Johnson Matthey Technol. Rev., 2015, 59, (2), 152
本文还探讨了一种使用汽车或引擎数据开发模具的新方法论，这种方法能更快地制出模具。本文还展示了使用这种方法开发柴油车氧化催化剂模型能够预测铂族金属(pgm)装载量的影响。当铂族金属(pgm)装载量在30 g ft–3到120 g ft–3范围内时，这个模型很好地预测了车辆测试的一氧化碳、碳氢化合物(HC)和二氧化氮(NO2)。
Computer Simulation of Automotive Emission Control Systems
Key developments in modelling of diesel emissions control components and catalysts are highlighted
- By Mehrdad Ahmadinejad, Jonathan E. Etheridge and Timothy C. Watling*
- Johnson Matthey Technology Centre,
- Blount's Court, Sonning Common, Reading RG4 9NH, UK
- Åsa Johansson
- Johnson Matthey Emission Control Technologies,
- Västra Frölunda, SE-421 31, Sweden
- Gareth John
- Johnson Matthey Emission Control Technologies,
- Orchard Road, Royston SG8 5HE, UK
Computer simulation has become an important tool for designing automotive emission control systems. This paper highlights some of the key developments made in modelling of diesel emissions control components and catalysts by Johnson Matthey. The general methodology for model development involves determination of the reaction kinetics using laboratory reactor data, followed by validation of the resulting model against vehicle or engine data. The development of models for diesel oxidation catalysts (DOCs), ammonia selective catalytic reduction (SCR) catalysts, lean nitrogen oxides (NOx) traps (LNTs) and diesel particulate filters (DPFs), including coated filters such as the SCR coated DPF (SCRF®), is discussed.
A new methodology for developing models (or at least adapting existing models) using engine or vehicle data, which offers a faster route to a finished model, is also discussed. The use of this methodology to develop a DOC model capable of predicting the effect of platinum group metal (pgm) loading is presented; the model gives a good prediction of carbon monoxide, hydrocarbon (HC), NOx and nitrogen dioxide (NO2) over a vehicle test cycle for pgm loadings in the range 30 g ft–3 to 120 g ft–3.
Catalytic exhaust gas aftertreatment has considerably reduced the emissions from automotive sources since its introduction in the late 1970s (1–5). The progress made in developing improved aftertreatment catalysts and technologies has been complemented by the development of mathematical models (6, 7), which are used for designing aftertreatment systems and for improving understanding of the way these systems function; computer simulation offers a faster and more cost effective route for the system design than a purely experimental ‘trial and error’ process. This paper gives an overview of some of the work done at Johnson Matthey on developing models for computer simulation of aftertreatment systems.
Johnson Matthey has been working on computer simulation of emission control systems for more than 25 years. The original work used computational fluid dynamics to understand the impact on performance of flow maldistributions across the catalyst (8). Subsequent work has focused on developing models for predicting the performance of aftertreatment systems using a chemical reaction engineering approach with the major effort devoted to the development of detailed reaction kinetics (rate equations) to describe the multitude of reactions occurring over the catalyst.
Earlier work has already been reviewed in a paper published in 2007 (9); this paper will highlight some of the work done since then. This review only covers diesel aftertreatment, but we have also been active in modelling gasoline aftertreatment catalysts (9, 10). The rest of this paper is structured as follows. Firstly, our general approach to modelling will be outlined. Next the paper covers individual aftertreatment technologies, highlighting key developments. Finally, the paper will look at a faster method of developing models.
A model for an aftertreatment catalyst consists of two parts, the physical model and the kinetics (chemical model). The physical model describes the heat, mass and possibly momentum balance equations for the gas and solid (catalyst) phases, while the kinetics describes how the rate of each reaction varies with the concentration of the reactants (and other species present) and the temperature, i.e. the chemistry of the system. We use one-dimensional models, i.e. models in which concentrations and temperature vary along the length of the catalyst but not in a radial direction. Such models offer a good compromise between accuracy and runtime. Our model for a flow-through monolith catalyst is discussed in more detail elsewhere (9, 11); equations for our DPF model are given in (12).
Generally, we use laboratory reactor data for developing kinetics, as, unlike engine or vehicle tests, they allow complete control. This allows the kinetics of individual reactions to be studied separately, unlike the real system where all reactions occur together. It also allows a wide range of conditions (such as concentration and temperature) to be explored, something which is difficult to do with an engine. Reactor data allows the kinetics of a given reaction to be studied with a simple reaction mixture, enabling the basic rate equation to be defined; further experiments with more complicated gas feeds then allow extra terms to be added to the rate equation to describe additional inhibiting effects.
Once kinetics have been developed from the reactor data, it is important to validate the model against engine or vehicle test data; these models are used to predict catalyst performance over an engine or vehicle test so it is crucial to verify that they actually give a good prediction over such tests. Model validation is important to give confidence in the predictions, to understand the limitations of the model and to know where (if) improvements need to be made. The model development process is described in more detail elsewhere (9).
3. Diesel Oxidation Catalysts
A DOC serves not only to remove CO and unburnt HCs from the exhaust of diesel engines, but also to generate exotherms for active regeneration and for producing NO2 used by downstream components (13–16).
