Computer aided engineering in diesel exhaust aftertreatment systems design

Transcription

1 545 Computer aided engineering in diesel exhaust aftertreatment systems design A M Stamatelos*, G C Koltsakis, I P Kandylas and G N Pontikakis Laboratory of Applied Thermodynamics, Mechanical Engineering Department, Aristotle University, Thessaloniki, Greece Abstract: Computer aided engineering (CAE) methodologies are increasingly being applied to assist the design of spark-ignition (SI) engine exhaust aftertreatment systems in view of the stage III and IV emissions standards. Following this trend, the design of diesel exhaust aftertreatment systems is receiving more attention owing to the capabilities of recently developed mathematical models. The design of diesel exhaust systems must cope with three major aftertreatment categories: diesel oxidation catalysts, diesel particulate filters and de-no x catalytic converters. An integrated CAE methodology that could assist the design of all these classes of systems is described in this paper. It employs the following computational tools: a computer code for modelling transient exhaust system heat transfer, a computer code for modelling the transient operation of a diesel oxidation or a de-no x catalytic converter, a database containing chemical kinetics data for a variety of oxidation and de-no x catalyst formulations and a computer code for modelling the loading and regeneration behaviour of a wall-flow filter, assisted by catalytic fuel additives. Application of the CAE methodology, which helps the exhaust aftertreatment system design engineer to meet the future emissions standards, is highlighted by referring to a number of representative case studies. Keywords: diesel exhaust emissions, CAE methodology, optimization, mathematical modelling, catalytic converters, particulate traps NOTATION A i activity factor (mol K/m 2 s) c j species concentration (mol/m 3 ) CAE computer aided engineering methodology E i activation energy (J/mol) Errc normalized error in the prediction of cumulative emissions ETC European transient cycle FIGE European transient cycle FTP Federal test procedure G i inhibition factor IDI indirect injection diesel engine k permeability K a adsorption equilibrium constant m NEDC mass transfer rate (kg/s) new European driving cycle OICA Organisation Internationale des Constructeurs d Automobiles p filter back pressure (N/m 2 ) R gas constant (N m/mol K) R i reaction rate (mol/m 3 s) R&D research and development t time (s) T temperature (K) velocity (m/s) VOF volatile organic fraction w thickness of the deposit layer (m) channel wall thickness (m) w s Subscripts exhaust gas viscosity (kg/m s) The MS was recei ed on 7 September 1998 and was accepted after re ision for publication on 23 March *Corresponding author: Process, Equipment Design Laboratory, Mechanical Engineering Department, Aristotle Uni ersity, Thessaloniki, Greece. c i j m p s w computed index component of exhaust gas measured particulate layer solid, substrate wall D04998 IMechE 1999 Proc Instn Mech Engrs Vol 213 Part D

2 546 A M STAMATELOS, G C KOLTSAKIS, I P KANDYLAS AND G N PONTIKAKIS Table 1 Proposed diesel exhaust emission standards for the European Community Test cycle NEDC (g/km) for light-duty vehicle engine Test cycle OICA* (FIGE) (g/kwh) for heavy-duty engine Implementation date Particulate CO HC+NO x Particulate NO x 1996 (EURO II) 0.10 (DI) 0.9 (DI) 0.08 (IDI) (IDI) (EURO III) * 5.0* 2005 (EURO IV) (0.06)* 3.0* 1 INTRODUCTION The diesel engine has long been the most energy efficient powerplant for road transportation. Both heavy duty and passenger car diesels emit very low levels of hydrocarbons and carbon monoxide, and aftertreatment is usually not required to control these pollutants. However, it is difficult simultaneously to meet the projected standards for NO x and particulates (see, for example, the EC standards in Table 1) because of the conflict between the design changes needed to reduce each of them (particulate/no x trade-off) [1 4]. This situation represents a great challenge to the emission control engineer, who must select a combination of exhaust aftertreatment techniques and make compromises among a number of conflicting design factors in order to come up with an optimized exhaust system design that fulfils the future emissions standards (Fig. 1). Furthermore, good engine performance and fuel economy, improved filter and catalyst reliability and durability and low cost are in- Fig. 1 Computational aspects of the design of a diesel exhaust aftertreatment system

