Prediction Sciences (France)

In today's rapidly evolving industrial landscape, efficient predictive maintenance solutions are essential for minimizing downtime and enhancing productivity. This research introduces an adaptive cloud-based model pipeline for predicting the Remaining Useful Life (RUL) of machine components, specifically ball screws. The pipeline integrates local preprocessing, edge computing, and cloud-based adaptive model training, ensuring data privacy and reducing data transmission volumes. The system classifies wear states using various machine learning mod-els and predicts RUL through regression analysis, incorporating uncertainty quantification for robust maintenance scheduling. The experimental setup includes accelerated degradation of ball screws, with data collected via a three-dimensional accelerometer. Feature extraction and data augmentation techniques are employed to enhance prediction accuracy. Random Forest and Gradient Boosting models demonstrate superior performance, with Random Forest selected for its robustness and uncertainty quantification capabilities. Empirical results indicate high prediction accuracy, with Random Forest achieving up to 91% accuracy in Phase 2. This cloud-ready predictive maintenance framework leverages scalable cloud infrastructure for efficient data processing and real-time updates, offering a practical solution for industrial applications. The proposed approach significantly advances the adoption of digital business models within the manufacturing industry, providing a reliable and efficient tool for predictive maintenance.

International audience

The fatigue properties of filled elastomers are strongly related to the inclusions’ population induced by their complex recipes and the successive stages of the manufacturing process (mixing, injection and curing). The improvement of these properties involves at first an ability to describe the statistical features of these inclusions’ population in terms of size, geometry, nature and spatial distribution. Then a detailed understanding of the damage mechanisms is required in order to define the mechanical criticality of inclusions according to their characteristics under cyclic loading. The aim of the study is to take advantage of tomography observations that enable a three dimensional vision to access the statistical description of the inclusions population and of the related damage mechanisms. First of all, a tool to detect and characterize accurately inclusions and cracks is presented. Then, this tool is applied to an interrupted specimen in fatigue, for a compound filled with carbon black. A statistical analysis is performed on the nature of inclusions leading to cracks, on the kind of damage mechanisms (polar crack, cavitation between 2 inclusions or failure of an inclusion) and on the density and the geometric properties of activated inclusions. An additional focus is given on the inclusions leading to the biggest cracks in order to assess more precisely the features (nature, size…) of the most critical inclusions for fatigue.

A latent heating peak in the PBL was detected in a simulation by a global GCM that failed to reproduce Madden-Julian Oscillation (MJO). The latent heating peak in the PBL was generated by very shallow convection, which prevented moisture from being transported to the free troposphere. Large amount of moisture was therefore confined to the PBL, leading to a dry bias in the free atmosphere. Suffering from this dry bias, deep convection became lethargic, and MJO signals failed to occur. When the latent heating peak in the PBL was removed in another simulation, reasonable MJO signals, including the eastward propagation and the structure of its large-scale circulation, appeared. We therefore propose that the excessive latent heating peak in the PBL due to hyperactive shallow convection may be a reason for a lack of MJO signals in some simulations by other GCMs as well.