Our original DOC model (9, 17, 18) was for a Pt-based DOC and contained effectively just four reactions, CO oxidation, HC oxidation, nitric oxide (NO) oxidation and HC SCR (Equations (i)–(iv) in Table I). (Actually the model has more than four reactions as it contains more than one HC). While this model worked reasonably well under some conditions, demand for more accurate models has resulted in an increase in the number of reactions included in our DOC models. Our latest DOC model is for a platinum-palladium based DOC (19) and contains nine reactions (Table I). The reduction of NO2 to NO by CO and HC (Equations (vii) and (viii)) was added as it is required for accurate prediction of NO2, which in turn influences the performance of downstream emission control components. HC steam reforming (Equation (v)) and H2 oxidation (Equation (vi)) were also added for use in emission control systems containing an LNT as these reactions occur during the rich conditions encountered during an LNT regeneration.
Table IReaction Scheme for a DOC Model (19, 25)
|CO oxidation: |
CO + ½O2 → CO2 (i)
|Oxidation of HC or oxygenate: |
CxHyOz + (x+y/4–z/2)O2 → xCO2 + (y/2)H2O (ii)
|NO oxidation: |
NO + ½O2 ⇌ NO2 (iii)
|HC SCRa: |
(1–SN2O/2)C3H6 + 9NO → 3(1–SN2O/2)CO2 + 3(1–SN2O/2)H2O + 4½SN2O N2O + 4½(1–SN2O)N2 (iv)
|Steam reformingb: |
CxHy + (x–NWGS)H2O + NWGS CO2 → (x+NWGS)CO + (x+y/2–NWGS)H2 (v)
|H2 oxidation: |
H2 + ½O2 → H2O (vi)
|NO2 reduction by CO: |
CO + NO2 → CO2 + NO (vii)
|NO2 reduction by HC: |
CxHy + (2x+y/2)NO2 → xCO2 + (y/2)H2O + (2x+y/2)NO (viii)
|Extra reaction to explain NOx conversion during vehicle test: |
NO2 → ½N2 + O2 (ix)
aSN2O is the selectivity to N2O formation
bNWGS is the number of molecules (apparently) reacting by the reverse water gas shift reaction
Another improvement in the Pt-Pd DOC model was the inclusion of what might be called the ‘oxide effect’ on NO oxidation (19). In this effect, the active metal surface can be oxidised by NO2 to form a layer of oxide which is inactive for NO oxidation. This oxide layer can be removed by reduction with NO. This oxide effect results in suppression of NO oxidation by NO2 and leads to hysteresis when the temperature is ramped up and then down in a reactor test (19, 20). This effect has been reported by other workers (20–22). Its inclusion can be important for accurate prediction of NO2 production over a DOC. For example, a model for a heavy-duty diesel (HDD) DOC including this effect was able to give a good prediction of the NO2/NOx ratio at the DOC outlet over a wide range of engine tests (transient and steady-state engine light-off tests, the heavy-duty Federal Test Procedure (HD-FTP) transient cycle, the World Harmonised Transient Cycle and the Non-Road Transient Cycle); we found that it was not possible to get a good NO2 prediction over such a wide range of engine test cycles without including the oxide effect in the model (19).
The Pt-Pd DOC model using the reaction scheme in Table I was based on about 75 reactor tests. This demonstrates the amount of data required for developing an accurate model. This data was used not just for developing the model, but also for gaining important insights in the chemistry occurring over the catalyst (19). Figure 1 shows that this model is capable of giving a good prediction of CO, total hydrocarbon (THC) and NO2 emissions over the New European Drive Cycle (NEDC) (23). This DOC model has been extended to predict the effect of Pt:Pd ratio on catalyst performance (24, 25)
4. Ammonia Selective Catalytic Reduction
NH3 SCR is an important technology for removing NOx from vehicle exhaust. It works by reducing the NOx to (predominantly) N2 by reaction with NH3 (usually supplied in the form of an aqueous solution of urea) over a base metal catalyst (26–29).
Our original SCR model was for a vanadia-based formulation (30). Since then we have developed models for copper-zeolite (31) and iron-zeolite catalysts (32), as well as for NH3 slip catalysts which include an SCR component (33). Since our original SCR model (30), we have considerably improved our methodology for developing SCR models. Our original model was based on reactor experiments in which the gas composition was kept constant and the temperature ramped up (30). Since then we have realised that at lower temperatures it can take a long time for the amount of adsorbed NH3 and hence the rate of reaction to stabilise, particularly with metal-zeolite SCR catalysts, which store a lot more NH3 than vanadia-based catalyst. For example, we observed that it took about 70 min to achieve constant activity with a Cu-zeolite catalyst at 200°C (31). Thus, it is much better to run experiments at constant temperature in which NH3 is added to the feed gas at a known time, so that the rate at which the reaction approaches steady-state can be followed. This experiment is then repeated for a series of different temperatures and gas compositions with each experiment started with the catalyst in a stored NH3-free state (31–32). The experimental procedure for studying NH3 storage and desorption on the catalysts has also been improved, for example, by making measurements over a wider range of temperatures (31).
The reaction scheme for our Cu-zeolite SCR model is shown in Table II. Reaction schemes for vanadia-based and Fe-zeolite SCR catalysts are similar, but there are some differences, for example, reduction of nitrous oxide (N2O) by NH3 occurs over Fe-zeolite catalysts (34, 35), but not over Cu-zeolite or vanadia catalysts. The reaction scheme in Table II is similar to that used by other workers (36, 37) with the exception of the last reaction (Equation (xviii)), the reduction of NO2 to NO by NH3. This reaction was added to correct for a failure to predict the decrease in NOx conversion with increasing temperature at high temperature when the catalyst inlet NO2/NOx ratio is high, as well as the observation that most of the NOx slipped from the catalyst at high temperature is NO, even when the catalyst inlet NO2/NOx ratio is high (31). However, in a more recent model of a Cu-zeolite SCR catalyst (38) we have found that this reaction is not necessary to achieve a good prediction. This suggests that this reaction was not ‘real’, but rather served to correct for inadequacies in the kinetics of the other reactions, which became apparent at high temperature when the inlet NO2/NOx ratio was high.
