Diesel Propulsion. Anna Stefanopoulou y and Roy Smith z. University of California, Santa Barbara, CA 93106

Transcription

1 Emissions an Performance Traeoffs for Avance Marine Diesel Propulsion Anna Stefanopoulou y an Roy Smith z y Dept. of Mechanical & Environmental Engineering z Dept. of Electrical & Computer Engineering, University of California, Santa Barbara, CA 3106 Abstract: A multivariable control scheme is esigne that reuces smoke generation on an experimental marine Diesel engine equippe with a variable geometry turbocharger (v). The variable geometry turbocharger allows the improvement of the steay-state ship hyroynamic an propulsion characteristics an requires coorination with the injecte mass fuel to achieve a goo transient performance. Copyright c18 IFAC Keywors: Marine propulsion Multivariable control Emissions 1 Introuction Currently many international ports consier imposing stringent emissions requirements in oshore vehicles. In view of these new requirements, the shipping inustry must evelop technology to control emissions uring ship transient maneuvers. Although, various innovative propulsion congurations which are not base on the IC engine [10] attempt to achieve this goal, the avantages of turbocharge iesel engines rener them common practice for the vast majority of ocean freighters an oshore vehicles. In marine propulsion, the turbocharger, the engine, an the propeller operation is optimize to achieve the rate power an maximum torque output. Thus, the turbocharger size is etermine for high torque output which usually leas to large eective ow area an consequently slow response. A mechanism that can potentially mitigate the inherent performance traeo between maximum power an transient torque response is the variable geometry turbocharger. A turbine with variable geometry uses inlet guie vanes (v) on the turbine stator to moify its eective owarea as shown in Figure 1. Changing the eective turbine ow area is avantageous. First, it provies a better match between the turbocharger an the iesel engine in steay-state [5]. Seconly, it moies the power transfer to the turbine an, hence, to the compressor which can potentially reuce the turbolag by speeing up the turbocharger response [7, 8]. In this paper we assess the feasibility an poteny Corresponing author, phone: (805) 83{8501, anna@engineering.ucsb.eu. Research supporte in part by the NSF uner contract ECS matching funs were provie by Turboyne Inc. z R. Smith supporte by NSFunercontract ECS{ Exhaust Manifol N tc Engine Compressor Intake Manifol Figure 1: Schematic representation of the VGT iesel engine. tial benets of variable geometry turbocharger iesel engine for transient ship maneuvers an emission control. Several stuies in automotive applications (see [3, 6] for passenger vehicles an [7, 8] for heavy commercial vehicles) have shown promising results. The intent here is to coorinate the injecte fuel (W f ) an the variable geometry turbocharger (v) to jointly manage the fuel/air mixture into the cyliners. The multi-input multi-output (MIMO) controller is shown to improve torque an emission generation in marine iesel propulsion. In comparison to the MIMO controller we consier the conventional iesel engine governor. Such a governor controls the combustion fuel supply uring changes in loa to maintain constant engine spee,

2 an consequently, constant heaing spee. A comprehensive review of mechanical an electronic fuel governors can be foun in [4] an[1]. The actual injecte fuel is restricte by the injection elays [] an limiters that are employe to protect the shaft an prevent smoke. Smoke is generate uring fast fueling transients by rich air-to-fuel ratio mixtures. Part of the energy in the combuste fuel is store in the exhaust gas which is use by the turbocharger to spin up to a new steay-state operating point aneventu- ally increase the amount of air to the engine. Thus, changes in fueling rate cause a isturbance in air-tofuel ratio () which is partially rejecte through the natural feeback that the turbocharger establishes between the engine exhaust an intake processes. It takes, however, a certain amount of time for the turbocharger to spin up an for the intake manifol to ll to a new level of pressure an air mass, thus, reaching