Reduction of Heavy Duty Diesel Engine Emission and Fuel Economy with Multi-Objective Genetic Algorithm and Phenomenological Model

Transcription

1 Reduction of Heavy Duty Diesel Engine Emission and Fuel Economy with Multi-Objective Genetic lgorithm and Phenomenological Model T. Hiroyasu, M. Miki, M. Kim Doshisha University S. Watanabe National Institute of dvanced Industrial Science and Technology H. Hiroyasu, H. Miao Kinki University Copyright c 2003 Society of utomotive Engineers, Inc. BSTRCT In this paper, the system to perform parameter search of heavy duty diesel engines is proposed. The proposed system consists of multi-objective genetic algorithm (MOG) and phenomenological model. Usually, optimization is performed to optimize only one objective. On the other hand, MOG optimizes several objectives at the same time. There is often trade off relations ship between objects, to derive the Pareto optimum solutions that express the relation ship between the objects is one of the goals in this case. MOG has strong search capability for Pareto optimum solutions. However, MOG needs a lot of iterations. Therefore, for MOG, diesel combustion simulator that can express combustion precisely with small calculation cost. In the numerical experiments, fuel injection shape, boost pressure, EGR rate, start angle of injection, duration angle of injection, and swirl ration are chosen as design variables. The values of these design variables are optimized to reduce SFC, NOx, and SOOT. Through the experiments, the following five topics are made clarified. First of all, the proposed system can find the Pareto optimum solutions successfully. Secondly, MOGs are very effective method to derive the solutions. Thirdly, phenomenological model is very suitable for MO- Gs, since it can simulate precisely with small calculation cost. Fourthly, multi step injection shape can affect the amounts of SFC, NOx, and SOOT. Finally, parameter optimization is essential for designing diesel engines. INTRODUCTION Diesel engines have considerable advantages in the aspect of engine power, fuel economy and durability. They are widely applied in the area of transport and ship propulsion and off-road applications, such as mining, construction and agriculture. In order to meet increasing environmental concerns and more stringent emission regulations, currently researches are carried out aiming the reduction of soot and nitric oxide (NOx) emissions simultaneously while maintaining reasonable fuel economy. There are several techniques to design diesel engines that have small amounts of NOx and Soot while maintaining fuel efficiency, these being multiple injection, exhaust gas recirculation (EGR) and boost pressure. However, to carry out parameter studies through experiments to find the optimum parameters, huge expenses and huge time are needed. For this reason, the optimization of parameters with the aid of computer simulation is very useful for design purposes. Efforts were carried out to solve the optimization problems related to diesel engines. To perform engine design optimization by simulation, an optimizer (which determines the next searching point) and an analyzer (which evaluates searching points) are needed. These days, Genetic lgorithm (G) is focused for optimization method. G is an algorithm that simulates the heredity and evolution of creatures[1]. t the same time, G is one of probabilistic and multi point searching methods. Therefore, G can apply for not only continuous function but also discrete functions. In designing diesel engines, normal combustion cannot be performed in some parts of design field. Thus, it is difficult to perform parameter search with conventional methods such as gradient methods. On the other hand, G can apply even in this situation. Therefore, G is suitable for parameter search in designing diesel engines. The University of Wisconsin group has researched optimization of diesel engine parameters. Montgomery and Reitz are used response surface method for optimization[2]. In the Reference [3] and [4], Gs are utilized. Many cases are treated as single objective problems. However, there are several points that should be 1

