Natural language interactions for anomaly detection in mono and multi-variate time series using fondation models and retrieval augmented generation
Anomaly detection in mono and multi-variate time series highly depends on the context of the task. State-of-the-art approaches rely usually on two main approaches: first extensive data acquisition is sought to train artificial intelligence models such as auto-encoders, able to learn useful latent reprensations able to isolate abnormality from expected system behaviors; a second approach consists in careful features construction based on a combination of expert knowledge and artificial intelligence expert to isolate anomalies from normal behaviors using limited examples. An extensive analysis of the literature shows that anomaly detection refer to an ambiguous definition, because a given pattern in time series could appear as normal or abnormal depending on the application domain and the immediate context within the successive observed data points. Fondation models and retrieval-augmented generation has the potential to substantially modify anomaly detection approaches. The rationale is that domain expert, through natural language interactions, could be able to specify system behavior normality and/or abnormality, and a joint indexing of state-of-the-art literature and time series embedding could guide this domain expert to define a carefully crafted algorithm.
Design of a new light-sheet microscope for the temporal monitoring of organoids-on-chip
The subject of the thesis is the development of a fluorescence light-sheet microscope for the optical characterization of organoids-on-chips and 3D organoids. The thesis will focus on the conception of a compact multi-color 3D system, to allow time-lapse imaging of multi-scattering 3D samples directly in a cell culture incubator. The work will begin with a clear understanding of the miniaturization process on the quality of the images. The excitation will be correctly model to avoid optical artefacts and to allow the deepest penetration into the biological tissue. The candidate will be responsible to test different optical strategies as well as different excitation wavelengths. As a final step, the system will be characterized in a cell culture incubator for the morphological and functional monitoring of organs-on-a-chip and 3D organoids using specific fluorescent markers. If needed, novel modifications of the microfluidic chamber with integrated optical functions will be proposed. The research program will mainly focus on the morphological and functional monitoring of two samples: pancreatic organoids on a microfluidic chip and 3D brain organoids.
Taylor bubbles: experiments and modeling
This doctoral position focuses on the microscale phenomena that occur in the near-wall region at bubble motion in a capillary tube (also known as Taylor bubble). These are elongated bubbles of vapor having bullet-like shapes that are formed in compact heat exchangers used in a variety of industrial applications such as cooling of electronics and steam generators in nuclear reactors, for instance. The phenomena associated include dewetting dynamics, formation of micrometric thick liquid layers and heat transfer. The PhD candidate will conduct an experimental study at STMF/CEA and SPEC/CEA (Paris-Saclay, France) using advanced non-intrusive optical diagnostics to evaluate these phenomena, in particular the film profile. The student will use the experimental data to validate a numerical approach in the open-source software OpenFOAM that will be developed in partnership with the University of Nottingham (United Kingdom).
Experimental study and physical modelling for the characterization of the flow in an inclined pipe and of the exiting jet trajectory
The thermohydraulic circuits found in the nuclear industry consist of a complex network of horizontal, vertical, or inclined pipes. In particular, the Emergency Core Cooling (ECC) pipes are connected to the primary circuit and are intended to inject cold water into it in accidental situations. Different configurations and inclinations can be found depending on the reactor type. The flow in the ECC pipe, as well as the efficiency of the reactor cooling, are influenced by these different configurations. Therefore, characterizing the flow in these pipes is crucial.
The objective of this thesis is to acquire new experimental data to characterize the water-air flow in an inclined pipe and the trajectory of the exiting jet at atmospheric pressure. The experimental data obtained will be used to develop and/or improve the modeling of flows in inclined pipes.
This thesis should provide the following contributions:
• Set up the experimental facility and the measurement methodologies ;
• Acquisition of the experimental data for different test section geometries (round vs. square), pipe diameter, roughness, inclination, fluid density/viscosity, and flow rate;
• Development of physical models to characterize the flow in an inclined pipe, including liquid hold-up and detachment length;
• Development of a mathematical formulation to predict the trajectory of the exiting jet;
• Study of the impact of stratification in the ECC pipe on the Cocci et al. jet condensation model;
• Optionally, simulation of the experiment using a CFD code (e.g., NEPTUNE-CFD) to extend code validation and identify potential improvements.
