Publications

The contact between the tool and the workpiece/chip in metal cutting is complex, resulting in high local temperatures and stresses, which may cause severe tool wear and failure. Developments in cryogenic-assisted machining have shown an ecological alternative to the classical metal working fluids, besides tool wear reduction during machining difficult-to-cut materials due to the good ability to dissipate the heat generated by this process. The objective of this work is to analyze the tribological conditions and performance of new coatings specially developed for cryogenic-assisted machining in terms of friction coefficient, volume of build-up material (adhesion) to the tool, and tool temperature. The results have shown that the sliding speed and cooling/lubrication strategy are two main factors that affect the friction coefficient and adhesion of Ti–6Al–4V alloy to the pins. These tribological tests should allow us to select the best coating(s) to be used in cutting tools for further tool wear analysis. Moreover, the obtained friction coefficients could be further implemented into metal cutting models to predict the machining outcomes, including the surface integrity of the machined parts and tool wear.

Numerical simulation of the foundry process, aimed at reducing costs and production lead times, directly impacts the responsiveness and competitiveness of the foundries that employ it. Three experiments were conducted to establish the values of specific heat capacity (Cp), density (q), and thermal conductivity (k) of green sand over a measurement range spanning from room temperature to over 800 C. These experiments yielded representative data for the real properties of silico-clay sand, which were subsequently substituted for the generic properties provided in the NovaFlow&Solid software database. It is worth noting that this study is not limited to a single software; the physical properties obtained can be transferred to other simulation software as well. A cooling simulation was conducted with the corrected physical properties of the sand, and the resulting curve better reflects reality than the one initially obtained with the original values. Having a more precise knowledge of the exact temperature of the metal at a given moment is crucial for predicting certain defects.

The composition of the mold atmosphere plays a critical role in mold-metal interactions during steel casting and is a root cause of certain casting defects. However, studying this atmosphere is challenging due to technical difficulties in extracting gases from sand molds and the harsh condi- tions of the metal casting process. Most of the studies realized in the field uses gas chromatography and mass spectrometry technics to investigate atmosphere in a foundry mold. Single-gas sensors, being readily available nowadays, present a cost-effective alternative to aforementioned technics. In this paper, we present an innovative gas analysis system designed for real-time, in-situ moni- toring of the mold atmosphere. This system has been developed in collaboration between ENSAM Cluny (LaBoMaP) and the industrial partner Safe Metal.

In the context of certain metallic alloys, the conformity of the product depends on its metallurgical structure. Addressing this, the implementation of a real-time monitoring system to control the evolution of the metallurgical structure and the geometry of the cogging part is proposed. Focusing on the microstructure's dependence on temperature, this article outlines the requested steps for developing data-driven reduced models for describing the temperature field in the billet. These models use temperature data collected from predictive numerical simulations conducted using FORGE® software. Applying the Proper Orthogonal Decomposition (POD) technique, the images illustrating the temperature field are reconstructed through a 2D matrix-based framework. This matrix, derived from non-discretized elements issued from FORGE®, underwent discretization through an objective method, resulting in a size of 100*100. The utilization of the POD technique in this approach provides a parametric vector description, facilitating rapid image reconstruction through manipulation of vector system parameters. With just two vectors, we can effectively reconstruct the image representing the temperature field.

Although several methods have been proposed for generating as-scanned point clouds, i.e. point clouds incorporating various realistic artefacts that would appear if the corresponding real objects were digitized for real, most of them still fail to take into account the complex phenomena that occur in a real acquisition devices. This paper presents a new way of artificially generating point clouds by combining simulation and machine learning. Starting from the CAD model of the object to be virtually scanned and from a scan configuration, structured light simulation first allows reconstructing a preliminary 3D point cloud. Then, a coverage prediction network is used to predict the regions that would be acquired if a real acquisition was to be done. The prediction model has been trained from a large database of scan configurations and point clouds scanned for real. Finally, filtering and cropping are performed to fine-tune the generated point cloud. Experiments confirm that this method can generate point clouds very close to those that a real scanner would acquire, as shown by several metrics characterizing both local and global similarity. Such a virtual scanning technique enables the rapid generation of large quantities of realistic point clouds, especially when compared to the time-consuming and costly processes involved in using physical acquisition systems. This opens up new perspectives in terms of access to realistic point cloud databases, in particular for the development of various AI-based approaches.