2 optimized are existed in designing diesel engines; SFC, NOx, SOOT, and so on. In this case, these several points are integrated into one object. Then, a single objective optimization method can apply to solve problem. However, it is very difficult to integrate these points into one. In our former studies, we pointed out that it is better to treat as multi-objective optimization problems[5, 6]. In multi-objective optimization problems, several objectives are optimized at the same time. For multi-objective optimization, it has also pointed out Neighborhood Cultivation G (NCG) and HIDECS are useful implementation. In this paper, target engines are heavy duty diesel engines. For these engines, the following topics are discussed. First of all, parameter optimization and multiple fuel injections are essential for designing diesel engines. Then, it is pointed out that multi-objective optimization is much useful than single objective optimization. Thirdly, G is effective for multi-objective optimization problems. Fourthly, phenomenological model is effective in optimization using Gs. Finally, the optimum results of heavy duty diesel engines are discussed. BCKGROUND The system that is proposed in this paper consists of simulator and optimizer. Several types of the models of diesel combustion exist[7] and can be used as an analyzer. Those models are roughly divided into three categories: thermodynamic models, phenomenological models and detailed multidimensional models. s the thermodynamic model only predicts the heat release rate and the calculation cost is considerably high to use detailed multi-dimensional models. When parameter search is performed by Gs, a lot of analyzer call is needed. Therefore, a model whose calculation cost is small and it can simulate precisely. In this paper, the phenomenological model is chosen as an analyzer. Since response equations are determined by the data that is derived by experiments, calculation cost is very small and it simulates combustion precisely. Phenomenological model is utilized as simulator and multi-objective genetic algorithm is chosen as optimizer. These two factors are described briefly. PHENOMENOLOGICL MODEL ND HIDECS In the past 30 year, the most sophisticated phenomenological spray-combustion model, HIDECS has shown great potential as a predictive tool for both performance and emissions in a wide range of direct injection diesel engines. It was originally developed at the University of Hiroshima and was named HIDECS recently. detailed discussion of this model, and the examples of its successful applications were given in references[8, 9, 10, 11, 12, 13, 14]. Only a brief description of the model is provided in this article. In HIDECS, the spray injected into the combustion chamber from the injection nozzle is divided into many small packages of equal fuel mass as shown in figure1. No intermixing among the packages is assumed. The spray characteristics are defined by the empirical equations of spray penetration. For example, the shaded regions shown in figure1 are the fuel packages injected at the start of injection that constitute the spray tip during penetration. ir entrainment into a package is controlled by the conservation of momentum, that is, the amount of entrained air is proportional to the decrease in package velocity. The fuel, which is mixed with the air, begins to evaporate as drops, and ignition occurs after an ignitiondelay period. * No-intermixing among the package is assumed. * Spray tip penetration is defined by the experimental equations Package pf Spray P(L, M, N) Breakup Length Spray tip penetration M L N Injected at the start of injection Figure 1: Schematic of the package distribution The air-fuel mixing processes within each package are illustrated in figure 2. Each package, immediately after the injection, involves many fine drops and a small amount of air. s a package moves away from the nozzle, air entrains into the package and the fuel drops evaporate. Thus, the package consists of liquid drops, vaporized fuel, and air. fter a short period of time after the start of injection, ignition occurs in the gaseous mixture, resulting in the rapid expansion of the package. Therefore, more fuel drops evaporate, and more fresh air entrains into the package. The vaporized fuel mixes with fresh air and combustion products as the spray continues to burn. Injection ir Entralnment Fuel Evaporation & Mixing Expansion & ir Entralnment Ignition & Evapolation Combustion Mixing & Combustion Mixing & Combustion Figure 2: Schematic of the mass system during combustion Figure 3 shows two possible combustion processes inside each package. The Case () is called evaporationrate-controlled combustion, while Case (B) is called the entrainment-rate-controlled combustion. When ignition occurs, the combustion mixture that is prepared before ignition burns in a small increment of time. The fuel-burning rate in each package is calculated by assuming stoichiometric combustion. When there is enough air in the package to burn all of the vaporized fuel, there are combustion products, liquid fuel and fresh air remaining in the package after combustion. This process is shown in Figure 3 as Case (). In the next small increment of time, more fuel drops evaporate and fresh air entrains into the package. t this point, if the amount of air in the package is 2

