A Simulation Study on Urea Maldistribution and Implications for NOx Reduction with a Multi-Channel Modelling Approach

Abstract

Urea-based selective catalytic reduction (SCR) is highly efficient for NOx abatement within a diesel aftertreatment system. However, abnormally high NOx emissions in the aftertreatment system tailpipe during WHSC (World Harmonized Steady-State Cycle) evaluation have been observed due to insufficient urea decomposition or mixing, which cannot be predicted by the current uniform 1D (one-dimensional) modelling approach with different urea dosing ratios. As a result, a multi-channel model has been developed to investigate the effect of urea maldistribution on aftertreatment system performance, where the uniformity index (UI) is used as a characteristic parameter to describe urea mixing efficiency. It was found that NOx emissions at the tailpipe can be successfully described with the multi-channel model even with a relatively high UI (UI = 0.95). Additionally, an improved segment UI factor as a function of mass flow rate has also been applied for maldistribution description, wherein better correlation with the measured NOx emission can be obtained.

1. Introduction

The emission of nitrogen oxides (NOxs) from heavy-duty diesel engines serves as a significant element leading to ambient air quality issues, ozone pollution, and the secondary formation of particulate matter (PM). Consequently, international regulatory agencies are imposing intensified, strict emission standards. Then, the removal efficiency of NOx from diesel vehicle exhaust becomes a major challenge in environmental catalysis for air pollution. Accordingly, various combustion techniques and exhaust gas aftertreatment systems have been developed to meet the increasingly stringent regulations. Among them, selective catalytic reduction (SCR) is chosen as a favourable effective method, which has been widely adopted since the mid-2000s [1].

Urea undergoes a series of complex chemical reactions within the SCR system, where the primary reactive species in the DEF (Diesel Exhaust Fluid, a 32.5% pure aqueous urea solution) is ammonia produced by hydrolysis [2]. Considering operational safety, the thermally catalyzed hydrolysis of urea solution typically occurs in the mixer located upstream of the SCR system. The specific reaction is as follows in Equation (1):

CO(NH₂)₂ + H₂O → 2NH₃ + CO₂

(1)

Thus, the current factors affecting the urea–SCR system performance include the ammonia mixing concentration distribution, exhaust temperature and flow rate, urea conversion, catalyst activity, and so on [3,4,5,6]. The urea decomposition efficiency and effective mixing distribution through the catalyst inlet could be the main challenges in the urea–SCR system. In the aftertreatment system, the process of the urea aqueous solution generating NH3 through hydrolysis and pyrolysis reactions is affected by temperature changes, gas-flow drag forces, and diffusion. Eventually, the differences in ammonia concentration distribution on the front end surface of the SCR (selective catalytic reduction) carrier constitute the problem of ammonia distribution uniformity.

Basically, the monolith reactor can be depicted as a single uniform channel. And the SCR model/design would typically demonstrate the supposition of single channel behaviour, featuring uniform concentration, flow, and temperature throughout all the channels of the monolith. However, due to the insufficient residue time, the packaging limitations, incomplete decomposition of the DEF, and poor DEF spray results, the urea/ammonia at the inlet of SCR within an aftertreatment system might not be evenly distributed in non-axial directions in practice, which may lead to a lower NOx conversion than expected, as shown in Figure 1, and ammonia slip in the outlet gas. There are several experimental and numerical studies focusing on the investigation related to the effect of maldistribution and the optimization of the design and operating conditions of injectors for achieving a more uniform ammonia profile [6,7,8,9,10,11,12]. It is generally recognized that the distribution of NH₃ at the catalyst inlet significantly affects the conversion effect of nitrogen oxides. Therefore, designing different mixer structures to reduce this non-uniformity is a widely pursued research direction at present.