Table IIReaction Scheme for a Cu-zeolite SCR Catalyst Model (31)
The Cu-zeolite SCR model was validated against a range of engine or vehicle test data (31). As with all the SCR models we have developed, we found that the prediction of post-catalyst NOx emissions was generally good, but that the accurate prediction of NH3 slip is much more challenging. Particularly disappointing was the prediction of NH3 slip over the HD-FTP where the model predicted NH3 slip, but the measured data apparently showed none. We subsequently discovered that there had been a problem with the NH3 measurements on these tests. Hence here we present further validation of the model over the HD-FTP (39) against new data (Figure 2). These HD-FTP tests were run using a 9 l engine. Three tests were run with a 20 min soak between the first and second test and no time gap between the second and third tests. These tests are denoted as the ‘cold’, ‘warm’ and ‘hot’ tests, reflecting the increase in SCR catalyst temperature from test to test. The SCR catalysts were free of adsorbed NH3 at the start of the cold test. Other experimental details are given elsewhere (31) under ‘transient HDD tests’.
The simulations of this data were run with the stored NH3 remaining on the catalyst at the end of one test still being present on the catalyst at the start of the next test (reflecting the way the engine tests were run); both the amount of stored NH3 and its distribution were unchanged. Note that these simulations were run using kinetics developed from reactor data without any adjustment (to improve the prediction of the engine data). The model gives a reasonable NOx prediction for all three tests (Figure 2), correctly predicting the fact that light-off occurs earlier in the warm and hot tests than in the cold test due to both the increase in catalyst temperature and the presence of stored NH3 at the start of the test.
The model predicts the NH3 slip to occur at the correct times in the test, although it fails to predict the magnitude of the large peak in NH3 slip at about 730 s in the warm and hot tests. In a transient test, NH3 slip occurs when a rise in catalyst temperature causes desorption of unreacted NH3 which has accumulated on the catalyst during the test. Thus, the shape of the NH3 slip curve reflects the catalyst temperature. The model correctly predicts the increase in NH3 slip from the cold test to the warm and hot tests due to build-up of stored NH3. While there is still room for improvement, Figure 2 represents an advance over the HD-FTP data previously presented (31).
5. Lean NOx Traps
The LNT is a commonly used technology for the removal of NOx from the exhaust of diesel and lean-burn gasoline engines (40–42). LNTs work by storing NOx during normal (lean) operating conditions in the form of nitrates of alkali or alkaline earth metals, or both. As the LNT fills up with NOx, the rate of storage falls. To maintain a sufficiently high rate of NOx removal from the exhaust, it is necessary to periodically regenerate the LNT by briefly running the engine rich; this results in the conversion of the nitrates to NOx, and the subsequent reduction of this NOx to (predominantly) N2 (40–42).
Our original LNT model was published in 2006 (43). This model included all the main reactions occurring over the LNT during the storage and regeneration (deNOx) phases of operation. Since then we have been focusing on improving the modelling of specific aspects of the LNT.
Improvements in the prediction of NOx storage over an LNT are illustrated in Figure 3 (44). This figure shows measured NOx breakthrough curves (points), together with model predictions (lines), for a wide range of temperatures (125°C–450°C). In these experiments, the LNT is first regenerated so that it is free of stored NOx. At time zero NOx is switched into the feed stream. Initially, all the NOx is stored and no NOx is seen at the reactor exit. However, as the NOx storage sites become occupied, the rate of NOx storage slows and NOx is seen at the reactor exit. Close examination of the breakthrough curves reveals that the initial part of the curves are the same irrespective of temperature, while the latter part of the breakthrough curves are temperature dependant. This suggested that there were two types of storage site, one giving rise to the temperature independent part of the curves and the other associated with the temperature dependant part of the curves; these sites were denoted ‘fast’ and ‘slow’ sites, the names reflecting the relative rates of NOx storage on the two sites.
In the paper three different models were considered for NOx storage on the slow sites (44). While there were some differences in the predictions of these models, all of the models tried gave a good prediction of the measured data. Figure 3 compares the predictions of one of these models with the measured data; the model gives a good prediction across a wide temperature window. From this it was concluded that key requirements for an effective NOx storage model are:
Fast storage sites, where the temperature independent storage rate and capacity are required to explain the initial part of the breakthrough curves, which is the same for all temperatures
Slow storage sites, where the temperature dependent rate of storage and capacity is required for the later part of the breakthrough curves
Kinetics for both sites need to account for the rate of storage decreasing faster than the number of available sites as storage sites become occupied.
Provided these basic criteria are met, there is no reason to use more complicated kinetics (for example microkinetics) for modelling NOx storage. It was also found to be crucial to use data covering a wide range of temperatures, so as to provide a sufficiently demanding test for potential models; many LNT modelling studies in the literature have used rather limited temperature ranges (44), which makes modelling the data easier, but makes it less likely that the resulting model will give a good prediction over the full range of conditions encountered in a real application.