3 enough to burn all the vaporized fuel under stoichiometric conditions, the same combustion process (Case ) is repeated. If the amount of air is not enough to burn all the vaporized fuel, however, the fuel-burning rate is dictated by the amount of air present. This process is shown in Figure 3 as Case (B). Therefore, the combustion processes in each package always proceed under one of the conditions shown in figure 3. Injection () Ignition Combustion Controlled by ir Fuel Evaporation Rate ve Open () Complete Combustion (B) Controlled by ir Entrainment Rate (B) : Liquid Fuel : ir : Products Incomplete Combustion Figure 3: Schematic of the package combustion process The heat release rate in the combustion chamber is calculated by summing the heat releases of each package. The cylinder pressure and bulk-gas temperature in the cylinder are then calculated. Since the time history of temperature, vaporized fuel, air and combustion products in each package are known, the equilibrium concentrations of gas composition in each package can be calculated. The concentration of NOx is calculated by using the extended Zeldovich mechanism. The formation of soot is calculated by assuming first-order reaction of fuel vapor. The oxidation of carbon is calculated by assuming second-order reaction between carbon and oxygen. During the past 30 years, the code, HIDECS has been validated against wide ranges of engine rig experiments. MULTI-OBJECTIVE OPTIMIZTION PROBLEMS Problems to find design variables x that minimize or maximize k objective functions within the m constraints are called Multi-objective Optimization Problems (MOPs). MOPs can be formulated as follows[15, 16], min f( x) s.t. x X = (f 1 ( x), f 2 ( x),..., f k ( x)) T = { x R n g j ( x) 0 (j = 1,..., m) Objective functions and constraints are consisted of design variables as follows, { fi ( x) = f i (x 1, x 2,..., x n ), i = 1,..., k (2) g j ( x) = g j (x 1, x 2,..., x n ), j = 1,..., m When the objective functions are in the trade-off relationship, it is difficult to minimize or maximize all objective functions at the same time. Therefore, the concept of the Pareto optimum Solution should be introduced in this case. {bf Difinition (Dominant): x 1, x 2 R n. (1) When f i ( x 1 ) f i ( x 2 ) ( i = 1,..., k) and f i ( x 1 ) < f i ( x 2 ) ( i = 1,..., k), x 1 dominates x 2. When x 1 dominates x 2, x 1 is the better solution than x 2. Therefore, it is a good way to select non-dominant solutions. Definition (Pareto optimum solutions): x 0 R n. a) There is no solution x R n that dominates x 0, x 0 is a (strong) Pareto optimum solution. b) There is no solution x R n that satisfies f i ( x ) < f i ( x 0 )( i = 1,..., k), x 0 is a week Pareto optimum solution. Usually, there is not only one Pareto optimum solution but plural solutions in multi-objective optimization problems. In Figure4, the concept of the Pareto optimum solutions are illustrated in the case of two objectives. These objectives are minimized. f2 f1 Figure 4: The Pareto optimum solutions The value of f 1 of is better than that of B. However, f 1 of B is better than that of. Therefore, it is difficult to conclude that which is the better solution. In this figure, the line of the Pareto optimum solution is called a Pareto front. In MOPs, to find Pareto optimum solutions is one of the goals. GENETIC LGORITHMS FOR MOPS The Genetic lgorithm is an algorithm that simulates creatures heredity and evolution[1]. Since the G is one of the multi point search methods, an optimum solution can be determined even when the landscape of the objective function is multi modal. Moreover, the G can be applied to problems whose search space is discrete. Therefore, the G is one of very powerful optimization tools and is very easy to use. In multi-objective optimization, G can find a Pareto optimum set with one trial because the G is a multi point search. s a result, the G is a very effective tool especially in multi-objective optimization problems. Thus, there are many researchers who are working on the multi-objective G and there are many algorithms B 3

