Laboratoire Génie de Production

The manufacturing of high-temperature environment parts (combustion chambers, turbojet or rocket injectors) with the Laser Power Bed Fusion (LPBF) process is very promising due to the freedom on geometries it provides. In order to gain a better understanding on a realistic flame–wall interactions, 316 L steel plates from LPBF and cast /forged processes were subjected to a reacting flow using robust cross-disciplinary methodology. The physical phenomena (thermal, physico-chemical and mechanical) involved in structural parts (full-scale) subjected to combustion-based burnt gases is studied via the use of a laboratory-scale combustion test rig. The generation of a methane-air flow enables to conduct a comparative analysis to understand the influence of a well-controlled thermal stress (1200°C, 1 h) on the material properties (hardness, residual stresses), depending on the manufacturing process and microstructure evolution during the thermal exposure. Results highlights that the interaction between the flame burnt gases and the material does not affect the microstructure and mechanical behaviour for cast whereas localized recrystallization in the range of 0–10 μm near the exposed surface and residual stress state modifications are revealed for LPBF. This is the first study on flame–LPBF alloy interactions.

Milling of hardened carbon steels like SKD11 generates complex physical interactions that can affect tool wear, surface quality, and process stability. This study presents a multiphysics experimental approach aimed at better understanding these physical interactions. The first objective is to monitor key features in real-time, including cutting force, vibration, cutting temperature, and sound signal, during the milling process. An integrated setup involving infrared thermography and thermocouples, a piezoelectric dynamometer, sound and vibration emission sensors was used, and data were collected under two different cutting conditions. The second measurement objective is to inspect the end-of-process features such as surface quality, tool wear, and chip formation. Tool wear was measured by different methods at the end of the cut in order to show the strong correlation between cutting forces, temperature, sound signal, and vibration patterns of the machining process. The results of this study are focused mainly on the design approach rather than on detailed numerical details. The tooth-passing frequency analyzed from the cutting-force signal was correlated with the sound signal. Cutting strategies affect the surface quality under the same cutting conditions. This work provides guidance on eliminating redundant measurements for multiphysics investigations of the milling process, and accounts for all features, in contrast to previous experimental studies. Besides those findings, there are some issues that need to be improved, such as repositioning the thermocouples, elongating the cutting path to inspect the tool state and surface quality; more cutting conditions should also be applied.

Physical human-robot interaction (pHRI) requires an adapted robot control to ensure both efficient task execution and human safety and comfort. In this regard, human-in-the-loop robot controls have been shown to be a viable option. Thus, accurately modeling human behaviors based on the task performed becomes necessary and optimization methods appeared to be efficient in this matter. In this study, we compare two minimization methods -Nelder-Mead simplex method, frequently used for this purpose, and genetic algorithm NSGAII- to identify the weights of cost functions in our optimal control problem (OCP). This problem was formulated to effectively model the movement of a human who performs a collaborative pick-and-place with a human partner, as we consider human-human collaboration to be the most natural form of collaboration for human comfort. The movement is an average one computed from data from 30 subjects, ie. 15 pairs. The results demonstrate that NSGAII found a solution better correlated with our measured data, achieved lower root mean-squared errors (RMSE) across all axes and converged faster than the Nelder-Mead method.

Compression-tension asymmetry (CTA) in yield strength is one of the issues that hinders widespread application of wrought Magnesium (Mg) alloys. Strong basal texture formed in thermomechanical processing and the difference in activation of deformation twinning in tension and compression are responsible for CTA. To reduce CTA, the readily implementable Double Equal Channel Angular Pressing (D-ECAP) is used in this study for altering the microstructure and texture of Mg samples. Mg bars containing 1, 5, and 10 wt% Gd were subjected to D-ECAP at 400 °C. Due to plastic strain, the average grain sizes decreased from several hundred microns to about 13.4 μm, 5.6 μm, and 5.1 μm, respectively. All processed samples showed high strength and characteristics shear textures. The crystallographic textures displayed the so-called C1-C2 and B fibers. C1-C2 were the major fibers in the 1 % and 5 % Gd samples, while the B and C1 fibers appeared in the 10 % Gd sample. C1-C2 requires high activity of pyramidal <c + a > slip, and B belongs to basal slip. The CTA was measured by the ratio of the compression/tensile yield stress and significant increase in CTA was obtained on the D-ECAP processed samples with respect to the base alloys. The CTA reached 0.85 in the Mg 10Gd alloy without sacrificing the yield strength. Polycrystal plasticity simulations were done using the experimental textures and good reproduction of the CTA values were achieved. The simulation results also revealed the relative activity of the slip and twinning systems for understanding the mechanisms that control the CTA. The results of this study revealed that D-ECAP is an efficient processing technology that can be widely used in the preparation of high-performance Mg Gd alloys. • D-ECAP reduces the compression-tension yield asymmetry (CTA) of Mg Gd alloy. • Polycrystal plasticity simulation reproduced the experimental CTA values. • For random texture with twinning, a CTA of 0.8 predicted, not the expected 1.0.

Predicting the remaining useful life (RUL) of fast-charging lithium-ion batteries using early-stage lifecycle data is remains challenging due to limited run-to-failure data and lack of knowledge on battery degradation mechanisms. To address this issue, a generic Physics-Informed Machine Learning (PIML) framework is developed. The PIML framework consists of two parallel branches: a physics-informed (PI) branch and a data-driven branch. The PI branch is a neural network stacked by the linear projection layers with embedded physics knowledge, while the data-driven branch is a task-specific machine-learning model. In addition, a three-step training strategy is introduced, including (1) Training the data-driven branch, (2) Training the PI branch for aligning physical consistency without updating the hyperparameters in the data-driven branch, and (3) Fine-tuning both branches simultaneously to achieve optimal performance. To validate this framework, a physics-based model that represents the growth of solid electrolyte interphase (SEI) and a dilated convolutional neural network are implemented in the PI and data-driven branches, respectively. The solid electrolyte interphase-informed dilated convolutional neural network (SEI-DCN) model is demonstrated on the Stanford–MIT–Toyota-battery dataset. Using only four lifecycle data, the SEI-DCN model achieves very high prediction accuracy compared to standard dilated CNNs and other state-of-the-art models under various testing conditions and lifetime ranges. Moreover, the framework is generalizable to different physics-based battery degradation models. • Novel dual-branch parallel PIML framework to merge varied knowledge on battery degradation. • The new learning strategy ensures that PIML is lower bounded by the performance of data-driven models with imprecise knowledge. • Building SEI-informed DCNN validated on the fast-charging lithium-ion batteries and outperformed the SOTA model. • SEI-informed DCNN preserves SOTA prediction accuracy on completely new data with different operation conditions and life spans of the training set. • Investigation of the adaptability of the proposed framework for knowledge replacement across various SEI models.