Ammonia has shown great potential as a carbon-free fuel, in particular for marine transportation and energy production. Its low laminar flame speed, and the tradeoff between ammonia slip and NOx emission, pose challenges for industrial applications, and a more in-depth understanding of the combustion of ammonia is therefore needed. Raman spectroscopy is a powerful diagnostic often employed to investigate turbulence chemistry interactions and resolve the thermo-chemical structure of hydrogen and hydrocarbon-air flames. However, this technique has been used extensively in hydrocarbons (HCs) and hydrogen flames, but no quantitative Raman spectroscopy is available for ammonia flames, despite the current interest in ammonia combustion. First, this work extends Raman spectroscopy to the instantaneous and spatially resolved measurement of major species concentrations and temperatures in ammonia flames. The lack of detailed ammonia spectra at high temperatures, the strong flame luminosity, and fluorescence interference are the major obstacles to the implementation of Raman spectroscopy to ammonia flames. This thesis introduces a novel approach to estimating the temperature dependence of the Raman signal and fluorescence interference contributions from a series of counterflow diffusion flames. Species concentrations and temperature profiles from measurements are shown, and their accuracy and precision are discussed. Next, this work obtains the first quantitative Raman measurements of temperature, mass fractions, and mixture fractions in two turbulent ammonia/hydrogen/nitrogen diffusion jet flames simulating 14% (CAJF14) and 28% (CAJF28) partial ammonia cracking ratio with Reynolds numbers of 11,200. The scalar structure in turbulent flames is examined using conditional mean and RMS radial profiles, scatterplots in mixture fraction space, as well as statistics conditioned on mixture fraction and physical spaces. Finally, the probability of localized extinction and the differential diffusion (diff-diff) effects are analyzed in two turbulent flames. Lastly, an improved Raman/Rayleigh system is introduced for ammonia combustion at atmospheric pressure, which enables the 1D simultaneous laser-induced fluorescence (LIF) measurements of the amidogen (NH2) radical and interference-free Raman/Rayleigh measurements of major species and temperature in non-premixed and premixed NH3/H2-air flames.

Une variété d’applications techniques dans la conversion d’énergie sont basées sur la combustion turbulente. En dépit du fonctionnement avancé de la technologie, la science de la combustion (turbulente) en est à un stade relativement jeune. Des études détaillées des flammes de référence sont essentielles pour mieux comprendre la physique sous-jacente aux processus mentionnés, ainsi que pour fournir une base de données fiable permettant de valider les modèles numériques développés pour simuler des problèmes de combustion turbulents. La spectroscopie Raman / Rayleigh est une technique extrêmement utile qui permet d’accéder simultanément à des informations sur la température et la concentration en espèces chimiques principales dans les structures de flammes. La data reduction strategy appliquée avec cette technique est cruciale pour extraire des informations fiables des investigations expérimentales. Dans cette thèse, une version modifiée de la stratégie, basée sur la Hybrid Method de Fuest et al., a été développée et testée à l’aide de techniques de résolution de NLP problem, dont des méthodes d’optimisation globale et un algorithme génétique. La stratégie proposée permet une réduction significative du temps de traitement des données, nécessite moins d’expertise de l’utilisateur et réduit l’erreur de mesure. La procédure modifiée a été appliquée à un data set fourni par une étude expérimentale de deux jet flame méthane/air prémélangés turbulentes appartenant au régime flamelet de la combustion prémélangée turbulente. Le data set est composé de plusieurs scalaires, y compris les principales espèces et la température, mesurés simultanément avec la single-shot 1D Raman/Rayleigh spectroscopy. Les résultats des mesures sont analysés et discutés. / A variety of technical applications in energy conversion are based on turbulent combustion. Despite the advanced contest of operation, (turbulent) combustion science is at a relatively young stage. Detailed investigations of benchmark flames are essential to achieve a better understanding of the physics behind the mentioned processes, as well as to provide reliable database for validating numerical models developed to simulate turbulent combustion problems. Raman/Rayleigh spectroscopy is a highly valuable technique which allows to access simultaneous information on temperature and main chemical species concentration within the flame structures. The data reduction strategy applied with this technique is crucial, in order to extract reliable information from the experimental investigations. In this this thesis, a modified version of the strategy, based on the Hybrid Method by Fuest et al., has been developed and tested using NLP problem solving techniques, including global optimization methods and a genetic algorithm. The proposed strategy allows for a significant reduction of the data processing time, requires less user’s expertise and returns better results in reduced measurements error. The modified routine has been applied to data set provided by an experimental investigation of two turbulent premixed methane/air jet flames belonging to the flamelet regime of turbulent premixed combustion. The data set is composed by multiple scalars, including major species and temperature, simultaneously measured with single-shot 1D Raman/Rayleigh spectroscopy. The results of the measurements are analyzed and discussed.