6. Diesel Particulate Filters
DPFs serve to remove soot (particulate matter) from diesel engine exhaust. The collected soot is then burnt off by oxidation either with NO2 or with O2; the former tends to dominate at lower temperatures, while the latter dominates at higher temperatures (45, 46). The functionality of a DPF can be increased by adding a catalytic coating to the DPF. Thus, addition of an oxidation catalyst coating (as would be found on a DOC), results in a catalytic soot filter, which is capable of removing (oxidising) CO and THC and oxidising NO to NO2, in addition to removing soot from the exhaust. Similarly, addition of an SCR coating to the DPF results in an SCRF®, which removes NOx, in addition to soot, from the exhaust.
Our physical model for a DPF, like that of most workers, is based on the pioneering work of Bissett (47), but with improvements. The development of DPF modelling has been reviewed by Koltsakis et al. (48).
One of the challenges of DPF modelling is that soot is a variable entity, so its reactivity will vary from engine to engine or even for the same engine operating under different conditions. We have developed a method for determining soot oxidation kinetics using engine bench tests (49), which ensures that we are studying the reaction of real diesel engine soot under realistic conditions. In these experiments, the DPF was first loaded with soot on the engine under one engine condition, followed by regeneration at constant temperature under a different engine condition. During regeneration, the DPF was periodically removed from the system and weighed, so that the regeneration could be followed by observing the change in weight of soot as a function of time. This was repeated for a series of temperatures (225°C–450°C). Key experimental innovations for getting good results were: (a) the use of another DPF in front of the DPF being studied, to remove soot from the engine exhaust and avoid the need to know accurately how much soot is coming from the engine during the regeneration; and (b) the use of a bypass around the DPF being studied, to allow the engine to stabilise before exhaust gas is directed over the DPF, ensuring that the gas flowing over the DPF is as consistent as possible.
The engine data collected was used to develop kinetics for soot oxidation. Measured and simulated soot loading as a function of regeneration time are compared in Figure 4. Generally, the model gives a good prediction across the full temperature range. When looking at this figure it is important to remember that these are difficult experiments to perform and, in particular, that the measurement of a relatively small change in soot mass on a heavy DPF means that the data is subject to error. Thus, there are points where the measurement shows an apparent increase in soot loading with time; clearly this is due to experimental error and the model cannot be expected to predict this. The other major discrepancy between model prediction and measured data is that the measured weight loss in the first period of regeneration is often larger than that predicted. This has been investigated in detail (49); it was found that in the first regeneration period the weight loss was much larger than expected from the NO2 consumption, while in the second regeneration period there was good agreement. (At these temperatures, soot oxidation is predominantly due to reaction with NO2.) Thus it was concluded that much of the weight loss in the first regeneration period was due to a process other than soot oxidation by NO2; it is speculated that some of this weight loss could be due to evaporation of volatile components in the soot. The fact that the weight loss during the first regeneration period is often much greater than the weight loss in subsequent regeneration periods is consistent with this idea.
When modelling coated DPFs it is important to correctly model the transport of reactants to the catalytic coating to obtain a good prediction of post-DPF emissions (50). In our original coated DPF model, reactant molecules reached the catalyst coating solely by travelling with the bulk gas flow through the filter wall and the coating. This makes no sense when one thinks of a flow-through coated monolith catalyst where the gas flows over the catalyst along the channels, so reactants must reach the coating by diffusion through the gas in the channels. Thus, it is logical that reactant molecules in a coated DPF should reach the catalytic coating by diffusion from the gas flowing along a coated channel as well as from the gas as it passes through the filter wall (and hence the catalytic coating); both transport routes are included in our current model for a coated filter (9, 50). This results in better prediction of post-DPF emissions and more plausible predicted concentration profiles along the filter compared to our original model (50).
Moving on to SCRF® modelling, we have demonstrated that with the transport model already mentioned, it is possible to obtain a good model for an SCRF® by adding SCR kinetics developed for a flow-through monolith catalyst to the model for a coated DPF. Such a model gives a good prediction of engine or vehicle data without having to make any changes to the SCR kinetics (apart from allowing for the difference in washcoat loading). This has been demonstrated for both Cu-zeolite and Fe-zeolite formulations (12).
One point of interest with SCRF®s is the interaction between the NOx reduction (SCR) and soot oxidation functionalities; NOx removal results in a reduction in the availability of NO2 for soot oxidation, while soot oxidation by NO2 lowers the NO2/NOx ratio, which is unfavourable for SCR if it moves away from the optimum value of 0.5. Therefore the SCRF® model has been applied to investigate this (12). The presence of soot on the SCRF® is predicted to have no significant impact on NOx conversion (Figure 5). Conversely, SCR activity (NOx reduction) is predicted to significantly retard the rate of soot removal at lower temperatures (200°C–400°C), where soot oxidation is predominantly by reaction with NO2, but to have little effect at higher temperatures (450°C–550°C), where soot is predominantly oxidised by O2 (Figure 6). Both predictions are in agreement with experimental results (12). In fact, the model predicts that SCR activity actually enhances soot oxidation at 500°C and 550°C; this is due to the heat generated by the SCR reaction increasing the rate of soot oxidation with O2.
7. Faster Methods for Model Development
One of the things we have been looking at recently is faster ways of developing models, ideally using data that would be routinely collected (rather than data specially collected for model development). For example, the Pt-Pd DOC model discussed in Section 3 required 75 reactor experiments to develop the kinetics. Clearly repeating this for every catalyst formulation, or indeed for a different ageing, pgm loading or Pt:Pd ratio with the same catalyst, would represent a considerable amount of effort and time. What is needed to enable model development to keep pace with the rapid development in catalyst formulation is a faster method for adapting models to different formulations/ageings/pgm loadings (rather than developing new models from scratch each time). Thus, we have developed a methodology for adjusting/optimising existing models to best match measured engine or vehicle data. Of course this is reliant on knowing the reaction scheme for the catalyst and the form of the rate equations, so it does not entirely replace reactor testing and kinetics development, but it does mean that this does not have to be done for every formulation or variation. We have previously used this method to extend the Pt-Pd DOC model described in Section 3 to be capable of predicting the effect of changing Pt:Pd ratio at a fixed total pgm loading across the full range from monometallic Pd to monometallic Pt (25). Here we discuss the application of this method to adapt the Pt-Pd DOC model described in Section 3, to a model for a different formulation and ageing, capable of predicting the effect of pgm loading on DOC performance.
