Valeo

About: Valeo is a company organization based out in Paris, France. It is known for research contribution in the topics: Heat exchanger & Clutch. The organization has 8904 authors who have published 14808 publications receiving 83614 citations. The organization is also known as: Société anonyme française du Ferodo.

Control Strategies for Open-End Winding Drives Operating in the Flux-Weakening Region

Abstract: This paper presents and compares control strategies for three-phase open-end winding drives operating in the flux-weakening region. A six-leg inverter with a single dc link is associated with the machine in order to use a single energy source. With this topology, the zero-sequence circuit has to be considered since the zero-sequence current can circulate in the windings. Therefore, conventional overmodulation strategies are not appropriate when the machine enters in the flux-weakening region. A few solutions dealing with the zero-sequence circuit have been proposed in the literature. They use a modified space vector modulation or a conventional modulation with additional voltage limitations. This paper describes the aforementioned strategies, and then, a new strategy is proposed. This new strategy takes into account the magnitudes and phase angles of the voltage harmonic components. This yields better voltage utilization in the dq frame. Furthermore, inverter saturation is avoided in the zero-sequence frame, and therefore, zero-sequence current control is maintained. Three methods are implemented on a test bed composed of a three-phase permanent-magnet synchronous machine, a six-leg inverter, and a hybrid digital signal processor /field-programmable gate array controller. Experimental results are presented and compared for all strategies. A performance analysis is conducted as regards the region of operation and the machine parameters.

Deep semantic segmentation for automated driving: Taxonomy, roadmap and challenges

Abstract: Semantic segmentation was seen as a challenging computer vision problem few years ago. Due to recent advancements in deep learning, relatively accurate solutions are now possible for its use in automated driving. In this paper, the semantic segmentation problem is explored from the perspective of automated driving. Most of the current semantic segmentation algorithms are designed for generic images and do not incorporate prior structure and end goal for automated driving. First, the paper begins with a generic taxonomic survey of semantic segmentation algorithms and then discusses how it fits in the context of automated driving. Second, the particular challenges of deploying it into a safety system which needs high level of accuracy and robustness are listed. Third, different alternatives instead of using an independent semantic segmentation module are explored. Finally, an empirical evaluation of various semantic segmentation architectures was performed on CamVid dataset in terms of accuracy and speed. This paper is a preliminary shorter version of a more detailed survey which is work in progress.

Adversarial Frontier Stitching for Remote Neural Network Watermarking

Abstract: The state-of-the-art performance of deep learning models comes at a high cost for companies and institutions, due to the tedious data collection and the heavy processing requirements. Recently, Nagai et al. (Int J Multimed Inf Retr 7(1):3–16, 2018), Uchida et al. (Embedding watermarks into deep neural networks, ICMR, 2017) proposed to watermark convolutional neural networks for image classification, by embedding information into their weights. While this is a clear progress toward model protection, this technique solely allows for extracting the watermark from a network that one accesses locally and entirely. Instead, we aim at allowing the extraction of the watermark from a neural network (or any other machine learning model) that is operated remotely, and available through a service API. To this end, we propose to mark the model’s action itself, tweaking slightly its decision frontiers so that a set of specific queries convey the desired information. In the present paper, we formally introduce the problem and propose a novel zero-bit watermarking algorithm that makes use of adversarial model examples. While limiting the loss of performance of the protected model, this algorithm allows subsequent extraction of the watermark using only few queries. We experimented the approach on three neural networks designed for image classification, in the context of MNIST digit recognition task.

Axial flow fan

Abstract: An axial flow fan has a hub and plural blades. Each blade extends from hub to a blade support ring and has a pitch which decreases over a first inner part of the radial extent and increases over a second outer part of the radial extent. In a preferred embodiment, the trailing edge of the blade tip and the median point of the blade root are situated on a common radial line. In a further preferred embodiment, a median point on the tip chord of the blade is disposed angularly ahead of a median point of the root chord. The fan provides improved noise performance.

Competing Broadband Noise Mechanisms in Low-Speed Axial Fans

Abstract: This paper compares two broadband noise mechanisms, the trailing-edge noise or self-noise, and the leading-edge noise or turbulence-ingestion noise, in several blade technologies. Two previously developed analytical models for these broadband contributions are first validated with well-defined measurements on several airfoils embedded in an homogeneous flow at low-Mach number. Each instrumented airfoil is placed at the exit of an open jet anechoic wind tunnel with or without a grid generating turbulence upstream of it. Sound is measured in the far field at the same time as the wall-pressure fluctuations statistics close to the airfoil trailing edge and the inlet velocity fluctuation statistics impacting the airfoil leading edge. The models are then compared in some practical cases representative of airframes, wind turbines, and automotive engine cooling modules. The airfoil models of the two mechanisms are then extended to a full rotating machine in open space. The model predictions of both mechanisms are compared with in-flight helicopter measurements and automotive engine cooling modules measurements. In both instances, the turbulence-ingestion noise is found to be a dominant source over most of the frequency range. The self-noise only becomes a significant contributor at high angles of attack close to flow separation.