3 COMPUTER AIDED ENGINEERING IN DIESEL EXHAUST AFTERTREATMENT SYSTEMS DESIGN 547 cluded in the targets that must be simultaneously met. In this context, the support of computer aided engineering methodologies and tools is considered indispensable [5, 6]. Figure 1 is a schematic representation of the design and operating parameters involved in the process of aftertreatment optimization. The required input data and the results of the mathematical models involved are shown in this figure. In recent years, most major automotive, exhaust system and catalyst manufacturers have invested in R&D activities aimed at the development of computational tools to support such methodologies for gasoline cars. These tools comprise exhaust system heat transfer models, transient three-way catalytic converter models and supportive tuning methodologies. The situation regarding gasoline engines is discussed in detail in reference [7], which illustrates the advantages of employing CAE at the exhaust system development stage. On the other hand, a number of computational tools have been developed over the last decade to support specific design requirements for diesel aftertreatment systems. The experience gained in a variety of such applications has led to the formulation of a preliminary CAE methodology and tools to assist this job, and these are presented in this paper. A flow chart summarizing the current status of the integrated CAE methodology to assist the design of diesel exhaust aftertreatment systems is presented in Fig. 2. The methodology consists of the following basic steps: 1. Definition of a starting point in terms of a base-line car, diesel engine and exhaust system. The main geometrical, material, emissions, catalyst and filter performance and control system characteristics datasets that are required as input for the computational tools must be compiled at this initial stage. 2. Definition of certain objectives that must be met by the optimized prototype. These may refer to particulate and NO x emissions levels, along with acceptable performance, fuel consumption, durability and cost levels. 3. Definition of a number of alternative concepts that are considered as possible solutions in view of the objectives. The concepts could be based on diesel oxidation catalysts, de-no x converters, diesel particulate filters or combinations thereof. This initial definition is based on background experience in exhaust system design. This phase is also called concept analysis and results in the initial layout drawings of the exhaust system pertaining to the different concepts. 4. As a next step, a preliminary analysis is carried out with the aim of predicting the performance of the alternative systems in the legislated tests (NEDC, FTP-75, US Transient, FIGE, etc.). Especially for diesel filters, additional, custom-designed cycles could be considered in order to assess their loading and regeneration behaviour. Performance prediction is carried out by employing the exhaust system heat transfer code, the transient oxidation catalytic converter model, the lean-no x converter code and the particulate filter code. The main characteristics of these models are shown in Fig. 3. A number of kinetic constants required to model the reactions in different catalytic formulations should be accessed by a specific procedure, employing a simple set of laboratory tests and models. This procedure referred to as a tuning procedure is critical to the success of the preliminary analysis and will be discussed later. 5. Probably, the calculated predictions from the preliminary analysis will favour the adoption of just one or two from the list of possible alternative concepts. The best candidate(s) selected will deserve preliminary prototyping and testing on the engine bench or chassis dyno. 6. Based on the results of preliminary testing, along with the computational results of the preliminary analysis, the best candidates selected will be subjected to the detailed analysis phase. In this phase, more detailed computations will be employed in order to predict with high accuracy the effect of further improvements in the selected system versions, by, among other things, modification of the following design parameters: (a) washcoat technology and formulation (oxidation and lean-no x converter); (b) precious metal loading (oxidation and lean- NO x converter); (c) converter material, size, geometry and position, substrate thickness; (d) filter material, size, geometry and position; (e) fuel additive type and dosimetry; (f) exhaust manifold and downpipe layout (geometry, materials, insulation); (g) modification of fuel post-injection for lean- NO x converter (engine management). With the conclusion of the detailed analysis phase, enough data will be available to the design engineer to finalize the exhaust aftertreatment system design. In that phase, detailed drawings of the complete exhaust system are produced and an improved prototype exhaust system is built accordingly. Only this improved prototype then needs to be subjected to the extensive testing that is required in order to confirm overall system performance. This explains the significant reductions in development time and cost associated with the application of CAE methodologies. 2 COMPUTATIONAL TOOLS SUPPORTING CAE IN EXHAUST AFTERTREATMENT SYSTEMS Figure 3 presents a short description of the computational tools, that are intended to be utilized throughout D04998 IMechE 1999 Proc Instn Mech Engrs Vol 213 Part D

4 548 A M STAMATELOS, G C KOLTSAKIS, I P KANDYLAS AND G N PONTIKAKIS the computer aided design and optimization phases of the catalytic exhaust aftertreatment systems. In the same figure, the input and output data pertaining to each tool are summarized. The following tools have been employed to date: (a) a transient exhaust system heat transfer computation code; (b) a tuneable, transient oxidation catalytic converter code accompanied with an in-house catalyst kinetics database; Fig. 2 Flow chart of a computer aided engineering methodology assisting the design of diesel exhaust aftertreatment systems