Wardana et al. investigated the ammonia uniformity in the SCR system through the application of two types of urea injection devices [13]. Simultaneously, the ammonia uniformity along with urea injection time was examined to anticipate the NOx conversion reduction efficiency of the SCR system [14,15]. Song et al. employed 1-dimensional (1D) model simulations to explore the impacts of the non-uniformity of temperature, flow velocity, and the maldistribution of NH3 on the performance of the Cu-SCR system. It was ascertained that the SCR inlet NH3 maldistribution led to a 3.5% reduction in NOx reduction efficiency throughout a surrogate HD-FTP cycle [16]. McKinley et al. put forward that the enhancement of the uniformity coefficient of the flow field and gas-phase components could enhance NOx conversion efficiency and minimize ammonia leakage [17]. Dammalapati et al. utilized a CFD (computational fluid dynamics) model comprising a mixing chamber coupled with an SCR convertor to study the system, and the effects of temperature, NO:NO2 ratio, and inlet velocity were probed [18]. Despite numerous endeavours in modelling and experimental validation, a more comprehensive understanding of the influence of various non-uniformity parameters on SCR performance is still needed. An efficient approach featuring shorter calculation durations and superior prediction outcomes is still highly sought after. In this work, a multi-channel approach was adopted for predicting transient SCR performance in the 1D model, taking account of the urea maldistribution quantified by the uniformity index for realistic SCR inlet parameters, while the urea decomposition kinetic has been explored in previous work [19]. Based on the experimental results measured in the WHSC cycle, the corresponding simulation conditions were established. An approximate calculation method was also developed to investigate urea mixing efficiency on SCR system performance, introducing a better tool for further work on SCR modelling and optimization. At the same time, more accurate prediction could be obtained for the design and validation of an efficient urea injection SCR system.

2. Modelling and Methods

2.1. One-DimensionalModel Development

For the SCR model, an internal proprietary simulation tool designed by BASF, CatSim, was utilized for advanced modelling and computation, encompassing numerical models for the catalyst and pipes, cones, injectors, junctions, and exhaust gas recirculation (EGR) valves, as stated in the previous work [19]. Homogeneous reactions in the gaseous phase were taken into account, whereas non-catalytic surface reactions along the exhaust pipe were disregarded. The SCR reactions comply with the Eley–Rideal reaction mechanism, namely, gaseous nitrogen oxides react with strongly adsorbed NH₃, and the adsorption and desorption processes of NH₃ are contemplated. The SCR reactions comply with the Eley–Rideal reaction mechanism; namely, the gas-phase NOx reacts with strongly adsorbed NH₃, and the process of NH₃ adsorption and desorption is contemplated. The global kinetics form is used to create an overall kinetic model for each component and the kinetics form is expressed as an Arrhenius equation. The reaction framework for the Cu-SCR model includes the following reactions:

NH₃(g) ↔NH₃(ads)

(2)

NO + O₂ → NO₂

(3)

4NH₃(ads) + 4NO + O₂ → 4N₂ + 6H₂O

(4)

2NH₃(ads) + NO + NO₂ → 2N₂ + 3H₂O

(5)

2NH₃(ads) + 2NO₂ → N₂ + 3H₂O + N₂O

(6)

4NH₃(ads) + 3NO₂ → 3½N₂ + 6H₂O

(7)

4NH₃(ads) + 3O₂ → 2N₂ + 6H₂O

(8)

4NH₃(ads) + 5O₂ → 4NO + 6H₂O

(9)

2NH₃(ads) + 3NO₂ → 3NO + N₂ + 3H₂O

(10)

Assuming that the conditions within each individual channel in the monolith catalyst model are precisely identical, there exists no radial fluid flow velocity nor component concentration gradient in the normal section perpendicular to the flow direction. The SCR model is constructed by means of 1D single-channel approximation in the monolith catalyst, derived from the processes of each component’s reactions, the mass balance equation (Equation (11)), the energy balance equation (Equation (12)), and the momentum balance equation (Equation (13)), along with kinetic reactions on the washcoat surface [20]. The SCR reaction kinetics for both the Cu–zeolite SCR employed for calculation were developed and calibrated using testing data from a bench reactor [19].

