Modeling and simulation of surface reactions and reactive flow of a nitriding process

Due to wide use of ammonia in gas nitriding processes, a numerical study has been undertaken to determine the ammonia decomposition rates over an injector body which was used as a steel catalyst under conditions likely to be employed in such applications. With the extrapolation of the data available for noble surfaces, intrinsic rate data have been obtained for the process temperatures of around 520 °C in furnace conditions with pressures varying between − 100 and 900 Pa around the atmospheric pressure. The computational setup for surface reactions and furnace flow modeling was designed to match the in-service furnace used in Bosch production plants and contained 4.6 million mesh points and results were obtained with URANS k–ε turbulence model.Simulations revealed spatial variations of nascent nitrogen coverages over the injector surfaces and this distribution well agreed with the experimental thickness measurements on the same injector. Furthermore results indicated that the time scale defined by means of local average velocities was not the most relevant parameter to describe the advance of the surface reactions. Hence, the definition of time scale needed to be modified to include the scale of the fluctuation velocity of the turbulent transport in the vicinity of the wall and the local length scale associated with the surface. Such a time scale was shown to imitate the reactivity on the surface very well. It was also clear that when the velocities around the parts in the furnace were set high, flow separated and vortices were formed at both ends of the injectors, which blocked and further destabilized the flow inside the injector passages.

► We performed a numerical simulation of flow inside a nitriding furnace.
► The CFD setup was matched in-service furnace used in Bosch plants.
► We used chemical mechanism including 12 reactions and 9 species.
► Simulations put in evidence of variations of nitrogen coverages over the injector
► Computations well agreed with experimental thickness measurements