The samples used for this DOC pgm loading study were of a commercial light-duty diesel (LDD) Pt-Pd DOC with five different pgm loadings (30 g ft–3, 50 g ft–3, 70 g ft–3, 90 g ft–3, 120 g ft–3). All samples had a uniform pgm and washcoat loading, contained zeolite as a HC storage component and contained Pt and Pd in equal amounts by weight (i.e. a Pt:Pd ratio of 1:1). The same washcoat was used for all samples. The catalysts were hydrothermally aged for 15 h at 750°C. This ageing is correlated to the thermal exposure typically experienced inside the catalyst over 160,000 km of use on an LDD application where elevated temperatures are periodically encountered during DPF regeneration. The catalysts were coated on 5.66″ × 3.82″ (143.8 × 97.0 mm) (diameter × length), 400/4.3 (620,000 cells m–2, 0.11 mm wall thickness), cordierite substrates.
Engine data was collected on an engine bench dynamometer using a 2.4 l, LDD engine with a common rail fuel system, EGR and an EU5 calibration. European ultra-low sulfur diesel was used. NEDC (23) tests were run according to standard procedures. In each case, a number of repeat tests were done to ensure the data were representative. During the NEDC there was no rich operation, as would be encountered during an LNT regeneration, or unusually high HC levels for exotherm generation. Emissions were measured simultaneously before and after the DOC using standard analysers and methods. The catalyst gas inlet temperature was measured with a 1.5 mm diameter thermocouple placed 25 mm in front of the catalyst with the tip in line with the centre of the catalyst.
To be able to model the effect of pgm loading on catalytic activity, it is necessary to have a function which describes how the rate constants vary with pgm loading. In general, while catalytic activity initially increases with pgm loading, this increase in activity with pgm loading will not go on forever, but rather the increase in activity will plateau off, reflecting the tendency for the larger metal particles with lower surface area to volume ratios to form at higher loadings instead of highly dispersed small particles. We have found the following arbitrary function works well (Equation (xix)):
where k is the rate constant, kMax is the maximum or limiting value of the rate constant at high pgm loading, A is a constant and x is the pgm loading. This equation fulfils a number of key criteria for the dependency of a rate constant on pgm loading, viz. it predicts zero activity at zero pgm loading, it is single valued, it produces a continuous smooth curve without discontinuities and it predicts the rate of increase in activity with pgm loading to tail off at high pgm loadings with the rate constant eventually reaching a plateau value at high pgm loading. Ideally we would have used a more fundamental function, which relates the activity of the catalyst to the changes in metal particle morphology and dispersion as the pgm loading is increased, but this simple equation was found to work well.
More details on the optimisation method/procedure are given elsewhere (25). In the optimisation process, rate constants for CO oxidation (Equation (i)), HC oxidation (Equation (ii)), NO oxidation (Equation (iii)) and NO2 reduction by CO and HC (Equations (vii) and (viii)) were assumed to be functions of the pgm loading. The model includes a number of HCs; the rate constants for the reaction of all HCs were assumed to scale with pgm loading in the same way, i.e. the same value of A in Equation (xix) was used for the oxidation of each HC, but the value of kMax was different for each HC. Similarly, the rate constants for all the NO2 reduction reactions were assumed to scale in the same way with pgm loading. The parameter optimisation was carried out for all pgm loadings at once with the parameters in Equation (xix) being optimised, rather than optimising the kinetic parameters for each pgm loading and then fitting a function to the resulting rate constants obtained for each pgm loading, as this ensures that the rate constants vary smoothly with pgm loading. The rate constants for the main NOx reduction reaction (Equation (ix)) was also optimised, but was assumed to be independent of pgm loading as previous work suggests that the dominant NOx reduction reaction (Equation (ix)) does not occur over the pgm; we observe that an uncoated substrate gives comparable NOx reduction over the NEDC to a DOC (19).
In addition to optimising the constants controlling the variation of rate constants with pgm loading, the activation energies for CO oxidation, NO oxidation and NO2 reduction by HC were also optimised at the same time; (other activation energies were kept the same as in the original model). However, unlike the rate constants, the same activation energy for each reaction was used for all pgm loadings, i.e. it is assumed that changing the pgm loading changes the number of catalytic sites, but not the nature of the site. This may not be entirely true, but it results in a model capable of predicting the effect of pgm loading on catalytic activity. It was necessary to change the activation energies from the values in the original model (19), as this model was for a different formulation. The constant for inhibition of CO oxidation by CO was also included in the optimisation as we have previously shown that it is difficult to obtain a reliable value for this constant using only reactor light-off data (11); again the same value for this constant was used for all pgm loadings.
The variation in catalytic activity for the key reactions in the model with pgm loading is shown in Figure 7. Here the relative activity for each reaction is defined as the activity (or rate constant) of that reaction relative to that of the 120 g ft–3 sample, i.e.:
The variation in activity for NO2 reduction (Equations (vii) and (viii)) and NO oxidation (Equation (iii)) with pgm loading exhibits a pronounced curve, while the line for CO oxidation (Equation (i)) exhibits only slight curvature. In this model the variation in activity for HC oxidation was observed to be linear in pgm loading over the range of pgm loadings considered (Figure 7). However we would not expect this trend to continue ad infinitum with further increase in pgm loading.