Study of catalysis on stainless steels
The materials (mainly stainless steels) aging of the spent nuclear fuel reprocessing plant is the focus of an important R&D activity at CEA. The control of this aging will be achieved by a better understanding the corrosion mechanisms the stainless steels in nitric acid (the oxidizing agent used in the reprocessing steps).
The aim of the PhD is to develop a model of corrosion on a stainless steel in nitric acid as a function of temperature and the acid nitric concentration. This PhD represents a technological challenge because currently few studies exist on in situ electrochemical measurements in hot and concentrated nitric acid. The PhD student will carry out by coupling electrochemical measurements, chemical analyses (UV-visible-IR spectrometry...) and surfaces analyses (SEM, XPS,…). Based on these experimental results, a model will be developed, which will be incorporated in the future in a more global model of the industrial equipments aging of the plant.
The laboratory is specialized in the corrosion study in extreme conditions. It is composed of a very dynamic and motivated scientific team which has the habit to receive students.
Design and realization of a high-temperature optical neutron detector. Application to an experimental program in the JOYO reactor
As part of the development of fourth-generation sodium-cooled fast reactors, the CEA/IRESNE Dosimetry and Instrumentation Laboratory is working on innovative neutron measurement systems capable of operating at temperatures of the order of 600°C, and insensitive to the parasitic phenomena that occur under these conditions. Recently, a new type of optical signal neutron detector (ODN) has been developed at the laboratory. Despite a more complex signal interpretation, this instrument has several advantage: it can be miniaturized and it is intrinsically insensitive to problems of partial discharges and leakage currents that occur in ionization chambers at high temperature.
We propose to pursue the theoretical and experimental development of ODNs to adapt them to high temperatures. The PhD student will further develop the modelling tools already available in the laboratory for simulating the detector response. The work will investigate heavy ion-noble gas interaction cross sections, also a radiative collisional model to predict emission spectra and their temporal dynamics. Part of the work will involve dimensioning a high-temperature prototype and testing it in the JSI TRIGA reactor. Ultimately, the detector will be qualified in the JOYO research reactor as part of a broader experimental program.
Effect of substitution on the ferroelectric and photocatalytic properties of Barium Titanate nanoparticles
As part of the energy transition, the production of hydrogen from solar energy appears to be an extremely promising means of storing and then producing energy. To develop on a large scale, water photoelectrolysis requires materials with high catalytic efficiency. Among the candidates considered, materials derived from barium titanates appear promising because their ferro- and piezoelectric properties could increase their photocatalytic effect. Therefore, we propose to synthesize BaTiO3 nanoparticles by flame spray pyrolysis and to make substitutions on Ba and O in order to study the effect of these modifications on the ferroelectric properties of the material. The addition of noble metal inclusions to the surface of the particles, likely to improve catalysis, will also be addressed. Finally, photocatalysis and piezocatalysis tests will make it possible to establish the links between ferroelectric and catalytic phenomena in this family of materials. This subject will be carried out in collaboration between the LEEL of the CEA and the SPMS of Centrale – Supelec.
Experimental investigation, modeling and impact of corium thermophysical properties in a nuclear reactor severe accident
In a nuclear reactor, certain scenarios can lead to extreme conditions called serious accident conditions with thermal meltdown of the core. In this case, a multiphase mixture, called "corium" based on nuclear fuel and molten metals, will form at high temperature (>2000°C), such as in the Chernobyl reactors in 1986 and Fukushima-Daiichi in 2011. Knowledge of the thermophysical properties of corium remains limited to date due to the domain of very high temperatures, the multiphasic nature (solids, liquids, gases) and the multiplicity of corium compositions. Knowledge of these properties remains a fundamental issue for the safety assessments of nuclear reactors. Limited current knowledge leads to a lack of robustness of these assessments.
To respond to these fundamental scientific challenges, the CEA has been developing three areas of research for several years:
• a first experimental part concerning the measurement the thermophysical properties - in particular density, surface tension, and viscosity - of corium,
• a second part concerning the parametric modeling of the corium thermophysical properties,
• a third component linked to small-scale modeling using molecular dynamics.