5 COMPUTER AIDED ENGINEERING IN DIESEL EXHAUST AFTERTREATMENT SYSTEMS DESIGN 549 Fig. 3 Tools supporting the computer aided engineering methodology (c) a tuneable, de-no x catalytic converter code accompanied with an in-house catalyst kinetics database; (d) a computer code predicting diesel particulate filter operation (with or without catalytic assistance). The task of each tool is briefly outlined below. The transient exhaust system heat transfer code [8] allows the computation of the second-by-second evolution of exhaust gas temperatures along the exhaust line during a specified cycle (e.g. NEDC, FTP-75, etc.), based on the second-by-second evolution of exhaust gas temperature at the exhaust manifold inlet or exit. The successful determination of the exhaust temperatures is crucial for the accuracy of the oxidation and the de- NO x catalytic converter code, as well as for the fuel additive-assisted diesel particulate filter code. The heat transfer modes accounted for in this code are convective heat transfer inside the exhaust piping, axial heat D04998 IMechE 1999 Proc Instn Mech Engrs Vol 213 Part D

6 550 A M STAMATELOS, G C KOLTSAKIS, I P KANDYLAS AND G N PONTIKAKIS conduction through the pipes and natural and forced convection, as well as radiation from the outer pipe surface to the ambient. Furthermore, double-wall pipes (with an air gap or insulation material present) are also modelled by accounting for radiative, conductive and convective heat transfer through the insulation or the air gap respectively. The two-dimensional heat transfer effects induced by flanges in the exhaust system are also computed by the model. Convective heat transfer augmentations resulting from the curvatures of the exhaust piping are also taken into account, based on well-accepted semi-empirical approaches. The code has been extensively applied to simulate the transient thermal behaviour of a variety of engine exhaust systems, based on in-depth theoretical and experimental knowledge of the heat transfer coefficients involved [9 11]. The diesel oxidation catalytic converter code models the transient behaviour of the diesel oxidation catalyst in any mode of engine operation. This is actually a modification of the previously developed three-way catalytic converter model as regards the reaction scheme and the kinetics [12, 13]. The code is based on a twodimensional axisymmetric model, which approximates a real diesel oxidation catalyst (usually of an oval shape) with a cylindrical equivalent consisting of concentric rings. This approximation allows possible exhaust gas flow maldistribution effects, as well as effects of monolith heat losses and radial temperature gradients to be taken into account. In the classic work of Voltz et al. [14] extensive measurements on pellet-type Pt catalysts are processed in order to derive kinetic rate expressions for the oxidation reactions of CO and C 3 H 6 under oxygen-rich conditions. These relations are of the Langmuir Hinshelwood type and account for the inhibition of CO, C 3 H 6 and NO. In practice, Voltz et al. used a numerical integration optimization computer program to find the best combination of kinetic parameters for a given set of rate equations of this type. The abovementioned type of rate expressions has proven reliable also for the case of diesel catalysts. Of the multitude of reactions known to occur on the catalytic surface, the following reaction scheme may be considered as fundamental, as it is able to describe the basic chemistry of a typical diesel oxidation catalyst: CO+ 1 2 O 2 CO 2 (1) H O 2 2 H 2 O (2) C x H y + x+ 4 y O2 xco 2 + y 2 H 2O (fast HC) (3) C x H y + x+ 4 y O2 xco 2 + y 2 H 2O (slow HC) The problem of accounting for the kinetics of the large number of hydrocarbon species present in the diesel (4) exhaust can be handled in practice by their classification into fast (long-chain and non-saturated) and slow (small-chain and saturated) reacting hydrocarbons. The rate expressions employed in the above reactions are of the following type (for the example of the first reaction CO/O 2 ): R 1 = A 1e E 1 /RT c CO c O2 (5) G 1 (T S, c S ) with the following type of expressions accounting for the various inhibition terms: G 1 =T s (1+K a1 c CO +K a2 c C3 H 6 ) 2 (1+K a3 c 2 2 CO c C3 H 6 ) 0.7 (1+K a4 c NO ) (6) The modelling approach in matching the real-world behaviour of a diesel oxidation catalyst is supported by an original tuning procedure of the chemical kinetics submodel to the characteristics of the catalytic converter under examination. This procedure, presented in reference [7], is treated in more detail in a later section of this paper. The de-no x catalytic converter code models the transient behaviour of a diesel NO x reduction catalytic converter. The following reaction scheme is employed in the example of a Pt-based catalyst: CO+ 1 2 O 2 CO 2 (7) H O 2 2 H 2 O (8) C x H y + x+ 4 y O2 xco 2 + y 2 H 2O (9) C x H y + 2x+ y 2 NO xco2 + y 2 H 2O+ x+ y 4 N2 (10) The oxidation of CO and H 2 present in the diesel exhaust is also taken into account due to its contribution to the exothermal heat generated. Following the practice used in the diesel oxidation catalyst model, the same form of Arrhenius-type rate expressions are employed. The above reaction scheme is capable of simulating the temperature window effect observed in the de-no x converters. In the simulation of a light-off test presented in Fig. 4 the model kinetics have been tuned to match qualitatively the reaction characteristics of a Pt-based and a zeolite Cu/ZSM-5 catalyst. In both cases, the HC+NO reaction is initiated at lower temperatures than the HC+O 2 reaction. However, after a certain temperature level the rates of the HC+O 2 reaction become significantly higher, leading to the consumption of the HCs available for NO reduction. The selectivity of each catalyst formulation as regards the above reactions can thus be predicted by successful kinetic parameter tuning.

