International audience

In current sensor-based monitoring solutions, each application involves a specified deployment and requires significant configuration efforts to adapt to changes in the sensor field. In this paper, we propose a generic solution that relies on the massive deployment of battery-powered sensors. More precisely, we present a solution for LPWAN sensors emissions scheduling to ensure overall regular sensor data emissions over time (at a rate chosen by the user) while limiting management costs incurred by sensors’ arrivals and departure. Our objectives include monitoring quality that we evaluate through a "diversity" metric encompassing that information value depletes with time, plus management cost quantified by the number of orders sent to sensors. Modeling arrivals and departures as random processes, we compute those performance metrics as functions of the overall data reception period selected and evaluate them against alternative scheduling methods. We show that our solution is better suited for Massive IoT contexts.

In the industry of hot water tanks, welding quality plays an important role in the durability of the final product. Welds are often inspected visually by the operator, which tends to be time-consuming and prone to a high error rate. Machine learning and Deep learning offer solutions for the automation of this task. In this paper, we propose a system for welding anomalies detection from weld images. We begin by developing an image acquisition system and then a software based on the You Only Look Once v3 (YOLOv3) network. The results show that the model identifies and localizes welding anomalies with high accuracy and fast inference time.

In the semiconductor industry, Failure Analysis (FA) is an investigation to determine the root causes of a failure. It also involves an intermediate analysis to build the steps of the failure analysis in order to mitigate future failures and to facilitate the future FA. In the framework of the FA 4.0 project, the reporting system records three items of information using natural language: the failure analysis request description (input space) and analysis steps (paths), as well as generic categories of root cause conclusion (output space). The main objective of this article is to develop a system which is able to automatically help industries carry out fault analysis diagnoses with Artificial intelligence (AI). This article extends and validates the adapted methodology proposed by [1] to transform text data into numeric data based on Natural Language Processing (NLP). It transforms the text data from the input space and output space. Different deep learning algorithms based on a Variational AutoEncoder (VAE) are applied to the output space to reduce the dimension of the numeric data, and the performance of each VAE is evaluated with different metrics. The Generalized-Controllable VAE (GCVAE) is the one best suited to our case. A Gaussian Mixture Model (GMM) is then used to perform clustering in the latent space generated by the GCVAE. A centroid analysis is also conducted to verify the similarity of each cluster.

In today’s monitoring solutions, each application involves custom deployment and requires significant configuration efforts to accommodate sensor changes. In contrast, in this paper, we consider a massive deployment of battery-powered sensors to propose a more versatile monitoring solution that is not tied to the physical deployment of devices.First, we define a framework for the definition of a monitoring strategy, for which we propose a generic monitoring accuracy metric, which, weighted to the lifetime of the monitoring network, allows the characterization of a multi-objective problem.We then introduce a specific two-parameter instantiation for the period update function, that ensures strictly periodic emissions from sensors even when new sensors join the system over time. We show through simulations how the two parameters– target emission period and number of jointly used sensors–can be chosen according to the objectives for the monitoring, by highlighting the Pareto front for accuracy and energy-efficiency.1