4 of the multi-objective G[17, 18]. These algorithms of the multi-objective G are roughly divided into two categories; those are the algorithms that treat the Pareto optimum solution implicitly or explicitly. The most of the latest methods treat the Pareto optimum solution explicitly. Typical algorithms are SPE2[19] and NSG-II[20]. In multi-objective Gs, the most remarkable feature compared to conventional Gs is setting fitness function. In multi-objective Gs, the Pareto ranking is often used for determining the fitness value[21]. The Pareto ranking is determined in the following procedure. For each solution, the number of the solution that is dominant to the focused solution is counted. The Pareto ranking is this number + 1. When the solution is non-dominant, the Pareto ranking becomes 1. The concept of the Pareto ranking is shown in figure5. In this figure R denotes the Pareto ranking. The fitness value of each individual is a reciprocal number of the Pareto ranking. f2 PROPOSED SYSTEM R=1 f1 R=3 R=4 R=2 R=1 Figure 5: Pareto Ranking The overview of the system is illustrated in figure6. s it is described before, the HIDECS is an implementation code of phenomenological model that is originally developed at the University of Hiroshima. It had previously demonstrated potential as a predictive tool for both performance and emissions in several types of the direct injection diesel engine. detailed discussion of the HIDECS spray-combustion model and some examples of its previous applications are given in Reference [8, 9, 10, 11, 12, 13, 14]. Neighborhood Cultivation Genetic lgorithm (NCG) is one of multi-objective genetic algorithms. NCG has the searching mechanism that NSG-II[20] and SPE2 [19] have. t the same time, NCG has another searching mechanism that is called neighborhood crossover. Usually, the parent individuals are chosen randomly. However, in the neighborhood crossover, adjoining individuals are chosen for parent individuals. This mechanism helps to derive distributed solutions uniformly. The precise procedures and validity of NCG are explained in the references[5, 22]. The NCG is a multi point search method. Therefore, several searching points are evaluated at the same time. For this reason, this system is very suitable for parallel processing. The system is implemented as master slave model and performed on PC cluster system. TRGET ENGINES In this paper, our HIDECS-NCG system is applied to heavy duty diesel engine. s mentioned in introduction, Reitz and his fellow researchers carried out G optimization of diesel engine parameters in Reference[2]. It would be very interesting to use the same engine as their researches by applying different G methodology. Therefore, our investigation on treating the diesel engine design as a multi-objective problem is based on the same engine as theirs. CTERPILLR 3400 SERIES The target engine is a single cylinder version of the Caterpillar 3400 series truck engine. The baseline engine operation condition was used the same as that of Reference[2]. The specification of this engine is summarized in table1. Figure 6: System construction In Figure6, the NCG is used as an optimizer and the HIDECS is used as an analyzer. Between the optimizer and analyzer, text files are exchanged. Therefore, basically several types of the Gs and analyzers can be used in this system. Table 1: Specification of Caterpillar 3400 Series Bore (m) Stroke (m) Connecting Rod (m) 0.24 Cavity (m) 0.06 EPS Compress Ratio 15.6 Nozzle Number 6 Nozzle Diameter (m) Displacement (l) 2.44 In this paper, design starts from the baseline. The specification of baseline is summarized in Table 2. HIDECS is applied to simulate this engine. The calculated and the measured in-cylinder pressure trace are com- 4

5 Table 2: Operation conditions of baseline case Engine Speed (rpm) 1737 Load (% of Maximum) 57 Start of Injection (TDC) 3.5 Injection Duration (C) 20.5 Fuel Rate (kg/hr) 6.97 Intake Temperature (Co) 32 Intake Pressure (kpa) 184 Exhaust Pressure (kpa) 181 Exhaust Pressure (kpa) 181 EGR rate 0% Rate of injection Percent of fuel in the first part Dwell between injections pared in figure 7 and show good agreements. Cylinder pressure HIDECS Measured Crank angle Figure 7: Cylinder pressure of base line design EXPERIMENTS In this paper, three experiments are performed. The design condition of the first experiment is the same as Reference[2] and the second is the same as [3]. In these experiments, the target engine is the caterpillar engine. Two step injection is applied in these experiments. In the third experiment, the caterpillar engine is also targeted, however, the three step injection is applied. Crank shaft angle Figure 9: Description of two step injection shape Table 3: Range of design variables (Experiment 1) Item Min Max bit for G Dwell between injections (angle) Percentage of first part (%) Boost Pressure (kg/cm2) EGR rate For NCG, 24 bits are used for expressing the total design variables. These bits are explained in table3. The used parameters for Gs are summarized in table4. Table 4: G parameters (Experiment 1) Population Size 200 Crossover Rate 1.0 Mutation Rate 1/bit length Terminal Generation 100 Trials 