A first thesis has led to the development of an original experimental device on the VITI facility of the PLINIUS severe accidents platform of the IRESNE institute, operated by the LEAG laboratory in Cadarache, capable of measuring the surface tension of corium up to 2700°C using the MBP (Maximum Bubble Pressure) technique. In a complementary approach, a thesis made it possible to develop the ATTILHA installation of the ISAS institute operated by the LM2T laboratory in Saclay, allowing access to precise density measurements at high temperatures (> 2000°C). Finally, a thesis in progress at the LMAG laboratory of the IRESNE institute in Cadarache on the modeling of the interfacial tension of corium using the Calphad approach and Butler formalism has led to first results on certain compositions representative from in-vessel corium.
In continuation of the previous works, the LEAG, LMAG and LM2T laboratories jointly propose a thesis which aims to expand the experimental database by relying on the following approach:
• the first part includes the extension of the characterization tests to an expanded grid (composition, temperature) for various coriums of interest. The PhD student will have to carry out these measurements of properties of uranium samples using the cutting-edge instrumentation implemented on the VITI and ATTILHA experiments. It will also be a matter of improving data post-treatment for test interpretation by improving existing inverse method tools.
• The second part deals with the modeling of these thermophysical properties, in order to explore in a more complete way the (vast) grid (composition, temperature), which will remain partly inaccessible to measurement, due to the difficulty of the latter in extreme conditions. The PhD student's work will focus on improving existing ad hoc tools, based in particular on the Calphad approach.
• The last part involves considerations at the system scale, where the doctoral student will apply reactor calculations to severe accident scenarios, via the PROCOR tool in order to evaluate the sensitivity of accident scenarios to the new correlations obtained for the thermophysical properties of interest.
DESIGN OF A MONOLITHIC PIXEL SENSOR FOR PARTICLE PHYSICS WITH AN EMBEDDED ADAPTIVE READOUT ELECTRONICS
In current and future high-energy physics experiments (i.e. upgrades of large detectors at the LHC and experiments in future colliders), the granularity of particle detectors continues to increase, and the use of multi-channel submicron integrated circuits has become a standard.
This granularity was taken one step further in the field of "Monolithic Active Pixel Sensor" (MAPS) technology, where pixel sizes can be as small as 10 x 10 µm2. These small pixels make it possible to achieve record spatial resolutions or greatly improve the radiation resistance of the trace detector, at the cost of a large quantity of data produced. This large amount of data is acceptable where a maximum spatial resolution is required, but can be prohibitive when this is not necessary, or when space and consumption constraints put limits on the number of fast downstream links.
Each experiment therefore requires to redefine the combination of the pixel size and the architecture of the detector's readout electronics, in order to meet the occupancy rate requirements of each physics experiment, and the detector's readout capabilities.
A major innovation in the design of pixel sensors for particle physics is to decouple the pixel matrix from the data rate sent.
As part of a team that has been developing MAPS since 1999, the approach required for the thesis is in a first step to study the existing trace detector architecture in order to understand its limitations in terms of radiation resistance. In a second step, the thesis will focus on information grouping options, assessing the impact of these options on data reduction as well as on induced information loss.
This will be supported by the design of a system-on-chip architecture, including pixel array optimization and digital processing, validating the work carried out in an integrated circuit.
To this end, this thesis will focus specifically on one of the major experiments at the European Center for Nuclear Research (CERN): the Upstream Tracker detector for the LHC Beauty Quark Experiment (LHCb).
In situ Magic Angle spinning NMR analysis of Li-ion batteries
In situ solid-state Nuclear Magnetic Resonance (ssNMR) is a valuable characterization tool to decipher the electrochemical phenomena during battery operation. However, the broad signal lineshapes acquired from the sample static condition often retrain from the full potential of ssNMR characterization. Ex situ ssNMR experiments, using Magic-Angle sample Spinning (MAS), are often necessary to interpret the in situ data. As in any ex situ characterizations, the analyses do not always represent the real electrochemistry because of unwanted artifacts from the ex situ sample preparation, i.e., cell dismantling and electrode separations. Consequently, in situ ssNMR applications have been limited. The PhD student will address this limitation by developing a spinning battery cell for acquiring high-resolution ssNMR data under MAS for in situ study, including a new method of spatially-resolved ssNMR spectroscopy. Combining in situ, MAS, and localized spectroscopy would lead to an unprecedented in situ ssNMR tool for deciphering fundamental insights into battery chemistry, which the student will emphasize by studying phenomena such as interfaces and dendrite formation in operating Li-ion batteries.