7 COMPUTER AIDED ENGINEERING IN DIESEL EXHAUST AFTERTREATMENT SYSTEMS DESIGN 551 The code may be applied in the prediction of the de-no x converter operation in full driving cycles. Here, the effect of various techniques for hydrocarbon enrichment of the exhaust gas can be assessed. For the case of vehicles equipped with a diesel particulate filter the problem of filter design optimization is quite complicated. Apart from sufficient filtration efficiency, a successful diesel filter design should combine low exhaust backpressure and regeneration safety as regards thermal damage to the filter. Here, the computer aided techniques focus on predicting: (a) filter loading and induced backpressure (b) onset and evolution of regeneration of the accumulated particulate. At present the CAE methodology does not account for filtration efficiency with high accuracy, since it is considered that the filter designs examined are all able to achieve the filtration needed to attain the emission standards. A simple pressure drop model is employed, accounting for the flow resistance through the soot layer and the filter wall, based on Darcy s law [15]. It has been shown that the following expression approximates well the total pressure drop through the loaded filter: p= k p w w+ k s w w s As regards the problem of filter regeneration, a set of related mathematical models have been developed and presented by the authors in the literature, starting from the simplest zero-dimensional approach of a wall-flow filter without catalytic aids [16]. An extended regeneration model, taking account of the catalytic activity of fuel additives, is presented in reference [17]. Recently, this model has been further extended in order to include the effect of oxidation of the volatile hydrocarbons adsorbed in the soot [18]. Adsorbed hydrocarbon oxidation is observed at low temperatures of the order of C by the catalyst oxides present in the soot originating from the catalyst mixed in the diesel fuel, but also from evaporation and desorption of the adsorbed hydrocarbons. At higher temperatures ( C), regeneration is not observed because of the evaporation and desorption of the volatile organic fraction of the particulate. When temperatures rise above C, the dry particulate can be oxidized by the catalyst oxides, and catalytic regeneration is observed. This may lead to a complete filter regeneration, because at higher temperatures the particulate is also oxidized from exhaust gas oxygen (Fig. 5). The model in its present state of development is demonstrated to predict, at least qualitatively, the lowtemperature stochastic regeneration behaviour that is observed during real-world city driving of diesel vehicles equipped with diesel filters and run on catalyst-doped fuel. Employing this model in the prediction of real-world filter operation (loading and regeneration behaviour) is already a valid tool in the hands of the system designer. Fig. 4 Computed temperature window for a Pt-based and zeolite de-no x converter in a light-off test (qualitative) D04998 IMechE 1999 Proc Instn Mech Engrs Vol 213 Part D

8 552 A M STAMATELOS, G C KOLTSAKIS, I P KANDYLAS AND G N PONTIKAKIS Fig. 5 Computational prediction of equilibrium filter regeneration ranges during steady state filter operation with catalyst-doped fuel [18] The reaction scheme takes into account carbon oxidation by exhaust gas oxygen: C+ O 2 2( 0.5)CO 2 +2(1 )CO (11) as well as catalytic carbon oxidation by the catalyst oxides (oxygen atoms exchange theory). Additionally, catalytic oxidation of adsorbed hydrocarbons by the catalyst oxides is taken into account, with significantly faster kinetics. Thus, if it is assumed that the metal additive Me forms oxides with both its three- and four-valent state, the following redox sensitive reactions take place: 2MeO 2 +C Me 2 O 3 +CO (12) Me 2 O O 2 2MeO 2 (13) CH 1, MeO Me 2 O 3 +CO+0.925H 2 O (14) The rates of the above three reactions are assumed to be Arrhenius-type functions of temperature. The need for higher accuracy in the prediction of the filter regeneration has prompted the development of a more complicated one-dimensional model, which computes the flow distribution, heat transfer and reactions across the channels of a wall-flow particulate filter [19]. This model offers the possibility of computing localized temperature peaks, as well as axial temperature gradients along the filter, which are important for filter safety under uncontrolled regenerations. 3 TUNING METHODOLOGY The development of useful catalytic converter and diesel filter models became possible only recently, by employing and integrating tunable kinetics submodels in the simulation of actual converter operation. The main obstacle in the evolution of this class of models was the absence of adequate kinetics data able to cover a significant number of reactions for the multitude of catalyst formulations and washcoats, fresh and aged, used in automotive applications. Employment of tunable kinetics expressions, which provide just the necessary and sufficient number of degrees of freedom for the simulation of real-world catalyst operation, is now becoming common practice [12]. Although the employed rate expressions in the chemical kinetics submodels are of empirical nature, the activation energies for specific categories of catalysts are in good agreement among different researchers [12, 13]. Experience from modelling and experiments indicates that a single set of activation energies gives good results for a particular family of catalysts, provided that the frequency factors are suitably adapted to the measured activity of each catalyst under consideration [20]. The determination of the unknown kinetic constants is supported by a set of simple laboratory tests with the specific catalyst formulation. A semi-empirical closedloop procedure, employing a sequence of model runs, is then followed to assess the best set of kinetic parameters [7]. The overall tuning success may be confirmed by comparing computed and experimentally measured results in a transient test, like the NEDC. In this final verification phase of the tuning procedure, the computational assessment of the catalyst performance during the full legislated cycles must give a clear