2 EXPERIMENT 1 Experiment Setup and Design Condition In this experiment, Caterpillar 3400 series Engine is applied that is explained in the former section. The base line condition is shown in table2. For this engine, fuel injection shape, boost pressure, and EGR are chosen as design variables. In this experiment, two step injection is performed. To express fuel injection shape, duration angle, dwell between injections, percent of fuel in the first part are used. This is shown in figure9. In this experiment, dwell between injections and percent of fuel in the first part are design variables. This design setting is same as Reference[2]. Therefore, there are four types of design variables. The minimum and maximum values of each design variable are described in table3. Results In figure8, the derived Pareto optimum solutions are illustrated. In this figure, the solutions are illustrated in three object space. t the same time, solutions are projected on the surface of two objectives. In these figures, point B indicates the base line design. Point is the optimum solution that is derived in Reference [2]. In this reference, the optimum solution is derived by response surface method. From these figures, it is obvious that there is trade off relation ship between SFC and NOx or Soot and SFC. On the other hand, there is no trade off relationship but linear relation ship between SFC and SOOT. It is also figured out that the base line design is far from the Pareto optimum solution. t the same time, point is one of 5

6 B B B Figure 8: Pareto Optimum Solutions (Experiment 1) Pareto optimum solutions. Therefore, the value of NOx is very small. However, the values of NOx and Soot are not so good. Since the solutions that have good values of NOx are weak Pareto optimum solutions in this case, the values of SFC are almost the same. In this condition, to determine the weights that are used to integrate several objective functions into one object is very difficult. In Reference[2], the optimum solution is derived thorough several steps. On the other hand, the Pareto optimum solutions are derived at once in this experiment. Gs have strong search capability to find Pareto optimum solutions. However, Gs need many iterations. Phenomenological model is a simulator that need not high calculation cost. Thus, by using phenomenological model, G can perform several iterations. In figures10, 11, and 12,fuel injection shapes that provide minimum values of SFC, SOOT, and NOx are illustrated. Figure 11: Injection Shape that gives minimum NOx (Experiment 1) From figure10, it is found that most of fuel should be injected at the first part to derive the minimum SFC. This is the same for the case where SOOT is minimized. To Figure 10: Injection Shape that gives minimum SFC (Experiment 1) Figure 12: Injection Shape that gives minimum SOOT (Experiment 1) 6

7 derive the minimum NOx, uniform rate of fuel should be injected during the injection duration. Like this way, by solving multi-objective optimization problems, several types of solutions are derived at the same time. This information is very useful for diesel engine designers. Calculation Time This experiment is performed on the PC cluster. Spec of PC cluster is shown in table5. Figure 14: Injection Shape that gives minimum SFC (Experiment 2) Table 5: Spec of PC Cluster Number of CPUs 64 CPU type Pentium III 933 MHz OS RedHat Linux 7.1 The total execution time to derive the Pareto optimum solution is 6602 [s] and each HIDECS call takes [s]. This calculation time is very small compared to other diesel engine combustion simulators. This is a strong characteristic of phenomenological model and this feature is fit to Gs. Figure 15: Injection Shape that gives minimum NOx (Experiment 2) EXPERIMENT 2 Experiment Setup and Design Condition In this experiment, the design conditions are the same as experiment 1, but start angle and duration angle of injections are chosen as design variables. The fuel injection shape is also double step injection. The range of design variables is summarized in table6. In this table, the bit for G to express each design variable is shown. This design condition is the same in Reference[3]. Table 6: Range of design variables (Experiment 2) Item Min Max bit for G Dwell between injections (angle) Percentage of Dwell between first part (%) Boost Pressure Dwell between (kg/cm2) EGR rate Start ngle Duration ngle The tendency of the results is also the same as Experiment 1. Most of fuel should be injected at the first part to derive the minimum SFC. To derive the minimum NOx, uniform rate of fuel should be injected during the injection duration. EXPERIMENT 3 Experiment Setup and Design Condition In this experiment, three step injection that is illustrated in figure17 is applied. Results The derived Pareto solutions are described in figure 13. The tendency of the results is the same as experiment 1. There is trade off relation ship between SFC and NOx. The fuel injection shapes that give the minimum values of SFC, NOX, and SOOT are shown in figures 14,15,16. Figure 16: Injection Shape that gives minimum SOOT (Experiment 2) 7