Measured and simulated DOC outlet cumulative CO, THC and NO2 emissions over the NEDC for all pgm loadings are compared in Figure 8; predicted emissions are shown with a solid line, while the measured outlet is shown by a dashed line. Generally, the model is giving a good, although not always perfect, prediction of the measured data through the NEDC. Figure 9 compares the measured and simulated conversions over the whole NEDC. Again the model is giving a good prediction over the full range of pgm loadings from 30 g ft–3 to 120 g ft–3. It is worth noting that while the simulated conversions show a smooth variation with pgm loading, the measured conversions in some cases show a more irregular trend with pgm loading presumably due to experimental error/test-to-test variability.
As expected, CO and THC conversion increase continuously with increasing pgm loading. NO2 conversion, on the other hand, decreases with increasing pgm loading. This is because the NO2 conversion depends on the NO2 reduction reactions (which remove NO2) and NO oxidation (which produces NO2). During the urban part of the NEDC (first 800 s (23)), levels of unreacted CO and HC are high, so NO2 reduction (Equations (vii) and (viii)) dominates over NO oxidation (Equation (iii)) and the net result is conversion of NO2 to predominantly NO. During the extra-urban part of the NEDC (after 800 s (23)), temperatures are higher resulting in high CO and HC conversions which means there is little reductant available for NO2 reduction and so NO oxidation is the dominant process. In this part of the test cumulative NO2 increases with pgm loading, indicating that the rate of NO oxidation to NO2 increases with pgm loading, as expected.
This review has highlighted some of the key developments in diesel aftertreatment modelling made by Johnson Matthey. Increasingly stringent emissions legislation has created a demand for more accurate models, which in turn has resulted in models with more comprehensive reaction schemes and the need for larger experimental studies to provide the data for developing kinetics.
Demand for faster model development has led to the need for a new methodology for developing models (or at least adapting existing models) using engine or vehicle. This has been used to model the effect of Pt:Pd ratio and total pgm loading on DOC performance.
|ANR||Ammonia to NOx ratio|
|DOC||Diesel oxidation catalyst|
|DPF||Diesel particulate filter|
|EGR||Exhaust gas recirculation|
|FTP||Federal test procedure|
|HD-FTP||Heavy-duty FTP transient cycle|
|LNT||Lean NOx trap|
|NEDC||New European drive cycle|
|pgm||Platinum group metal|
|SCR||Selective catalytic reduction|
|SCRF®*||SCR coated DPF|
- K. C. Taylor, Catal. Rev.: Sci. Eng., 1993, 35, (4), 457 LINK http://dx.doi.org/10.1080/01614949308013915
- I. Chorkendorff and J. W. Niemantsverdriet, ‘Environmental Catalysis’, in “Concepts of Modern Catalysis and Kinetics”, Wiley-VCH Verlag GmbH & Co KGaA, Weinheim, Germany, 2003, pp. 377–391
- M. V. Twigg, Appl. Catal. B: Environ., 2007, 70, (1–4), 2 LINK http://dx.doi.org/10.1016/j.apcatb.2006.02.029
- H. S. Gandhi, G. W. Graham and R. W. McCabe, J. Catal., 2003, 216, (1–2), 433 LINK http://dx.doi.org/10.1016/S0021-9517(02)00067-2
- C. Morgan, Johnson Matthey Technol. Rev., 2014, 58, (4), 217 LINK https://www.technology.matthey.com/article/58/4/217-220/#
- G. B. Marin, Adv. Chem. Eng., 2007, 33, xi LINK http://dx.doi.org/10.1016/S0065-2377(07)33006-8
- O. Deutschmann, D. Chatterjee, R. E. Hayes and M. Votsmeier, Catal. Today, 2012, 188, (1), 1 LINK http://dx.doi.org/10.1016/j.cattod.2012.05.032
- N. S. Will and C. J. Bennett, ‘Flow Maldistributions in Automotive Converter Canisters and their Effect on Emission Control’, SAE Technical Paper 922339, 1992 LINK http://dx.doi.org/10.4271/922339
- M. Ahmadinejad, M. R. Desai, T. C. Watling and A. P. E. York, Adv. Chem. Eng., 2007, 33, 47 LINK http://dx.doi.org/10.1016/S0065-2377(07)33002-0
- T. C. Watling and J. P. Cox, SAE Int. J. Engines, 2014,7, (3), 1311 LINK http://dx.doi.org/10.4271/2014-01-1564
- J. E. Etheridge and T. C. Watling, Chem. Eng. J., 2015, 264, 376 LINK http://dx.doi.org/10.1016/j.cej.2014.11.089
- T. C. Watling, M. R. Ravenscroft and G. Avery, Catal. Today, 2012, 188, (1), 32 LINK http://dx.doi.org/10.1016/j.cattod.2012.02.007