9 COMPUTER AIDED ENGINEERING IN DIESEL EXHAUST AFTERTREATMENT SYSTEMS DESIGN 553 indication of the inaccuracies involved as a function of time. For this reason, it has been found quite helpful to define the normalized error in the prediction of cumulative emissions, e.g. for HC, according to the following formula: Errc HC = t 0 t m HC, c (t) dt m HC, m (t) dt t 0 0 m HC, m (t) dt 4 APPLICATION CASE STUDY 4.1 Diesel oxidation catalyst A quite common application of CAE in the design of diesel oxidation catalyst aftertreatment systems is related to the prediction of the effect of applying different catalyst and converter systems to an existing car model with the aim of attaining lower emissions standards. The starting point for the computation in this case is the measurement of the second-by-second variation in exhaust gas temperature, mass flowrate and pollutant concentrations at, say, the exhaust manifold outlet during the legislated cycle (NEDC for Europe). Based on this input data, the pollutant concentrations at the catalytic converter exit may be computed by the model and compared with the experimental results (car fitted with oxidation catalyst, Fig. 6). In order to allow better assessment of the attained accuracy in tuning, Fig. 7 presents the evolution of the cumulative prediction error regarding HC emissions, according to the formula given above. After the necessary tuning of the model to represent the behaviour of the catalytic converter under consideration, the second-by-second variation in CO, HC and NO x conversion efficiency at the catalytic converter exit may be predicted. Based on these predictions, the capability of the system to attain the desired emissions level standards may be checked, and the effect of various design modifications may be assessed. When applying substrates of different geometry (size, cell density, wall thickness, material) and different precious metal loading, the complex interactions of the various parameters that are modified in each design version (heat capacity, thermal conductivity, mass, cell geometry, precious metal availability in a single channel, convection and mass transfer between exhaust gas and channel wall, three-dimensional effects, etc.) make the experimental assessment quite time consuming [21]. As an example of this type of application of the CAE methodology, Fig. 8 summarizes the results of an assessment of the emissions reduction potential of a diesel oxidation catalyst with size and precious metal loading variation, installed on a 1.9 l engined car. According to the results of Fig. 8, increasing the precious metal loading is effective up to a certain level. The same is true as regards the size of the converter. However, it becomes apparent that the assessment of the combined effect of size and precious metal loading optimization, which is always a difficult job for the design engineer, may be substantially assisted by the CAE methodology. Fig. 6 Diesel oxidation catalyst: comparison of computed and measured instantaneous HC and particulate conversion efficiency over NEDC D04998 IMechE 1999 Proc Instn Mech Engrs Vol 213 Part D

10 554 A M STAMATELOS, G C KOLTSAKIS, I P KANDYLAS AND G N PONTIKAKIS Fig. 7 Normalized cumulative error assisting optimal kinetic submodel tuning: diesel oxidation catalyst Fig. 8 Diesel oxidation catalyst: HC efficiency as a function of converter size for different precious metal loadings 4.2 De-NO x catalyst A possible technique for enhancing the efficiency of de-no x catalysts is to increase the HC availability in the catalyst either by post-injection using a common rail diesel engine [22] or by secondary fuel injection. The additional fuel injection should be carefully designed and controlled in order to maximize the efficiency benefits with minimum effects on fuel consumption and undesired HC breakthroughs. In the following, the CAE methodology will be briefly illustrated for such an optimization problem. Figures 9a to c present the measured [23] and computed HC and NO x efficiencies of a Pt-based catalyst in a light-off test with different degrees of hydrocarbon enrichment of the exhaust gas, corresponding to 0, 3 and 6 per cent secondary fuel injection (or post-injection) [24]. The kinetic parameters of the reaction model have been tuned to match the behaviour of the catalyst in the case of 3 per cent secondary fuel. Based on the results, it can be stated that the de-no x converter model simulates the behaviour of the real converter with sufficient accuracy for all three cases. Having evaluated the prevailing kinetic constants in the light-off test, the catalyst performance in a full driving cycle, namely the NEDC for a 1.9 l diesel passenger car, has been modelled. A series of

