Driven by innovative and miniaturization needs for future missiles and smart ammunitions, the thermal management of on-board electronics will have to overcome some of the drawbacks related by traditional cooling solutions currently used in such systems. Based on microsystem technologies, new solutions embedded as close to the electronics, directly into the printed circuit boards (PCB) provide interesting solutions as they meet both reliability, cost and integration requirements especially in harsh military environments (wide temperature range, vibrations, high acceleration). Fluidic cooling systems, also called heat pipes, demonstrate outstanding performances in terms of thermal conduction. Even if direct integration of them in PCBs seems possible as related by different patents, they are still commonly used as external cooling solutions. Despite their high cooling performances, heat pipes also exhibit some challenging weaknesses. First, they are highly sensitive to acceleration due to their working principle based on the capillary pumping of the coolant fluid. Furthermore, implementing such devices requires the use of heavy and cumbersome interface plates so as to make relevant efficient thermal exchange. These constraints, in addition of costly operator-dependent assembly steps, confirm that the use of heat pipes for on-board electronics is not trivial. However, improvements are still possible in this domain and we propose to investigate some of them in this project. As part of this project, a consortium involving C2N (University Paris-Saclay), MBDA Group and EGIDE Company was being set up. The aim is to design, realize and characterize an innovative cooling system, providing both high performances, compactness and reliability, with a targeted TRL of 4. The proposed fluidic cooling device can work using monophasic mode and two-phase mode even if it is not optimized for at first sight. Here, most parts of the fluidic system are embedded in the bulk of the PCB without increasing the thickness of the electronic board. The pumping mechanisms is ensured using an innovative MEMS pump. This device, in the form of a module, will be soldered as the other components at the PCB surface. So, in this way, the pumping will not be ensured by capillarity but will be electrically assisted and will not be subjected to acceleration limit. Most parts of the system are realized using an innovative thermocompression (Cu/Sn/Au/Cu) technique during PCB manufacturing without any modification of the process. This technique allows the realization of thin metalized free shape cavities and so an efficient way to create a perfectly sealed fluidic loop. Based on a THALES patent, this process is provided using a license for exploitation owned by the MEREDIT consortium in which MBDA is a partner. Apart from the fluidic aspect, one of the main challenges is highly linked to the pumping system. In order to set in motion the cooling fluid meeting both, high efficiency and harsh environments needs, we plan to use a micro-pump based on the Electro-Hydro-Dynamic (EHD) phenomena. Such devices are known to be non-mechanical pumping systems, offering extremely high robustness compared to traditional mechanical ones. The innovative 3D module, in ceramic, is manufactured using a well-known Temperature Co-fired Ceramic (HTCC) process thanks to the EGIDE expertise. The pump is then mounted at the PCB surface as a classical component by brazing, allowing a long term hermeticity of the whole fluidic loop. In addition to a heater and a Peltier TEC to mimic the chip to be cooled and the cold point, the final demonstrator envisaged, as close as possible from the industrial solution, will include a set of temperature sensors. This test strategy will allow us to characterize as finely as possible the whole system, in an automated manner over a long period of time.

During the very last years, computer vision has made a significant breakthrough with the emergence of deep learning techniques. Indeed, it successfully benefits to image classification where deep learning outperforms the state-of-the-art in challenges such as the Imagenet Large Scale Visual Recognition Competition (ILSVRC) since 2012. In this project we propose to use this kind of method, more particularly convolutional neural networks (CNN), for the detection and the recognition of multiple small size objects in images. Two applications are considered in this project, the detection and mapping of wales using satellite imaging and the detection and recognition of objects (vehicles) in infrared images. For both applications, the object size ranges typically from 5x5 to 10x10 pixels. To address this detection and recognition problem we propose to design 2 different architectures. The first one is the most common approach and consist in a sliding window that extract patches (small part of the full image). Then, the patches are introduced in a trained CNN to differentiate objects from the background, and possibly, classify them. The second approach deals with the full image in one step. In this case, we design a deep classification net for semantic segmentation. In the final segmented map each pixel gets a label representing (hopefully) its class. These two approaches will be developed by the members of the consortium using the synthetic database provided by MBDA. Note that these CNN architectures must be designed regarding the operational constraints. Following this work, we will deal with the variability of the background in the test database in comparison with the background available in the training database. The objects to be detected and recognize will be considered available in the training database. The goal is to evaluate possible operational situations and the potential losses in the final results. We will also study how the CNN trained on simulated data performs on real data. To fulfil this experiment, new acquisitions in operational situations will be conducted by MDBA in order to complete the existing real images database. Considering the whale mapping application, the idea is to evaluate the adaptation capacity when lower resolutions are used in the test phase. Besides, thanks to the available real data, we also propose to evaluate common methods for incremental learning to specialize the proposed architecture. Along each step of the project, we will evaluate the performances of the CNN and the results obtained. The goal is to monitor the learning process and to use criteria to quantify the final detection and recognition results. Finally, as introduced before, the CNN design will take into account the possible operational constraints. Thus, we will analyse the potential solutions to reach real time implementation in embedded systems. We will deal with both the material (GPU, energy consumption…) and software (how to reduce the computational time) parts.

The thermal accelerometer appears as a real technological breakthrough compared to traditional pendulum accelerometers which are very poorly suited to the high-level shock measurement and their resistance to harsh environments. Indeed, thanks to their architecture, thermal accelerometers are resistant to accelerations up to 50 000g (1g=9,8m/s-2). Besides the ability to survive these extreme environments, the sensor must also allow the measurement of this high level of acceleration with high bandwidth as we have begun to demonstrate with the preliminary results of the initial project. The main purpose of this study is to use the principle of thermal accelerometer to measure strong shock levels with a bandwidth from DC to more than 10 KHz. The use of microtechnologies will significantly reduce footprint and volume but also achieve competitive manufacturing costs. This technology allows us to measure both high acceleration or gravity in continuous operation. This would envisage dual applications both in civilian and defense fields. Indeed in the latter case, there would be applications for the measurement of strong deceleration levels on impact of the specimen in concrete target and then trace the movements with the ability to measure continuous acceleration . In the second case, these devices could be used as shock or vibration sensors in areas such as transport, aerospace, civil engineering or oil exploration, ... At present, there are only very few sensors operating in this range of accelerations having good stability and accuracy. The technology that is used most often is "piezo-resistive" and is North American. They are subject to strong export constraints (ITAR). It is essential to have a European technology or French in order to overcome a US dependency. It is by the way strategic for the defense industry. That is why the development of such a MEMS sensor (Micro Electro Mechanical System) based on the heat transfer would in itself a major innovation in this area.