recommendations such as indentation with the lower teeth which may increase injury or fracture risk.

Terahertz (THz) imaging is gaining attention as a nondestructive testing technique for assessing damage due to its high axial resolution and nonionizing nature, presenting a promising alternative to conventional methods such as ultrasound and X-ray imaging. Its practical implementation, however, remains limited by the reliance on expert interpretation and the frequent need for validation using supplementary techniques such as X-ray microcomputed tomography (µCT), particularly for complex damage modes. This study focuses on woven-fabric-reinforced thermoplastic composites subjected to low-velocity impact, which typically causes barely visible impact damage (BVID). The damage is subtle yet critical, potentially leading to failure under subsequent loading. The multilayered and spatially distributed characteristics of BVID make it especially challenging to identify. To overcome these challenges, this work integrates deep learning with pulsed THz time-of-flight tomography (TOFT) imaging to enable automated damage detection in composite laminates. In contrast to existing research that mainly targets delamination using A- or C-scan data, this study emphasizes the detection of low-velocity impact damage by leveraging THz B-scans, which offer nondestructive depth-resolved cross-sectional imaging. The training dataset is labeled by correlating THz TOFT scans with X-ray CT images used as ground truth. A transfer learning approach, based on convolutional neural network (CNN) architectures, is employed for binary classification to distinguish damaged from undamaged regions. The resulting classifier achieves over 95 % accuracy, demonstrating the viability of this method for industrial applications such as quality assurance and in-service inspection of composite structures.

Advancements in artificial intelligence (AI) have ushered in the era of the fourth industrial revolution, transforming workplace dynamics with AI's enhanced decision-making capabilities. While AI has been shown to reduce worker mental workload, improve performance, and enhance physical safety, it also has the potential to negatively impact psychosocial factors, such as work meaningfulness, worker autonomy, and motivation, among others. These factors are crucial as they impact employee retention, well-being, and organizational performance. Yet, the impact of automating decision-making aspects of work on the psychosocial dimension of human-AI interaction remains largely unknown due to the lack of empirical evidence. To address this gap, our study conducted an experiment with 102 participants in a laboratory designed to replicate a manufacturing line. We manipulated the level of AI decision support—characterized by the AI's decision-making control—to observe its effects on worker psychosocial factors through a blend of perceptual, physiological, and observational measures. Our aim was to discern the differential impacts of fully versus partially automated AI decision support on workers' perceptions of job meaningfulness, autonomy, competence, motivation, engagement, and performance on an error-detection task. The results of this study suggest the presence of a critical boundary in automation for psychosocial factors, demonstrating that while some automation of decision selection can nurture work meaningfulness, worker autonomy, competence, self-determined motivation, and engagement, there is a pivotal point beyond which these benefits can decline. Thus, balancing AI assistance with human control is vital to protect psychosocial well‑being. Practically, industry and operations managers should keep employees involved in decision making by adopting partial, confirm‑or‑override AI systems that sustain motivation and engagement, boosting retention and productivity.

Augmented Reality (AR), a technology that overlays digital content onto the physical environment, holds promise for enhancing creativity and usability in product design education. However, despite the advantages of Additive Manufacturing (AM) in enabling complex and customizable designs, designers often struggle to grasp its abstract principles. Grounded in theories of immersive learning and multimodal visualization, this study investigates whether integrating AR visualization can facilitate better understanding and stimulate creativity in AM education. A controlled experiment was conducted with 34 master's students in product design, randomly assigned to either an AR-based learning group or a traditional card-based learning group. Participants engaged with AM principles through either an interactive AR application featuring manipulable 3D cube models or static information cards. Usability perceptions and creativity of design outputs were assessed respectively through structured questionnaires and expert evaluations by five domain specialists. Mann–Whitney U tests, appropriate for non-normally distributed data, revealed that the AR group reported significantly higher usability ratings and produced more original design outcomes compared to the card-based group. These findings demonstrate that AR-based educational tools can directly improve the usability and creative engagement of students in learning AM principles. This study contributes to advancing the understanding of how immersive technologies can be effectively integrated into design education to foster both practical skills and innovative thinking.

Previous work has shown that telecollaboration is a suitable solution for remote assistance of industrial maintenance operations, provided that an audio chat solution is available. There are several reasons why audio chat may not be available: the quality of the available Internet network, both in terms of bandwidth and stability (jittering), but also the presence of too much noise at the site of the operation, which interferes with voice capture. This paper presents a methodology to evaluate the quality provided by an audio chat solution. This methodology is then tested on a specific audio chat solution built on a lossy compression algorithm based on the grouping of successive similar values to overcome the jittering problem and significantly reduce bandwidth requirements. We suggest evaluating the audio quality by assessing the intelligibility of different audio recordings using standard speech therapy methods. Our results suggest that an audio chat can be provided even in a low bandwidth scenario and in a noisy environment, which provides promising insights for the further development of telecollaboration. Moreover, the assessment of audio quality using restitution exercices to evaluate intelligibility, tested on a real use case gives interesting results on the usability of an audio chat solution as well as detailed feedbacks on which part of the altered signal is to be improved.