11 COMPUTER AIDED ENGINEERING IN DIESEL EXHAUST AFTERTREATMENT SYSTEMS DESIGN 555 Fig. 9 De-NO x catalysts: comparison between predicted and measured HC and NO x conversion efficiencies in light-off tests, for 0, 3 and 6 per cent additional fuel injection

12 556 A M STAMATELOS, G C KOLTSAKIS, I P KANDYLAS AND G N PONTIKAKIS simulations have been conducted, corresponding to different secondary fuel injection and different fuel injection patterns. In the first pattern, a constant secondary fuel flowrate is maintained during the test. In the second pattern the fuel injection is restricted only during vehicle acceleration phases, whereas in the third pattern, fuel is injected during the acceleration and cruise modes of the cycle. In all three patterns the total amount of fuel injected during the whole cycle is the same. Figure 10 presents the results of this parametric test case in terms of computed overall HC and NO x conversion efficiencies during the NEDC. Clearly, even 3 per cent of secondary fuel injection is beneficial for NO x conversion, since it doubles the efficiency compared with the case with no additional fuel. On the other hand, the HC conversion decreases dramatically, taking negative values of the order of 100 per cent, which is translated into a significant HC breakthrough. For the 3 per cent case, the best strategy seems to be the one where the secondary fuel is injected only during the vehicle acceleration with high raw NO x emissions. The above trends are generally valid also for the 6 per cent case. Here, the somewhat better NO x efficiency is accompanied with an even larger HC breakthrough. Fig. 10 De-NO x catalysts: overall conversion efficiency in NEDC for different additional fuel injection quantities and patterns Fig. 11 De-NO x +oxidation catalysts: overall conversion efficiency in NEDC for different additional fuel injection quantities

13 COMPUTER AIDED ENGINEERING IN DIESEL EXHAUST AFTERTREATMENT SYSTEMS DESIGN 557 Fig. 12 Diesel filters: measured and computed filter backpressure of a van equipped with a 4 6 in ceramic wall-flow filter in continuous repeated UDC driving as a function of the distance travelled. The filter is placed 1 m from the exhaust manifold outlet Fig. 14 Predicted backpressure as a function of distance for the 100 km UDC+short 100 km/h excursion scenario converters, which is a very frequently observed design trend. As a further example, the characteristic temperature window of high NO x conversion in the lean-no x catalytic converters necessitates the combination of close-coupled and underfloor converters with bypassing at various parts of the cycle, attaining high NO x conversion temperatures at different parts of the cycle. Such combinations may further increase NO x conversion up to 50 per cent and beyond [22]. 4.3 Diesel filters Fig. 13 Diesel filters: predicted backpressure as a function of distance travelled for different filter sizes, positioned 0 and 1 m from the exhaust manifold outlet In order to overcome the problem of breakthrough HC emissions, an oxidation catalyst can be coupled to the de-no x catalyst. Figure 11 presents the computed NO x and HC efficiencies for a de-no x +oxidation catalyst system for the case of a constant secondary fuel injection flowrate. This figure is a good example of the usefulness of CAE in the study of the emissions reduction capabilities of various combinations of catalytic The diesel filter is the most effective technology for meeting extremely stringent particulate emissions standards. However, the design of diesel exhaust aftertreatment systems based on particulate filters is a challenging problem, since a multitude of design criteria should be concurrently met. Apart from sufficient filtration efficiency, a diesel filter system should invoke minimum exhaust backpressure (and thus additional fuel consumption) and should exhibit high reliability and durability. To this end, the periodic regeneration of the accumulated soot should be well controlled. Based on the above reasoning, the evaluation of a diesel filter system requires the testing of the system in a long-duration representative driving mode of the vehicle. The selection of this driving mode depends on the expected driving conditions of the specific vehicle. In the example presented here, the tested 2.5 l engined van is expected to run in urban driving conditions. A possible driving mode to be used in diesel filter evaluation is thus the repeated legislated urban driving cycle (UDC). D04998 IMechE 1999 Proc Instn Mech Engrs Vol 213 Part D