$$\mathsf{\phi }\mathsf{\rho }{\displaystyle \frac{\mathrm{d}{\mathsf{\mu }}_{\mathrm{i}}}{\mathrm{d}\mathrm{t}}}=-{\displaystyle \frac{\mathrm{d}\left(\dot{m}{\mathsf{\mu }}_{\mathrm{i}}\right)}{\mathrm{d}\mathrm{x}}}-\mathsf{\phi }\sum _{\mathrm{j}}\mathrm{M}{\mathsf{\omega }}_{\mathrm{i},\mathrm{j}}{\mathrm{R}}_{\mathrm{H}\mathrm{j}}-{\mathrm{k}}_{\mathrm{m}\mathrm{i}}\mathrm{S}\mathsf{\rho }\left({\mathsf{\mu }}_{\mathrm{i}}-{\mathsf{\mu }}_{\mathrm{s}\mathrm{i}}\right)$$

(11)

$$\left(1-\mathsf{\phi }\right){\mathsf{\rho }}_{\mathrm{s}}{\mathrm{C}}_{\mathrm{p}\mathrm{s}}{\displaystyle \frac{\mathrm{d}{\mathrm{T}}_{\mathrm{s}}}{\mathrm{d}\mathrm{t}}}=\left(1-\mathsf{\phi }\right){\mathsf{\lambda }}_{\mathrm{s}}{\displaystyle \frac{{\mathrm{d}}^2{\mathrm{T}}_{\mathrm{s}}}{\mathrm{d}{\mathrm{x}}^2}}-\mathrm{h}\mathrm{S}\left({\mathrm{T}}_{\mathrm{s}}-\mathrm{T}\right)+\sum ∆{\mathrm{H}}_{\mathrm{R}}{\mathrm{R}}_{\mathrm{s}}$$

(12)

$${\displaystyle \frac{\mathrm{d}\dot{\mathrm{m}}}{\mathrm{d}\mathrm{t}}}=-{\displaystyle \frac{\mathrm{d}\left(\dot{m}v\right)}{\mathrm{d}\mathrm{x}}}-\mathrm{A}\mathsf{\phi }{\displaystyle \frac{\mathrm{d}\mathrm{P}}{\mathrm{d}\mathrm{x}}}-2{\displaystyle \frac{\mathrm{f}\dot{m}v}{{\mathrm{d}}_{\mathrm{h}}}}$$

(13)

where φ is porosity, ρ is the density of the gas phase, ρ_s is the density of the washcoat phase, μ_i is the mass fraction of component i in the gas phase, μ_si is the mass fraction of component i in the solid phase, $\dot{m}$ is the inlet mass flow rate, R_Hj is the reaction rate j in the gas phase, ω_i,j is the stoichiometry of component i in reaction j, Rs is the reaction rate in the solid phase, k_mi is the diffusion coefficient, S is the solid-phase specific area, C_ps is the specific heat capacity of the solid phase, T is the temperature of the gas phase, Ts is the temperature of the solid phase, h is the heat transfer coefficient, ΔH_R is the reaction enthalpy, λ_s is the thermal conductivity of the solid phase, d_h is the hydraulic diameter of the substrate channel, v is the superficial velocity, and A is the open front area.

2.2. Urea Dosing Uniformity

It is a common approach that the non-uniform flow distribution is not considered in this effective single channel model when performing larger parametric simulation, but the effect of this could not be neglected in practice [21]. For the urea maldistribution measures and expectations, either experimental measurements or CFD calculation could be used to estimate the spatial distribution of urea/ammonia across the catalyst inlet. However, a much simpler way without detailed testing or modelling could be applied, that is, the uniformity of urea distribution within a cross section is described with the uniformity index (UI) as given in Equation (6), where the X can be the concentration of urea/NH₃/HNCO, Ai is the area of each grid, X mean is the average of X within the cross section, and A is the area of the cross section. In accordance with the definition of the uniformity index, its scope lies between 0 and 1. A value that is closer to 1 indicates a superior uniformity distribution. The uniformity index at the SCR inlet typically ranges from 0.64 to −0.98. An aftertreatment system without a static mixer will have a UI value of ~0.75, and a good mixer design will result in a UI value between 0.85 and 0.9, while an excellent mixer design can help to obtain a UI value of ~0.95 or even higher. A probability distribution function, such as normal distribution or beta distribution, was used for generating urea index non-uniform input based on the uniformity index for urea dosing.