- W. A. Majewski,‘Diesel Oxidation Catalyst’, Revision 2012.02a, DieselNet Technology Guide, Ecopoint Inc, Ontario, Canada, 2012 LINK http://www.dieselnet.com/tech/cat_doc.php
- W. A. Majewski, ‘Commercial DOC Technologies’, Revision 2011.02, DieselNet Technology Guide, Ecopoint Inc, Ontario, Canada, 2011 LINK http://www.dieselnet.com/tech/cat_pm.php
- D. Jollie, ‘Palladium Use in Diesel Oxidation Catalysts’, in “Platinum 2009”, Johnson Matthey Plc, Royston, UK, 2009, pp. 39–41 LINK http://www.platinum.matthey.com/documents/market-review/2009/special-feature-downloads/palladium-use-in-diesel-oxidation-catalysts-.pdf
- A. Russell and W. S. Epling, Catal. Rev.: Sci. Eng., 2011, 53, (4), 337 LINK http://dx.doi.org/10.1080/01614940.2011.596429
- A. P. E. York, J. P. Cox, T. C. Watling, A. P. Walker, D. Bergeal, R. Allansson and M. Lavenius, ‘Development and Validation of a One-Dimensional Computational Model of the Continuously Regenerating Diesel Particulate Filter (CR-DPF) System’, SAE Technical Paper 2005-01-0954 LINK http://dx.doi.org/10.4271/2005-01-0954
- A. P. E. York, M. Ahmadinejad, T. C. Watling, A. P. Walker, J. P. Cox, J. Gast, P. G. Blakeman and R. Allansson, SAE Technical Paper 2007-01-0043 LINK http://dx.doi.org/10.4271/2007-01-0043
- T. C. Watling, M. Ahmadinejad, M. Ţuţuianu, Å. Johansson and M. A. J. Paterson, SAE Int. J. Engines, 2012, 5, (3), 1420 LINK http://dx.doi.org/10.4271/2012-01-1286
- W. Hauptmann, M. Votsmeier, J. Gieshoff, A. Drochner and H. Vogel, Appl. Catal. B: Environ., 2009, 93, (1–2), 22 LINK http://dx.doi.org/10.1016/j.apcatb.2009.09.008
- L. Olsson and E. Fridell, J. Catal., 2002, 210, (2), 340 LINK http://dx.doi.org/10.1006/jcat.2002.3698
- J. Després, M. Elsener, M. Koebel, O. Kröcher, B. Schnyder and A. Wokaun, Appl. Catal. B: Environ., 2004, 50, (2), 73 LINK http://dx.doi.org/10.1016/j.apcatb.2003.12.020
- DieselNet, ‘Emission Test Cycles: ECE 15 + EUDC / NEDC’, Revision 2013.07, Emission Test Cycles, Ecopoint Inc, Ontario, Canada, 2013 LINK http://www.dieselnet.com/standards/cycles/ece_eudc.php
- B. M. Shakya, B. Sukumar, Y. M. López-De Jesús and P. Markatou, ‘The Effect of Pt:Pd Ratio on Heavy-Duty Diesel Oxidation Catalyst Performance: An Experimental and Modeling Study’, SAE Technical Paper 2015-01-1052 LINK http://dx.doi.org/10.4271/2015-01-1052
- J. E. Etheridge, T. C. Watling, A. J. Izzard and M. A. J. Paterson, ‘The Effect of Pt:Pd Ratio on Light-Duty Diesel Oxidation Catalyst Performance: An Experimental and Modelling Study’, SAE Technical Paper 2015-01-1053 LINK http://dx.doi.org/10.4271/2015-01-1053
- W. A. Majewski, ‘Selective Catalytic Reduction’, Revision 2005.05l, DieselNet Technology Guide, Ecopoint Inc, Ontario, Canada, 2005 LINK http://www.dieselnet.com/tech/cat_scr.php
- W. A. Majewski, ‘SCR Systems for Mobile Engines’, Revision 2005.05d, DieselNet Technology Guide, Ecopoint Inc, Ontario, Canada, 2005 LINK http://www.dieselnet.com/tech/cat_scr_mobile.php
- S. Brandenberger, O. Kröcher, A. Tissler and R. Althoff, Catal. Rev. Sci. Eng., 2008, 50, (4), 492 LINK http://dx.doi.org/10.1080/01614940802480122
- B. Guan, R. Zhan, H. Lin and Z. Huang, Appl. Therm. Eng., 2014, 66, (1–2), 395 LINK http://dx.doi.org/10.1016/j.applthermaleng.2014.02.021
- A. P. E. York, T. C. Watling, J. P. Cox, I. Z. Jones, A. P. Walker, P. G. Blakeman, T. Ikenhans, R. Allansson and M. Lavenius, ‘Modeling an Ammonia SCR DeNOx Catalyst: Model Development and Validation’, SAE Technical Paper 2004-01-0155 LINK http://dx.doi.org/10.4271/2004-01-0155
- T. C. Watling, M. Tutuianu, M. R. Desai, J. Dai, P. Markatou and A. Johansson, ‘Development and Validation of a Cu-Zeolite SCR Catalyst Model’, SAE Technical Paper 2011-01-1299 LINK http://dx.doi.org/10.4271/2011-01-1299
- P. Markatou, J. Dai, A. Johansson, W. Klink, M. Castagnola, T. C. Watling and M. Tutuianu, ‘Fe-Zeolite SCR Model Development, Validation and Application’, SAE Technical Paper 2011-01-1304 LINK http://dx.doi.org/10.4271/2011-01-1304
- B. Sukumar, J. Dai, A. Johansson, P. Markatou, M. Ahmadinejad, T. Watling, B. Ranganath, A. Nande and T. Szailer, ‘Modeling of Dual Layer Ammonia Slip Catalysts (ASC)’, SAE Technical Paper 2012-01-1294 LINK http://dx.doi.org/10.4271/2012-01-1294
- M. Devadas, O. Kröcher, M. Elsener, A. Wokaun, N. Söger, M. Pfeifer, Y Demel and L. Mussmann, Appl. Catal. B: Environ., 2006, 67, (3–4), 187 LINK http://dx.doi.org/10.1016/j.apcatb.2006.04.015