8 B B B Figure 13: Pareto Optimum Solutions (Experiment 2) Rate of injection duration 1 rate 1 dwell 1 duration 2 rate 2 dwell 2 duration angle Crank shaft angle Figure 17: Description of three step injection shape To express this shape, seven parameters are needed. Not only fuel injection shape but also boost pressure, EGR rate, start angle, duration angle, and swirl ratio are chosen as design variables. When the number of design variables is increased, the search space becomes bigger. This means that users have huge space to design but it is difficult to search optimum solutions. Each range of design variables are summarized in Table7. Table 7: Range of design variables (Experiment 3) Item Min Max bit for G Duration of First Injection Step Duration of Second Injection Step 6 5 Duration of Third Injection Step mount of Each Injection Step (%) 5 5 mount of Each 0.5 degree (%) 8 5 Boost Pressure (kg/cm2) EGR rate Start ngle Duration ngle Swirl Ratio CONCLUSIONS Results In figures 18, the derived Pareto optimum solutions are described. The tendency of the results is almost same as experiment 1 and 2. There is trade off relation ship between SFC and NOx. There is a linear relation ship between SFC and SOOT. Figure 19,20 and 21illustrate the optimization results which provide the minimum SFC, NOx emission and soot emission respectively. lthough three step injection shape is used as the design parameter, these results suggest that two step injection may be good enough for the diesel engine economy and emissions optimization. In this paper, multi-objective genetic algorithm (MOG) and phenomenological model are applied for parameter Figure 19: Injection Shape that gives minimum SFC (Experiment 3) 8

9 Figure 18: Pareto Optimum Solutions (Experiment 3) - NCG that is one of MOGs derived the Pareto optimum solutions successfully. Users can derive the information of the relation ship between the objective functions from the derived Pareto optimum solutions. This information is very useful for diesel engine designers. For example, there is a trade off relation ship between SFC and NOx. On the other hand, between SFC and SOOT, the relation is linear. Since the sensitivities with respect to each objective are derived, even when designers know the trade off relation ship exists between the objectives, this information is useful. Figure 20: Injection Shape that gives minimum NOx (Experiment 3) Figure 21: Injection Shape that gives minimum SOOT (Experiment 3) search of diesel engine combustion problems. The proposed system is applied to heavy duty diesel engine. Through the experiments, the following topics are made clarified. - To derive the Pareto optimum solutions by Gs, a lot of calculation iterations are necessary. Therefore, diesel engine combustion simulator must describe combustion phenomenon with small calculation cost. HIDECS that is an implementation of phenomenological model can simulates diesel engine combustion precisely. t the same time, the calculation cost of HIDECS is very small compared to the other methods. This feature is suitable for parameter search by Gs. - Two step and three step injection shapes are applied. Boost pressure, start angle and duration angle of injection, EGR, swirl ratio are also chosen as design variables. With increase of the number of design variables, the search space becomes bigger. This fact indicates that users can obtain huge design space with higher number of design variables. On the other hand, it takes a lot of calculation cost. It is found that NCG with HIDECS can treat these design variables in the experiments. CKNOWLEDGMENTS This work was supported by a grant to RCST at Doshisha University from the Ministry of Education, Science Sports and Culture, Japan. 9