Extended Reality (XR) technologies—including Virtual, Augmented, and Mixed Reality—offer useful possibilities for improving industrial processes such as training, design validation, maintenance, and quality control. However, their adoption in industry remains limited, often because existing solutions do not fully meet practical needs. This study focuses on the co-development of XR applications better adapted to industrial requirements. In a first phase, a literature review helped identify 37 evaluation criteria, grouped into eight categories. Definitions were refined based on expert input. In a second phase, over 20 professionals from different industrial sectors assessed the relevance of each criterion using a 5-point Likert scale and a ranking method. The results showed differences between academic and industrial perspectives. While academic work often highlights technical or sensory aspects, indus-trial stakeholders emphasized usability, relevance of content, and potential for innovation. These find-ings provide a clearer view of industry expectations and will inform future development of XR tools.

The integration of data capture, analysis, monitoring, and control technologies is rapidly becoming the cornerstone of next-generation smart buildings. However, developing digital twins that dynamically interact with these buildings presents a significant challenge. In this paper, we study the most appropriate data models for leveraging a digital twin from data-driven smart buildings. We propose a framework that exploits a knowledge graph to directly address the challenges encountered in real-world building management systems, ensuring that the information is comprehensible as a preliminary step to intelligent decision-making. Furthermore, we validate this proposal for improving building performance and sustainability through a real-world use case. The experimental results, utilizing dynamic data streams from the Internet of Things (IoT), demonstrate promising outcomes. This research paves the way for using graph-based models and algorithms as digital twin enhancers for managing data-driven smart buildings.

This paper improves on the performance of the Deep Learning Additive Manufacturing driven Topology Optimization (DL-AM-TO) approach that was proposed in [4]. DL-AM-TO is a data-driven generative method that integrates the mechanical and geometrical constraints concurrently at the same conceptual level and generates a 2D design accordingly. Furthermore, DL-AM-TO tailors the design's geometry to comply with manufacturing criteria, which facilitates the designer's interpretation phase and prevents him/her from getting stuck in a loop of drawing the CAD and testing its performance. The geometry needs less support structure and hence is printed faster. Consequently, DL-AM-TO accelerates the Design for AM process.

The physical Vapor Deposition (PVD) surface treatment process con-sists of numerous steps involving of multi-physical and multiscale phenomena. i These phenomena are beyond the ability of human perception in their entirety which is a scientific challenge for learning PVD. The present article proposes a Virtual Reality (VR) approach dedicated to the PVD process learning and a pro-totype is developed with different modules. The virtual immersion includes two modalities. One ex-situ, in the surface treatment laboratory, at a real scale (1:1), allowing users to explore the process, the machine components, and to experi-ment with technical gestures such as handling the machine door or installing sub-strate-holder rods inside. The second modality is in-situ, enabling the user to fol-low the process steps immersed in an environment inaccessible to humans and multi-scale. These experiments help to understand the physical phenomena oc-curring thro

Predictive dynamic simulations of virtual machining rely on accurate representation of eigenmodes and damping factors. Historically, the modeling of flexible workpieces requires experimental updating of general modal properties, especially due to a simplified definition of fixtures. In the present work a substructuring-based approach for a virtual machining simulation is developed. It is demonstrated on a vibration-prone boring of a thin-walled automotive workpiece. Fixture-affected zones are modeled via MacNeal-type approach. This enables for addressing the influence of clamping in the mechanical modeling of dynamics, and for creating specific models of typical fixture configuration. During simulation vibrations occur on similar frequencies to those observed on real machining. Resulting surface defects follow alike patterns in simulation and experiment.