This PhD dissertation was carried out at the Technical University of Denmark in Department of Mechanical Engineering and supervised by Associate Professor Jesper Schramm. The PhD project was funded by the Technical University of Denmark. Demands on reducing the fuel consumption and harmful emissions from the compression ignition engines (CI engines or diesel engines) are continuously increased. To comply with this, better modeling tools for the diesel combustion process are desired from the engine developers. The complex combustion process of a compression ignition engine may be divided into three separate combustion phases; a pre-mixed homogeneous combustion phase, a non-homogeneous pre-mixed combustion phase and a diffusion controlled combustion phase. This PhD thesis comprises a study of the non-homogeneous pre-mixed combustion process. Experimental work was performed with a pre-mixed flat flame burner at atmospheric pressure. The fundamental mechanisms of this laboratory flame are identical with those of an IC engine. The high control of the laboratory flame and the easy access to it makes it valuable for model validation. Modeling of the pre-mixed flat flame was performed with computational fluid dynamics (CFD) including combustion chemistry, soot formation and thermal radiation. First main topic of the project was to deal with the post flame instability, also known as flame flicker, of a premixed flat flame. A nonintrusive method to stabilize the flame, called helium stabilization, was developed. The basis of the method is to ensure almost equal gas densities inside and outside the flame by diluting the surrounding gas with helium. This newly stabilized flame offers some features which are important in combustion research. The flame is very uniform and may be highly lifted, thus it is well suited for optical line of sight diagnostics in both pre and post combustion regions. The work also includes some preliminary studies of radiant emissions from helium stabilized ethylene/air and methane/oxygen flames. It is demonstrated that nano particles below the sooting threshold actually are weakly luminous. Second main topic of the project was to enable valid temperature measurements of sooting flames. A thermometry method based on IR emission and absorption by CO2 between 2100 and 2400 cm-1 was developed specifically for pre-mixed flat flames. Detection was performed with a commercial FT-IR spectrometer. An investigation of the method’s precision and accuracy was performed separately. The investigation shows that an important strength of the method is its reliability with regard to variations in the experimental setup. Even with non-optimized installation and operation of the method, the average line of sight temperature is determined with a precision of +/- 2 %. With a thorough installation and operation the method detects the temperature of the flame core within +/- 1 %, in spite of the line of sight principle. Temperature measurements from the literature performed with different thermometry techniques were reproduced and compared. Some of the methods from literature were valid only for a limited range of flame conditions. Thus the consistency of the most established flame thermometry techniques was evaluated, with the IR emission absorption method used as a link. An optical method was also used to measure the soot content in terms of soot volume fraction. The measurement was based on absorption of light with wavelength between 500 and 520 nm. The third main topic of the project was to develop a CFD model of a pre-mixed flat flame. The commercial CFD software CFX 11.0 was used. Three cases of ethylene/air flames well known from the experimental work, was used for the model validation. Two cases were helium stabilized flames with φ = 1 and 2.14. The third case was an unstable flame with φ = 2.14. The unstable case was used to test whether a transient model would be able to predict the frequency and shape of the unstable sooting flame. The stable cases were easier to treat both experimentally and numerically, and were used for more detailed validation. The model was developed stepwise from the simplest possible starting point. The first part of the development was made on a model of thermal fluid flow only. The model was incrementally extended with; temperature dependent thermal conductivity, temperature dependent viscosity, temperature dependent mass diffusion, thermal gas radiation, water gas shift reaction and soot radiation. In the second part combustion was introduced. A simple reaction model well known from the literature did not predict the features of the real flame, thus another simple reaction model was developed specifically for this purpose. A simple soot formation model developed for acetylene diffusion flames was validated against the soot measurements. The model parameters were fitted to predict the measured soot volume fractions in the pre-mixed ethylene flame. The provided results offer insight to the sensitive of the predictions to the model variations.