10 REFERENCES [1] Goldberg, D. E., Genetic lgorithms in search, optimization and machine learning. ddison-wesly, [2] Montgomery, D. T. and Reitz, R. D., Optimization of Heavy-Duty Diesel Engine Operating Parameters Using Response Surface Method, SE Paper , [3] Senecal, P.K. and Reitz, R. D., Simultaneous Reduction of Diesel Engine Emissions and Fuel Consumption using Genetic lgorithms and Multi- Dimensional Spray and Combustion Modeling, SE Paper , [4] Yun, H. and Reitz, R. D., n Experimnetal Study on Emissions Optimization Using Micro-Genetic lgorithms in a HSDL Diesel Engine, SE Paper , [5] Hiroyasu, T., Miki, M., Kamiura, J., Watanabe, S. and Hiroyasu, H., Multi-Objective Optimization of Diesel Engine Emissions and Fuel Economy Using Genetic lgorithms and Phenomenological Model, SE paper , [6] Hiroyasu, H., Miao, H., Hiroyasu, T., Miki, M, Kamiura J. and Watanabe, S., Genetic lgorithms Optimization of Diesel Engine Emissions and Fuel Efficiency with ir Swirl, EGR, Injection Timing and Multiple Injections, SE paper , [7] Hiroyasu, H., Diesel Engine Combustion and its Modeling, International Symposium on Diagnostics and Modeling of Combustion in Reciprocating Engines. Pp.53-75, [8] Hiroyasu, H. and Kadota, T., Models for Combustion and Formation of Nitric Oxide and Soot in Direct Injection Diesel Engines, SE Paper , [9] Hiroyasu, H., Kadota, T. and rai, M., Development and Use of a Spray Combustion Modeling to Predict Diesel Engine Efficiency and Pollutant Emissions (Part 1, Combustion Modeling), Bulletin of the JSME, Vol.26, No.214, pril, [10] Hiroyasu, H., Kadota, T. and rai, M., Development and Use of a Spray Combustion Modeling to Predict Diesel Engine Efficiency and Pollutant Emissions (Part 2, Computational Procedure and Parametric Study), Bulletin of the JSME, Vol.26, No.214, pril, [11] Kuo, T. W., Evaluation of a Phenomenological Spray-Combustion Model for Two Open-Chamber Diesel Engines, SE Paper , [12] Nishida, K. and Hiroyasu, H., Simplified Three- Dimensional Modeling of Mixture Formation and Combustion in a D.I. Diesel Engine, SE Paper , [13] Yoshizaki, T.Ishida, K. and Hiroyasu, H. pproach to Low NOx and Smoke Emission Engines by Using Phenomenological Simulation, SE Paper , [14] Imanishi, H. Yoshizaki, T and Hiroyasu, H., Simulation Study of Effects of Injection Rate Profile and ir Entrainment Characteristics on D.I. Diesel Engine, SE Paper , [15] Steuer, R. E., Multiple Criteria Optimization: Theory, Computation, and pplication, Wiley, New York, [16] Ringuest, J. L., Multiobjective Optimization: Behavioral and Computational Considerations, Kluwer, Boston, [17] Cantu-Paz, E., survey of parallel genetic algorithms., Calculateurs Paralleles, 10(2), [18] Coello, C.., Handling preferences in evolutionary multiobjective optimization: survey., In 2000 Congress on Evolutionary Computation, volume 1, pages 30 37, [19] Zitzler, E., Laumanns, M. and Thiele., L. Spea2: Improving the performance of the strength Pareto evolutionary algorithm., In Technical Report 103, Computer Engineering and Communication Networks Lab (TIK), Swiss Federal Institute of Technology (ETH) Zurich, [20] Pratab,., Deb, K., grawal, S. and Meyarivan, T., fast elitist non-dominated sorting genetic algorithm for multi-objective optimization: NSG-II., In Kan- GL report , Indian Institute of Technology, Kanpur, India, [21] Fonseca. C.M, and Fleming, P. J., n Overview of Evolutionary lgorithms in Multiobjecctive Optimization, Evolutionary Computation, Vol. 3, No. 1, pp. 1-16, [22] Watanabe, S., Hiroyasu, T., and Miki, M., Multi- Objective Rectangular Packing Problem and Its pplications,proceedings of Second International Conference on Evolutionary Multi-Criterion Optimization (EMO 03), pp ,2003. PPENDIX : PHENOMENOLOGICL MODEL: HIDECS In the past 30 years, the most sophisticated phenomenological spray-combustion model, HIDECS has shown great potential as a predictive tool for both performance and emissions in a wide range of direct injection diesel engines. It was originally developed at the University of Hiroshima and was named ehidecsf recently. HIDECS is based on phenomenological model that is explained in section hoge. In this appendix, some examples of its successful applications are given in this PPENDEX. 10

11 The code, HIDECS has been validated against wide ranges of engine rig experiments. Both the in-cylinder pressure, the emissions formation and the detailed information of the diesel spray were obtained. Some of them are discussed below. EXMPLE 1 YNMR NFD 170 The in-cylinder processes of a four-stroke diesel engine, whose details are shown in table 8, were calculated. The in-cylinder pressure was measured under the operation condition in table 9. s shown in figure22, the calculated in-cylinder pressure via time trace matches well with the measured results. Table 8: Engine Dimensions Bore m Stroke m Connecting Rod Length 0.11 m Cavity Diameter m Number of Nozzle Hole 4 Nozzle Diameter 2.9E-4 m Table 9: Engine Operation Condition Intake ir Pressure KPa Intake ir Temperature 298 K Engine Speed 1800 rpm Swirl Ratio 2.2 Injection Timing -5 degree Injection Duration 18 Mass of Injected Fuel 70 mg/stroke Timing of Intake Valve Close -160 degree TDC Timing of Exhaust Valve Open 