Flow Phenomena in Selective Catalytic Reduction Systems used in Large Two-stroke Marine Diesel Engines

Abstract

The present thesis concerns experimental and numerical investigations of flow related to Selective Catalytic Reduction (SCR) systems on large low-speed two-stroke (LSTS) marine diesel engines. These SCR systems are used to remove NOx (mono-nitrogen oxides). These engines undergo large changes these years, where the allowed emissions, especially NOx, are decreased with up to 80%. These restrictions force the shipping industry to radically optimising or changing their engines. This thesis focus on two aspects of the SCR system. First the optimal way of predicting the mixing of a tracer gas into the main gas, related to the mixing of the reductant into the exhaust gas. Secondly how to uniformly distribute the mixture onto the catalyst surface with a minimum of pressure loss, related to the utilisation of the entire catalyst. Three experimental setups are created for these investigations. The first is for mixing of gases in pipes. The second and third setups are for investigating the flow in scale models of the SCR reactor, where the second one use steady flow and the third one use pulsation flow. The first experiment on mixing of gases in pipe flow is performed at two different Reynolds numbers, Re = 50000 and Re = 100000. The setup consists of a long pipe with the inner diameter D = 0.2m, where the upstream conditions to the injection point can be changed. At the injection point methane (CH4) is injected as a tracer gas at the centerline into the bulk flow of air. The measurements are conducted 5D and 10D downstream the injection point. The turbulent pipe flow is ensured with a vena contracta created at the sharp-edged inlet, making the flow less dependant on the Reynolds number. The upstream conditions to the injection point can be changed to: a 10D straight pipe, a 90◦ bend connecting two 10D straight pipes and a 10D straight pipe with a mixer plate mounted 2D upstream the injection point. Laser Doubler Anemometry (LDA) is used to measure the velocity in lines along the pipe diameter. At the same locations, the concentration of methane is determined with the use of probe-based Fourier-transform infrared spectroscopy (FTIR). A model is constructed numerically, with two different Reynolds Averaged Navier-Stoke (RANS) turbulence models: the SST k-ω model and the k-ε model. Both the RANS models fail to predict the change of inlet conditions. The k-ε model fits the experimental results the best. A scale resolving RANS-LES hybrid, also called detached eddy simulation (DES) model is also tested. It predicts concentrations profiles close to the measured results. This is important knowledge for the engine designer since trustworthy mixing predictions are needed. A scale model for the SCR reactor is created, with steady-state airflow as the main fluid. The model has an inlet diameter (d = 0.104m) and a sudden expansion which expand to a diameter of 2.79d within a length of 0.5d. In this expansion, massive flow separation happens and an object can be placed here for improving flow uniformity further downstream at the catalyst. The object is denoted a flow distributor. Different inlet conditions are tested; a straight pipe, a straight pipe with a 90◦ bend and a straight pipe with two 90◦ bends mounted in a out-of-plane configuration. Two different catalyst dummies can be placed downstream of the expansion. One using a single mesh to model the pressure resistance of the first catalyst layer (mesh dummy), and one using a bundle of straws to model the exit flow from several catalyst layers (strawdummy). The velocity field is measured with Stereoscopic Particle Image Velocimetry (PIV) in front of the mesh dummy and behind the straw dummy. The results show that the flow distributor improves the flow uniformity to an acceptable level. Furthermore, when changing the inlet condition from a straight or a bended inlet, to an out-of-plane bended inlet or an inlet with the flow distributor inserted, the uniformity is unaffected of the inlet conditions. This is very important knowledge for the engine designer, since this gives more freedom for laying out pipes in the tightly packed engine room. When looking at the flow upstream of the mesh dummy using the straight inlet an instability for the flow is observed. This instability is seen as a locking effect between two of the three channels in the flow distributor, where the main part of the flow uses one channel, observed up to ≈ 200s. This locking effect could have a negative effect on the catalyst performance and would be time-consuming to model numerical. The third scale model is similar to the second model, where the difference is that the flow is incompressible, pulsating and the inlet pipe diameter is d = 0.04m. For this model, the effect of the amplitude and the frequency of the flow is tested to simulate the pulsating flow coming from the engine exhaust manifold. The results show, that for a quasi-static setup the amplitude and frequency does not affect the flow. The models are also constructed numerically, where the experimental results to some extend are predicted correctly. The flow distributor is optimised with the numerical adjoint solver. The optimised flow distributor has a slightly higher flow uniformity at a lower pressure drop.