a new equilibrium. The slow ynamics in the engine air ow path causes a rop in the level which is associate to high hyrocarbon emissions an visible smoke generation. Any further rop in the level below the stoichiometric value aects the torque output as escribe in []. The variable geometry turbocharger provies an extra egree of freeom to (i) to spee up the air loop ynamics, an consequently, reuce smoke generation without compromising the engine torque response an (ii) to improve steay-state ship spee. The controller must achieve satisfactory traeos between conicting goals such as maneuverability, fuel economy an emissions. Designing a moel base controller requires a low orer an accurate representation of the engine an the propeller/surge ynamics. Section 3 escribes the moel for the turbocharger, engine, an surge ynamics. The steay-state optimization an plant linearization are presente in Section 4 an Section 5. The control esign an simulation results with projection on emissions benets are inclue in Section 6 an Section 7. 2 Nomenclature The variables use in this paper are ene below. Other plant constants use in the mathematical moel will be escribe in the text. Variables Super/Sub-scripts m mass (kg) 1 intake manifol W ow rate (kg/sec) 2 exhaust manifol V volume (m 3 ) f fuel T temperature (K) c compressor Q torque (Nm) t turbine N rot. spee (rpm) o ambient P power, (W) e engine force (N) p propeller u velocity (knots) r resistance 3 Plant Moel The variable geometry turbocharge iesel engine is escribe briey with the following equations. The intake an exhaust manifols are represente as nite volumes base on the \Filling an Emptying Methos" of plenum moeling escribe in [2]. The ynamic equations that characterize the manifol lling ynamics (Eq ) are base on the principles of conservation of mass, conservation of energy, an the ieal gas law. Intake Manifol Exhaust Manifol t m 1 =W c ; W e (3.1) t p 1 = R (W c T o ; W e T 1 ) V 1 (3.2) t m 2 =W e + W f ; W t (3.3) t p 2 = R ((W e + W f )T e ; W t T 2 ) V 2 (3.4) In the above, R = c p ; c v an = c p.for simplicity, cv the same values of specic heats at constant pressure, c p,anvolume, c v, are use for both intake an exhaust manifols. In the above equations the heat transfer in the walls is neglecte. This assumption is accurate for the intake manifol but it may not hol for the exhaust manifol ue to the high exhaust temperature. The air into the intake manifol is assume to be homogeneous. In aition to the above equations, the principle of conservation of momentum is also satise by assuming uniform pressure an temperature between the manifol restrictions. The right-han sie of the above equations are etermine using the engine an turbocharger manufacturer ata. In particular, engine manufacturer ata is use to escribe (i) the engine volumetric eciency, an consequently, the engine mass air ow rate, W e, (ii) the temperature rise between the engine intake an exhaust process, T e ; T 1, an (iii) the engine brake torque, Q e that is use in Eq. 3.6 below. Acceleration of the turbocharger (Eq. 3.5) an engine (Eq. 3.6)is erive using Newton's Secon Law. Rotational Dynamics t N tc = P t ; P c I tc N tc (3.5) t = Q e ; Q p I e + I a (3.6)

3 In the above, I tc an I e + I a are the mass polar moment of inertia of the rotating systems. The turbine power an the compressor power, P t an P c respectively, are calculate base on an ieal aiabatic process, an steay-state ata provie by thetur- bocharger manufacturer. In particular, a \turbine map", f t, is use to etermine the pressure ratio between the upstream an the ownstream turbine pressures, p o p 2, an the eciency, t, base on the turbine ow, W t, the turbocharger spee, N tc, an the inlet guie vanes position, v. po p 2 t =f t (W t N tc v) (3.7) P t =W t c p t m T 2 1 ; ( p o p 2 ) ( ;1 ) (3.8) Similarly, using ata from the \compressor map" we can obtain the compressor characteristics. p1 po c =f c (W c N tc ) (3.) 