145 degree TDC tablereftab:app2-1. s shown in figurefig:app2-1, the calculated in-cylinder pressure via time trace matches well with the measured results. Figurefig:app2-2 shows that the NOx and soot emissions are also well predicted. (The original data was reported in the SE paper , pproach to low NOx and smoke emission engines by using phenomenological simulation, by Takuo Yoshizaki, Keiya Nishida and Hiroyuki Hiroyasu.) Table 10: Engine Dimensions Bore m Stroke 0.13 m Connecting Rod Length 0.15 m Cavity Diameter 0.09 m Number of Nozzle Hole 6 Nozzle Diameter 1.8E-4 m Table 11: Engine Operation Condition Intake ir Pressure KPa Intake ir Temperature 298 K Engine Speed 1500 rpm Swirl Ratio 1 Injection Timing -7 degree TDC Injection Duration 16 Mass of Injected Fuel 70 mg/stroke Timing of Intake Valve Close -145 degree TDC Timing of Exhaust Valve Open 145 degree TDC Figure 23: Comparison of the calculated and the measured Pressure and heat release Figure 22: Comparison of the calculated and the measure in-cylinder pressure trace *: For engine details and the measured results, please refer to SE paper , Effect of High Squish Combustion Chamber on Simultaneous Reduction of NOx and Particulate from a Direct-Injection Diesel Engine, by Kidoguchi, Y., Yang, C. and Miwa, K. EXMPLE 2 BORE OF 0.135M The in-cylinder processes of a four-stroke diesel engine, whose details are shown in table10, were calculated. The in-cylinder pressure was measured under the operation condition in EXMPLE 3 CTERPILLR 3400 SERIES The target engine is a single cylinder version of the Caterpillar 3400 series truck engine. The baseline engine operation condition was used the same as that of [SE paper , Optimization of Heavy-Duty Diesel Engine Operating Parameters Using Response Surface Method, by Montgomery, D. T. and Reitz, R. D.]. Engine details are shown in table12. The in-cylinder pressure was measured under the operation condition in table13. Both the baseline case and the optimization case of this paper are calculated by HIDECS. The calculated and the measured in-cylinder pressure trace are compared in figure25 and show good agreements. *: For details please refer to SE paper

12 Figure 24: Comparison of the calculated and the measured concentrations of NOx and soot Table 12: Engine Specifications Bore m Stroke m Compression Ratio 15.6 m Number of Nozzle Hole 6 Nozzle Diameter 2.14E-4 m Table 13: Operating Conditions Engine Speed 1737 rpm Load: 57% Injection Timing -3.5 to 5.5 degree TDC Fuel Rate 6.97 kg/hr Intake Temperature 32 Intake Pressure 184 kpa EGR Rate 0% EXMPLE 4 BORE OF 0.133M The in-cylinder processes of a four-stroke diesel engine, whose details are shown in table14, were calculated. The in-cylinder pressure was measured under the operation condition in table15. s shown in figure26, the calculated in-cylinder pressure via time trace matches well with the measured results. Note: The operation condition is at full load for commercial use and it is used as baseline case for genetic algorithm optimization application. Figure 25: Comparison of the calculated and the measure in-cylinder pressure trace CONTCT Table 14: Engine Specifications Bore m Stroke m Connecting Rod Length 0.26 m Cavity Diameter 0.08 m Number of Nozzle Hole 8 Nozzle Diameter 2.9E-4 m Tomoyuki Hiroyasu: Department of Knowledge Engineering and Computer Sciences, Doshisha University, 1-3 Tatara Miyakodani, Kyotanabe-shi, Kyoto, , Japan, tomo@is.doshisha.ac.jp Hiro Hiroyasu: Research Institute of Industrial Technology, Kinki University, Takaya, Umenobe, Higashi Hiroshima, , Japan. hiro@hiro.kindai.ac.jp EXMPLE 5 BORE OF M four stroke, 6- cylinder line type engine is calculated by HIDECS. The engine dimensions are shown in table16. The operation conditions of the baseline case are shown in table17. The calculated engine power output, the specific fuel consumption (S. F. C.), thermal efficiency and emissions were compared with the measured values respectively, as below: The calculation result agrees with the experiment measured values well. 12

13 Table 15: Operating Conditions Engine Speed 2000 rpm Load: 4/4 Swirl Ratio 2.6 Injection Timing -9.5 degree TDC Injection Duration 34 Mass of Injected Fuel 268 mg/stroke Timing of Intake Valve Close -150 degree TDC Timing of Exhaust Valve Open 130 degree TDC Figure 26: Comparison of the calculated and the measure in-cylinder pressure trace Table 16: Measured Result Calculated Result Measured Calculated Result Result Power (kw/cylinder) S.F.C. (g/kwh) Thermal Efficiency (%) NOx (ppm) Soot (BSU) Table 17: Engine Specifications Bore m Stroke m Connecting Rod Length m Cavity Diameter m Compression Ratio 15.1 Number of Nozzle Hole 8 Hole Diameter 10-4 m