14 558 A M STAMATELOS, G C KOLTSAKIS, I P KANDYLAS AND G N PONTIKAKIS Fig. 15 Diesel filters: temperatures inside the filter during a low space velocity regeneration scenario (a) for a highly loaded filter. Measured (b) and computed (c) evolution of exhaust gas temperatures 5, 18 and 31 cm from the filter inlet face Figure 12 presents the measured mean cycle backpressure variation of a van equipped with a 4 6 in filter in continuous UDC driving as a function of the distance travelled [25]. The backpressure is also predicted using the backpressure and the zero-dimensional catalytic regeneration model. Apparently, the results of the computation are purely deterministic, lacking the stochastic nature of real-world operation. Nevertheless, the overall effect of trap loading on mean backpressure figures is predicted in a quite satisfactory manner. In Fig. 13 the effect of filter size and positioning on the duration of the loading phase during uninterrupted UDC driving is presented. The close-coupled (placed 0 m from the exhaust manifold outlet), undersized trap regenerates after 120 km urban driving. The same trap, if positioned 1 m downstream, would first regenerate after 180 km of urban driving, with a significant deterioration in fuel penalty. Differences in trap size only slightly affect the onset of regeneration in terms of backpressure levels. However, the above-mentioned continuous urban driving scenario is not always realistic. Certain categories of vehicles follow different driving patterns which may consist of, say, 100 km continuous urban driving followed by short high-velocity driving to a depot outside the city centre. Such different scenarios may be

15 COMPUTER AIDED ENGINEERING IN DIESEL EXHAUST AFTERTREATMENT SYSTEMS DESIGN 559 readily studied by the CAE methodology and lead to different trap system optimization directions. An example is presented in Fig. 14. The approximation of uniform filter temperature introduced in the zero-dimensional model is not always valid in the regeneration modes met in the real world. In these cases, the computation of the temperature profiles along the filter requires the employment of the more advanced one-dimensional model. As an example, Fig. 15 presents the measured and computed temperatures inside the filter during low space velocity catalytic regeneration of a heavily loaded filter. This type of computation is essential in assessing the effect of filter size and positioning on the maximum levels of filter wall temperatures (filter safety considerations). Larger filters tend to regenerate with a higher thermal loading of the filter. This is due to the higher soot mass available during the onset of regeneration. For the same reason, higher filter thermal loading is shown to result from a more distant filter positioning downstream from the exhaust manifold. 5 CONCLUDING REMARKS A computer aided engineering methodology, aiding the emissions control engineer in the design of efficient and optimized diesel exhaust aftertreatment systems, is presented in this paper. This CAE methodology is based on a number of computational tools that have been developed by the authors during recent years. These tools allow the prediction of exhaust system, catalytic converter and diesel filter performance in a variety of test conditions covering the full range, starting from routine laboratory tests to real-world legislated cycles and other driving scenarios. The various design tools have been in use and continuous development since 1990 in the study of real-world diesel exhaust aftertreatment design situations, in cooperation with major automotive and catalyst manufacturers. As a result, significant know-how has been gained in this process, which is being continuously introduced for a more global and detailed understanding as well as prediction of the performance of different families of exhaust aftertreatment systems. Incorporating CAE methodologies in regular exhaust system design procedures leads to a number of advantages, starting from apparent cost and time-to-market reductions and further extending to the introduction of every new optimization procedure in an in-house database. This makes it possible promptly to investigate any extensions in the design tree, continuing from the previous branch without the need to design new minor experimental investigations that might disturb the schedule in the testing facilities. In all investigations with oxidation or de-no x catalysts, additional account must be taken of the need for experimental assessment in a variety of test conditions (light-off tests, driving cycles and parts thereof), as well as the requirements of multiple repetition of each test to attain an adequate statistical confidence level [26]. All these complexities are avoided by the employment of CAE, which brings to the fore the major trends during parametric investigations and thus guides the limited experimental effort to the most critical design situations. As shown by the representative case studies in this paper, the CAE methodology produces quick and reliable results in a variety of design situations and is continuously being improved further to enhance the design of modern exhaust aftertreatment systems. Work is already in progress in the direction of modelling other types of aftertreatment device, including NO x storage catalysts, ceramic foam filters, fiber filters, etc. REFERENCES 1 Psaras, D., Summers, J. C., Das, P. K., Ceynow, K., Khair, M. K. and DiSilveiro, W. Achieving the 2004 heavy duty diesel emissions using electronic EGR and a cerium based fuel borne catalyst. SAE paper , Walsh, M. P. Global trends in diesel emissions control a 1998 update. SAE paper , Zelenka, P., Cartellieri, W. and Herzog, P. Worldwide diesel emission standards, current experiences and future needs. Appl. Catalysis B: En iron., 1996, 10, Bauder, R. Die Zukunft der Dieselmotoren-Technologie. Motortechnische Z., 1998, 59(7/8). 5 Koltsakis, G. C. and Stamatelos, A. M. Catalytic automotive exhaust after-treatment. Prog. Energy Combust. Sci., 1997, 23, FEA HDV 2000: requirements, technical feasibility and costs of exhaust emissions standards for heavy duty vehicle engines for the year 2000 in the European Community. Federal Environmental Agency, Berlin, June Stamatelos, A. M., Koltsakis, G. C. and Kandylas, I. P. Computergestützter Entwurf von Abgasnachbehandlungskonzepten. Teil 1. Ottomotor. Motortechnische Z., 1999, 60(2), Konstantinidis, P. A., Koltsakis, G. C. and Stamatelos, A. M. Transient heat transfer modeling in automotive exhaust systems. Proc. Instn Mech. Engrs, Part C, Journal of Mechanical Engineering Science, 1997, 211(C1), Chen D. K. S. A numerical model for thermal problems in exhaust systems. SAE paper , Wendland, D. W. Automobile exhaust-system steadystate heat transfer. SAE paper , Liu, Z., Hoffmanner, A. L., Skowron, J. F. and Miller, M. J. Exhaust transient temperature response. SAE paper , Siemund, S., Leclerc, J. P., Schweich, D., Prigent, M. and Castagna, F. Three-way monolithic converter: simu- D04998 IMechE 1999 Proc Instn Mech Engrs Vol 213 Part D