1 P c =W c c p T o ( p 1 ) ( ;1 ) ; 1 c p o (3.10) Finally, the propeller torque, Q p in Eq. 3.6, epens on the ship spee which is calculate base on Newton's First Law. Surge Dynamics t u = p ; r mr + m hyro (3.11) The ship isplacement an the hull wette surface ae virtual mass are enote by mr + m hyro in the equation above. The propeller torque, Q p,an thrust, p, are etermine using open water cavitation tank ata supplie by the manufacturer, k Q an k respectively, an are moie by hyroynamic interactions between hull an propeller: Q p =N 2 p D 5 pk Q (J) (3.12) p =(1 ; t)np 2 Dpk 4 (J) (3.13) u(1 ; w) J = N p D p (3.14) where, w an t are the wake fraction an thrust euction (assume to be constant), an k Q an k are the wake an thrust coecients (approximate bylinear polynomials in the avance coecient, J). The propeller spee, N p = =gr, is etermine assuming a xe gear ratio, an D p is the propeller iameter. The actuator ynamics are moele as rst orer ifferential equations. The break frequency is ene by the stable an close inner loop hyraulic controller. Actuator Dynamics t W f = 1 ( fuel ; W f ) f (3.15) t v = 1 ( vgt ; v) v (3.16) The actuation signals, fuel an vgt, are generate by the control system, an the time constants, f an v, are approximately equal to 0.1 sec. Equation , 3.11, an constitute the nine states of the ynamic plant moel, [m 1 p 1 m 2 p 2 N tc u W f v]. 4 Steay-State Optimization The ship operator comman usually involves a esire maneuver such as a suen acceleration/eceleration or maintaining constant ship spee uring wave inuce isturbances. The controller is esigne to meet this operator comman, while satisfying constraints in emissions an fuel economy. This objective imposes both steay-state an transient requirements. We analyze the steay-state requirements of the control esign goals by stuying the equilibrium points of the ynamic moel iscusse above. The moel is simulate for a range of fueling levels, fuel, an inlet guie vane positions, vgt, an the achieve equilibriums are tabulate. We note here that for each pairof( fuel vgt )values we obtain a unique equilibrium which is asymptotically stable. Figure 2 shows the steay-state ship spee, u,engine spee,, an air-to-fuel ratio,, plotte versus vgt for ierent values of fuel. Firstly, for ecient marine propulsion it is esire to choose an inlet guie vane position that maximizes ship spee for a given constant fueling level. The vgt values that achieve the maximum ship spee objective for each fueling level, from 5 kg/hr to 13 kg/hr, are shown with the otte line in the upper subplot of Figure 2. It is evient from the -subplot of Figure 2 that the steay-state engine spee an the steay-state ship spee are maximize for the same vgt values. Seconly, the emission constraints that we consier in this paper are relate to smoke generation, an more generally, to hyrocarbon (HC) an particulate emissions. Large air-to-fuel ratios,, correlate with low hyrocarbon (HC) an particulate emissions. Hence, to reuce these type of emissions it is esire to choose vgt that results in high air-to-fuel ratio, an in particular, an air-to-fuel ratio greater than 24 (limit for visible smoke generation). The vgt values that achieve the maximum for each fueling

4 level from 5 kg/hr to 13 kg/hr are connecte with the ot-ash line in the lower subplot of Figure 2. Obviously the \maximum u" an \maximum " objectives cannot be satise simultaneously an the soli line (enote by \selecte setpoints" in Figure 2) represents the vgt values for which the best traeo between fuel economy an emissions is achieve for each fueling level. The \selecte setpoints" line is also shown on the -subplot in Figure 2. Ship spee,u = Engine spee, 1800 Selecte setpoints Maximum u setpoints Selecte setpoints Maximum φ setpoints 28 Selecte setpoints = Figure 3 escribe the ynamic relation between the output scale variables, =2000 an =10, an the scale input variables, fuel =10 an vgt. The scales are chosen base on range of values for each variable an relative importance in the control objectives. The ominance of the iagonal elements of the two-input two-output system at DC inicates, as expecte, that fuel strongly aects the steay-state value of an vgt strongly aects the steay-state value of. The eects of fuel on an illustrate the severe traeo between fast engine spee tracking an tight air-to-fuel ratio control in a conventional xe geometry turbocharge iesel engine. Specically, the peak of the interaction between fuel an occurs at 5 ra/sec an causes fast fuel commans to act as a isturbance to the subsystem. Therefore, to