- M. Colombo, I. Nova, E. Tronconi, V. Schmeißer, B. Bandl-Konrad and L. Zimmermann, Appl. Catal. B: Environ., 2012, 111–112, 106 LINK http://dx.doi.org/10.1016/j.apcatb.2011.09.023
- L. Olsson, H. Sjövall and R.J. Blint, Appl. Catal. B: Environ., 2008, 81, (3–4), 203 LINK http://dx.doi.org/10.1016/j.apcatb.2007.12.011
- M. Colombo, I. Nova and E. Tronconi, Catal. Today, 2012, 197, (1), 243 LINK http://dx.doi.org/10.1016/j.cattod.2012.09.002
- P. I. Chigada, T. C. Watling, J. P. Cleeton, M. Gall and J. A. Wylie, ‘Experimental and Modelling Study of Cold Start Effects on a Cu-Zeolite SCR Catalyst’, SAE International Powertrains, Fuels and Lubricants Meeting, Kyoto, Japan, 1st–4th September, 2015, submitted
- DieselNet, ‘Emission Test Cycles: Heavy-Duty FTP Transient Cycle’, Revision 1999.03b, Emission Test Cycles, Ecopoint Inc, Ontario, Canada, 1999 LINK http://www.dieselnet.com/standards/cycles/ftp_trans.php
- W. A. Majewski, ‘NOx Adsorbers’, Revision 2007.10c, DieselNet Technology Guide, Ecopoint Inc, Ontario, Canada, 2007 LINK http://www.dieselnet.com/tech/cat_nox-trap.php
- W. S. Epling, L. E. Campell, A. Yezerets, N. W. Currier and J. E. Parks II, Catal. Rev.: Sci. Eng., 2004, 46, (2), 163 LINK http://dx.doi.org/10.1081/CR-200031932
- S. Roy and A. Baiker, Chem. Rev., 2009, 109, (9), 4054 LINK http://dx.doi.org/10.1021/cr800496f
- M. Ahmadinejad, T. C. Watling, A. P. E. York, A. P. Walker, H.-Y. Chen, H. S. Hess and J. P. Cox, ‘Development and Application of a 1-Dimensional Model for a NOx Trap System’, SAE Technical Paper 2006-01-3445 LINK http://dx.doi.org/10.4271/2006-01-3445
- T. C. Watling, P. D. Bolton and D. Swallow, Can. J. Chem. Eng., 2014, 92, (9), 1506 LINK http://dx.doi.org/10.1002/cjce.22015
- W. A. Majewski, ‘Diesel Particulate Filters’, Revision 2011.03b, DieselNet Technology Guide, Ecopoint Inc, Ontario, Canada, 2011 LINK http://www.dieselnet.com/tech/dpf.php
- W. A. Majewski, ‘Diesel Filter Regeneration’, Revision 2005.06b, DieselNet Technology Guide, Ecopoint Inc, Ontario, Canada, 2005 LINK http://www.dieselnet.com/tech/dpf_regen.php
- E. J. Bissett, Chem. Eng. Sci., 1984, 39, (7–8), 1233 LINK http://dx.doi.org/10.1016/0009-2509(84)85084-8
- G. Koltsakis, O. Haralampous, C. Depcik and J. C. Ragone, Rev. Chem. Eng., 2013, 29, (1), 1 LINK http://dx.doi.org/10.1515/revce-2012-0008
- M. Ahmadinejad, A. Tsolakis, J. M. Becker, C. F. Goersmann, A. D. Newman and T. C. Watling, SAE Int. J. Fuels Lubr., 2012, 5, (1), 359 LINK http://dx.doi.org/10.4271/2011-01-2083
- A. P. E. York, T. C. Watling, M. Ahmadinejad, D. Bergeal, P. R. Phillips and D. Swallow, SAE Int. J. Fuels Lubr., 2009, 2, (1), 578 LINK http://dx.doi.org/10.4271/2009-01-1266
Johnson Matthey Technol. Rev., 2015, 59, (3), 232
It has come to our attention that there was a mistake in the published article M. Ahmadinejad, J. E. Etheridge, T. C. Watling, Å. Johansson and G. John, Johnson Matthey Technol. Rev., 2015, 59, (2), 152
The equation should read:
Mehrdad (Mori) Ahmadinejad is a Senior Scientist in the Emissions Control Research Department at the Johnson Matthey Technology Centre (JMTC), Sonning Common, UK, where he works on the development of computer models for the simulation of vehicle emissions control catalysts.
Jon Etheridge was a Senior Scientist in the Emissions Control Research Department at JMTC, Sonning Common, where he worked on the development of computer models for the simulation of vehicle emissions control catalysts, in particular DOC and optimisation methods.
Tim Watling is a Senior Principal Scientist in the Emissions Control Research Department at JMTC, Sonning Common where he leads a team focused on the development of computer models for the simulation of vehicle emissions control systems.
Åsa Johansson is a Technical Programme Manager within the European HDD application team. She is based in Johnson Matthey Emission Control Technologies’ (ECT) Technology Centre in Gothenburg, Sweden, and supports customers with both simulations and engine measurements of HDD type catalyst systems.
Gareth John is a Senior Engineer within the European LDD application team. He is based at the ECT Technology Centre in Royston, UK, where he works on both simulation and evaluation of LDD aftertreatment systems.