In modern manufacturing, accurately predicting cutting forces is essential for the design and control of machining operations. Common mechanistic models of cutting forces rely on a precise description of the local uncut chip area. However, in milling, the specific trajectories of cutting edges create challenges in modelling this quantity. Existing analytical models are typically limited to 2D contexts or assume circular tooth trajectories, which are mostly valid for cylindrical end mills. These assumptions limit their applicability to high-feed milling, especially due to low lead angles and complex insert cutter geometries producing non-circular paths. This article presents a new three-dimensional analytical model for evaluating the local uncut chip thickness in high-feed milling. It relies on closed-form expressions derived from geometric analysis and Taylor expansions to approximate the uncut chip area and cutter-workpiece engagement, even in regions where conventional models fail. The model applies to linear-path milling and accounts for tool run-out and differential pitch. Compared to a Newton-Raphson numerical method, it achieves a relative error below 5% while being 3 to 9 times faster, enabling efficient integration in force models. Beyond its computational efficiency, the explicit formulation enables analysis of geometric influence, such as sensitivity to feed per tooth or tooth count-capabilities not easily accessible with purely numerical approaches. This work contributes a rigorous and interpretable alternative for improving cutting force prediction in high-feed milling.

Bibliographic data about dry mass loss coming from different thermal and hygrothermal modification processes versus time were collected in this work and analyzed together. The data sets were collected from 2 experimental campaigns involving different modification technologies: (A) spruce wood hygro-thermally modified under superheated steam conditions performed at relative humidity ranging from 35 to 92% at temperatures ranging from 110 to 170 °C; and (B) poplar wood samples thermally modified with Thermo-vacuum® technology performed at temperatures ranging from 150 to 240 °C. For both processes, conversation rates master curves at 150 °C were identified on the experimental points using the time-temperature and the time-temperature-humidity superposition method. The two master curves were then compared and a generalized kinetic model able to predict the mass losses when modifying wood at different temperatures and relative humidity implemented. A kinetic model is expressed with a master curve composed of 2 kinetic stages. That model can predict the mass variation occurring during any hygrothermal modification whatever the temperature applied and the environment medium relative humidity.

In this paper, the wave finite element(WFE) method is briefly presented and applied in order to extract the dispersion curves. The formulation of the laminated structure is detailed through the Timoshenko theory. The finite element technique is used to model the laminated beam and extract the mass and stiffness matrices for the bending vibration. The bending vibration of the laminated beam is simulated and discussed. The travelling and evanescent modes are illustrated to characterize the flexural wave propagation in laminated structure. The resolution of the equilibrium equation leads to the extraction of the analytical wave number as a function of the frequency in order to validate the dispersion curves simulated through the WFE method. The question of the influence of the layers thickness on the wave propagation is detailed. An uncertainty is introduced in the thickness as a Gaussian variable and the mean and the standard deviation of the dispersion curves are extracted through the Monte Carlo simulation. Among the contributions of this article, the laminated structures are modeled through the Abaqus software and the mass and stiffness matrices are extracted for the multimodal propagation. The multimodal wave number is presented and discussed for the travelling and evanescent modes.

With the aim of improving the lifespan of different steel tools by reducing their degradation, CrAlN nitride coatings were investigated. The CrAlN coatings were deposited on X38CrMoV5 steel substrates using the DC reactive magnetron sputtering process under various deposition temperatures between ambient temperature and 300 ◦C. The effects of deposition temperature were systematically explored: XRD, EDS, SEM, optical profilometer, contact angles, nanoindentation, and tribometry were carried out to establish the structural, mechanotribological, and wetting properties’ relationship. Results show that the high deposition temperature promotes the growth of (200) CrN preferentially orientation with the appearance of AlN phase. As the deposition temperature increases, the contact angle of the CrAlN surface films changes to a higher hydropholicity and the hardness of the coatings gradually increases to reach a maximum value of 28.1 GPa. The main wear mechanisms of CrAlN coating deposited at 300 ◦C against Al6061 ball are a combination of abrasive and adhesive features. This coating also has the lowest friction coefficient (0.63) and wear rate (1.72 × 10�� 3 mm3/N.m). Indeed, the preferred deposition temperature of about 300◦C could effectively adjust the microstructure and improve the mechanical and tribological properties of CrAlN coatings, thereby indicating its potential as an effective coating material for the protection of X38CrMoV5 steel in industrial fields.