$$UI=1-{\displaystyle \frac{1}{2}}\sum _{i=1}^n{\displaystyle \frac{\left|X_i-X_{mean}\right|A_i}{A{X}_{mean}}}$$

(14)

The DOE function within BASF CatSim modelling software (GT-suite V2016 version) can be adapted to study the effect of the UI factor on SCR performance and NH₃ slip. Firstly, a DOE design is prepared, where the local urea dosing/average dosed urea can be calculated with a normal distribution function to describe a probability density distribution under different maldistribution levels (UI),

$$\mathrm{local} \mathrm{urea} \mathrm{dosing}/\mathrm{average} \mathrm{dosed} \mathrm{urea}=\mathrm{f}(\mathrm{x})={\displaystyle \frac{1}{\sqrt{2\pi }\sigma }}{e}^{-{\displaystyle \frac{{(x-\mu )}^2}{2{\sigma }^2}}}(\sigma >0)$$

(15)

In addition, parameters within the normal distribution function can be tuned to obtain the desired UI. Afterwards, DOE calculations are performed with the urea dosing flow rate factor as a DOE parameter. Finally, the calculation results can be processed based on the probability density distribution for each channel, and an average NOx emission and NH3 slip can be obtained and used for additional analysis. The calculation process is illustrated in Figure 2.

3. Results

3.1. One-DimensionalModel Calibration

An 8L engine furnished with a comprehensive aftertreatment system (DOC/CSF/SCR/AMOX) was employed to undertake the WHSC cycle test. During the experiment, a standard urea aqueous solution with a concentration of 32.5% was utilized. Nitrogen oxides were measured by an AVL i60 emission analyzer, and ammonia leakage was measured by an ammonia analyzer. The entire experimental procedure was in compliance with the stipulations of GB/T 29518-2013 [22]. The model was calibrated on the basis of the experimental outcomes of the cycle test. The temperature profiles were calibrated by a 1D model previously developed for the Cu–zeolite SCR for the ATS [19] using thermal transfer parameter tuning, and a good correlation of the measured temperature data can be found in Figure 3. Afterwards, this validated model can be used for the following deNOx performance calculation.

Initially, a simulation was performed using the existing 1D model with average urea dosing across the SCR catalyst inlet as the experimental measurements. The simulated outlet NOx and NH3 emissions do not agree very well with the experimental result, as shown in Figure 4. Several factors could be taken into consideration for this gap. Firstly, it is believed that sometimes the temperature–velocity deviation may result in worse NOx conversion performance. However, a previous study’s preliminary investigation [9] has proved that the deNOx performance of the Cu-SCR system is not significantly affected by the non-uniformity of velocity and temperature. Given the insignificant effect of velocity and temperature non-uniformity on the deNOx performance of the Cu-SCR system, and with the good correlation shown in Figure 3, the temperature–velocity deviation factor could be excluded in the following discussion for this case. Secondly, the urea dosing ratio may be below 100% due to operating system variation. As shown in Figure 5, 85%, 90%, and 95% of the urea dosing ratio were estimated and simulations were performed with the obtained cumulative NOx profile. It can be noticed that even with the compatible final NOx emission (95% urea ratio), the transient outlet profile cannot be fitted well during the whole WHTC cycle.

3.2. Multi-Channel Model Validation and Results

It is well known that the conversion efficiency of nitrogen oxides is closely related to the ammonia uniformity coefficient [23]. The multi-channel model method (20 channels) was adopted to simulate the urea flow maldistribution. The local urea dose/average urea dose level was evaluated when the UI coefficients were 0.85, 0.9, and 0.95, respectively. Figure 6 shows the probability density distribution function of the maldistribution coefficient corresponding to the three UIs. The urea distribution at the SCR inlet was obtained by applying the UI factor to 20 inlet channels. The average value of the random distribution factor was maintained at 1 to ensure mass conservation. Parameters within the normal distribution function were tuned to obtain the desired UI. Afterwards, DOE calculations were performed with the urea dosing flow rate factor as a DOE parameter for 20 channels. Finally, the calculation results were multiplied by each corresponding probability density factor for each individual channel to obtain an average NOx emission or NH₃ slip.