avoi unesirable isturbances of the air-to-fuel subsystem, the close loop fuel comman must be esigne to roll o before the frequency of interaction resulting in slow torque response. The variable geometry turbocharger, on the other han, has enough authority to reject the isturbances that fuel imposes to without signicantly aecting. Hence, a twoinput two-output feeback control esign can potentially enhance fast engine spee transients without egraing emission performance. Engine spee, Engine spee, Air-to-fuel ratio limit (φ = 24) = Figure 2: Steay-state ship spee, u, engine spee,, an air-to-fuel ratio,. Using the above selecte ata an polynomial interpolation, a feeforwar controller that calculates the esire steay-state engine spee, N e, an air-to-fuel ratio,, base on the ship spee comman, u,is erive. In Section 6 a linear multivariable feeback controller is esigne that ensures tracking of the esire setpoints. Figure 3: Boe magnitue plots of the input, fuel an vgt, to the output, an at a nominal operating point. 5 Dynamic Open Loop Characteristics The frequency omain characteristics of the linearize plant are presente to better unerstan the interactions between the performance variables, an, an the actuators, fuel an vgt. Linearization is performe at a nominal point, fuel = :0 kg/hr, an vgt =0:38. Boe magnitue plots in We note here that regulation or isturbance rejection in engine spee can successfully be use to regulate ship spee or maintain constant heaing in rough seas. This argument is base on the similarity of the two spees, u an. In particular, ship spee exhibits slightly slower ynamics than the engine spee. In aition, engine spee is one of the easily measure variables, in contrast to ship spee, which reners it more suitable for control esign.

5 On the other han, ning an appropriate measurement for the air-to-fuel ratio regulation is not straight forwar [11]. An intake manifol pressure measurement is currently use for fuel limiting base on the steay-state estimation of combustion air an air-tofuel ratio. Robust an accurate estimation of transient air-to-fuel ratio will be the next stage of this work. 6 Control Designs Three control esign strategies are compare on the above moel. We assume that a measurement of the air-to-fuel ratio is available. This allows us to consier the VGT relate control authority issues inepenently of any issues that woul arise from the practical requirement of using only an estimate of the air-to-fuel ratio. The rst strategy uses a more stanar proportionalintegral (PI) esign to control the engine spee,, by actuating the fuel, fuel. This has been tune to give a goo engine spee response to commane engine spee step changes. Figure 4 illustrates the conguration. Note that this oes not allow the airto-fuel ratio to be set inepenently of the engine spee since the inlet guie vanes are hel xe at This conguration oes not require the measurement (or estimation) of the air-to-fuel ratio. u* Set-Point Map Ne * Σ PI Plant φ Figure 4: PI conguration for engine spee control It is known an will be evient in our simulations that this conguration can prouce unesirable transients in the air-to-fuel ratio. The stanar approach to alleviating this problem uses a fuel limiter when the air-to-fuel ratio rops below a specie value. This conguration is illustrate in Figure 5. In our system the limiter ajusts the commane fuel input to prevent the air-to-fuel ratio from ropping below =24. A measurement (or accurate estimation) of the air-tofuel ratio is assume. The PI controller now contains a stanar antiwinup scheme to prevent the fuel limiter from causing controller saturation inuce transients. u * N * Set-Point e Ne Fuel PI Map Σ Limiter Plant φ Figure 5: PI with fuel limiter conguration for engine spee an air-to-fuel ratio transient control The esign strategy for the variable geometry turbocharge engine (esignate as MIMO in the subsequent analysis) uses measurements of both engine spee,, an air-to-fuel ratio,, an is illustrate in Figure 6. This allows tracking of both engine spee, N e, an air-to-fuel ratio,.aweighte linear quaratic Gaussian (LQG) esign, base on the linearize moel, has been tune for both transient performance an integral type (zero steay state error) tracking of N e an. The