16 560 A M STAMATELOS, G C KOLTSAKIS, I P KANDYLAS AND G N PONTIKAKIS lation versus experiments. Chem. Eng. Sci., 1996, 51(15), Koltsakis, G. C., Konstantinidis, P. A. and Stamatelos, A. M. Development and application range of mathematical models for 3-way catalytic converters. Appl. Catalysis B, 1997, 12, Voltz, S. E., Morgan, C. R., Liederman, D. and Jacob, S. M. Kinetic study of carbon monoxide and propylene oxidation on platinum catalysts. Ind. Eng. Chem., Prod. Res. and De., 1973, 12, Mogaka, Z., Wong, V. and Shaded, S. Performance and regeneration characteristics of a cellular ceramic diesel particulate trap. SAE paper , Koltsakis, G. C. and Stamatelos, A. M. Modeling thermal regeneration of wall-flow diesel particulate filters. AIChE J., June 1996, 42(6), Koltsakis, G. C. and Stamatelos, A. M. Modeling catalytic regeneration in wall-flow diesel filters. Ind. Eng. Chem. Res., 1996, 35, Kandylas, I. P. and Stamatelos, A. M. Modeling catalytic regeneration of diesel particulate filters, taking into account adsorbed hydrocarbon oxidation. Ind. Eng. Chem. Res., 1999, 38, Koltsakis, G. C. and Stamatelos, A. M. Modes of catalytic regeneration in diesel particulate filters. Ind. Eng. Chem. Res., 1997, 36, Koltsakis, G. C., Kandylas, I. P. and Stamatelos, A. M. Three-way catalytic converter modelling and applications. Chem. Eng. Comm., 1998, 164, Stamatelos, A. M., Koltsakis, G. C. and Kandylas, I. P. Computer aided engineering in SI engine exhaust aftertreatment systems design. FISITA World Automotive Congress, Paris, Peters, A., Langer, H. J., Jokl, B., Müller, W., Klein, H. and Ostgathe, K. Catalytic NO x reduction on a passenger car diesel common rail engine. SAE paper , Kawanami, M., Horiuchi, M., Leyrer, J., Lox, E. and Psaras, D. Advanced catalyst studies of diesel NO x reduction for on-highway trucks. SAE paper , Kandylas, I. P., Koltsakis, G. C. and Stamatelos, A. M. Mathematical modelling of precious metals catalytic converters for diesel NO x reduction. Proc. Instn Mech. Engrs, Part D, Journal of Automobile Engineering, 1999, 213(D3), Pattas, K. N., Stamatelos, A. M., Koltsakis, G. C., Kandylas, I. P. and Mustel, W. Computer aided engineering in the design of catalytically assisted trap systems. SAE paper , Kandylas, I. P., Stamatelos, A. M. and Dimitriadis, S. G. Statistical uncertainty in automotive emissions testing. Proc. Instn Mech. Engrs, Part D, Journal of Automobile Engineering, 1999, 213(D5),