The simulated and experimental cumulative NOx emissions out of the ATS during the WHSC cycle are shown in Figure 7. And the traditional 1D model was equal to the model with UI = 1.0. Based on the predicted results comparison, relatively good agreement of the simulated and test results could be achieved by UI = 0.95, indicating that a relatively good mixer and urea dosing system leads to relatively uniform distribution of urea or ANR across the SCR inlet. The outlet NOx emissions were found to double when the UI dropped to 0.9 and they were found to triple with a UI of 0.85 compared to the uniform distribution condition. That is a 7% decrease in the overall deNOx efficiency (97% compared to 90%). At the same time, a higher NH₃ slip was observed due to the insufficient consumption of urea dosing flow. These results further proved that the improvement of the injected urea mixing system with exhaust flow can have an improving effect on both the enhancement of the deNOx conversion performance and the reduction of NH₃ slip. It means that the input amount of ammonia added must be controlled for an optimized urea injection dosing strategy with a higher UI level and must be compatible with the need to minimize the NOx emission, as well as to maintain the ammonia slip below the acceptable level.

However, it is noticed that even with a UI = 0.95, there is still slight gap between the experimental and simulated results. This may result from the urea maldistribution fluctuation with the cycle time and flow condition. Therefore, an improved segment UI factor was applied for maldistribution description, with UI1 corresponding to a flowrate ≥ 300 kg/h and UI2 corresponding to a flowrate < 300 kg/h. The final excellent transient simulated results were obtained as shown in Figure 7. And the corresponding calculated NOx efficiency is within 2% of the experimental data, similar to that in previous reports [16,24,25]. These results substantiate that, compared with the ideal perfect mixture with a UI equal to 1, the UI must exceed 0.95 to incur a minimum loss in conversion efficiency compared to the perfect mixture ideal (UI = 1.0), which is analogous to that suggested by Thomas [17].

4. Conclusions

To sum up, it can be concluded that in contrast to the increase in catalyst volume, copper loading, or the amount of urea dosing (ANR), a high UI could be the most cost-effective approach reflecting optimized deNOx conversion with the optimized urea dosing strategy and mixing method. Hence, engine calibration, catalyst technology, emission test cycle, and converter and mixer design optimization can subsequently be carried out with the assistance of thermal fluid analytical tools and precise monitoring devices to obtain the target ANR/ammonia/urea uniformity index, especially the dynamic UI during different cycles used to assess emission compliance

In this study, a multi-channel methodology was formulated for predicting transient SCR performance in the common 1D kinetic model throughout the WHTC cycle, in which urea maldistribution was quantified by the uniformity index with multiple channels operating in parallel to simulate the non-uniform input parameters. The final simulated results matched well with the transient data collected from engine experiments. The effect of urea maldistribution on the SCR performance and cumulative NOx emission was explored. The inlet urea maldistribution exerted a negative effect on the NOx reduction efficiency and NH₃ slip under transient circumstances. Up to a 9% enhancement in NOx reduction efficiency could be achieved by improving the uniformity of inlet ANR from 0.85 to 1.0. Probability density function (expressed as normal distribution)-based segment UI estimation demonstrated the most optimal correlation accuracy, with the corresponding calculated NOx efficiency within 2% for the transient cycle, which was efficient to operate and appropriate for parametric study.

It should be mentioned that this approach could be coupled with multi-dimensional computational fluid dynamics (CFD) modelling and a detailed comprehensive measurement apparatus to determine the maldistribution factor. The accuracy of modelling, measurement and analysis tools, and the procedure’s robustness had an impact on the prediction deviation, since relevant particulars of fluid, associated reactions, and the thermal field can be validated with a pre-determined accuracy. Thus, with the improvement of the measurement system, a better tool and more precise prediction could be provided for estimating real UI from any combination of catalyst technology, mixer design optimization, engine calibration, and emission test cycle. Further work could be conducted on SCR modelling and optimization, which could ultimately benefit the urea injection SCR system as well as aftertreatment system design to obtain the desired deNOx efficiency.