relative weighting between the regulate variables an the actuators is the same as that use in the Boe plots in Figure 3. u * Set-Point Map * φ* MIMO Plant Figure 6: MIMO conguration for engine spee an airto-fuel ratio control 7 Simulation Results The above controllers have been evaluate on the nonlinear simulation moel. The test maneuver involves a step change in ship spee from 18.6 knots to 21.2 knots. The set-point map generates a change in N e from 1435 rpm to 1770 rpm, an a change in from 24.8 to The close-loop engine spee an air-to-fuel ratio responses are given in Figure 7. The corresponing actuation signals are shown in Figure Time (secons) Engine spee, 22 Time (secons) Figure 7: Close-loop step responses: PI (ashe), PI with fuel limit (ot-ash), an MIMO (soli) control systems The PI controller (ashe lines) gives the best engine spee response. This is achieve by a relatively fast φ

6 Time (secons) Time (secons) Figure 8: Close-loop actuation signals: PI (ashe), PI with fuel limit (ot-ash), an MIMO (soli) control systems change in fuel level which has the eect of proucing a transient air-to-fuel ratio excursion which goes signicantly below the excess smoke limit ( = 24). As expecte, the PI with fuel limiter (ot-ash lines) limits the rate of change of the fuel into the engine an prevents the air-to-fuel ratio from ropping below the smoke limit. However, it also has the eect of signicantly egraing the engine spee response. The MIMO controller (soli lines) gives an engine spee response which is close to that of the PI controller without creating a signicant air-to-fuel ratio transient. This has been achieve by using the extra egree of freeom in the VGT actuation to control air-to-fuel ratio inepenently of the fuel ow rate. The VGT actuation is evient in Figure 8. Note also that the MIMO controller uses the VGT actuator to provie a ierent steay state air-tofuel ratio at the completion of the maneuver. The steay state value of has been chosen as iscusse in Section 4 an results in the same engine spee with less fuel hence giving more ecient steay-state operation. 8 Conclusions this problem require a traeo between engine spee response an air-to-fuel ratio transient behavior. A MIMO controller must be use to take avantage of this aitional capability. References [1] Haa, S., an N. Watson, Principles an Performance of Diesel Engineering, Ellis Horwoo, Chichester, 184. [2] J.B.Heywoo, Internal Combustion Engine Funamentals, McGraw-Hill, 188. [3] Kolmanovsky, I., Moraal, P., van Nieuwstat, M., an Stefanopoulou, A., \Issues in moelling an control of intake ow in variable geometry turbocharge engines," 18th IFIP Conference on System Moelling an Optimization, Detroit, July 17. [4] Lilly, L.R., Diesel Engine Reference Book, Butterworth, Lonon, 184. [5] Mooy, J.F., \Variable geometry turbocharging with electronic control," SAE paper No , 186. [6] Stefanopoulou, A., an I. Kolmanovsky, an J.S. Freuenberg, \Control of Variable Geometry Turbocharge Diesel Engine for Reuce Emissions," Proc. 18 American Control Conference. [7] Winterborne, D.E., an S. Jai-In (11), \The application of moern control theory to a turbocharge iesel engine powerplant,", Proc. Instn. Mech. Engrs, Part I, Journal of Systems an Control Engineering, Vol. 205, pp. 6-83, IMechE, 11. [8] Watson N., an K. Banisoleiman (188), \The Variable-Geometry Turbocharger Control System for High-output Diesel Engines," SAE paper No , 188. [] Woowar, J.B., an R.G. Latorre (184), \Moeling of Diesel Engine Transient Behavior in Marine Propulsion Analysis," SNAME Transaction, Vol 2, pp [10] Valenti, M., \Luxury Liners Go Green," Mechanical Engineering, pp , July, 18. [11] van Nieuwstat, M., an I.V. Kolmanovsky, \Sensor Selection for EGR-VGT Control of a Diesel Engine," submitte International Conf. on Avances in Vehicle Control an Safety, 18. The VGT gives an extra egree of freeom in the propulsion control system which allows some amount of inepenence between engine spee an air-to-fuel ratio. This gives signicant performance avantages: in steay-state operation the air-to-fuel ratio can be tune inepenently of engine spee to improveef- ciency. In transient maneuvers the VGT can be use to prevent smoke proucing transients in the air-to-fuel ratio without signicantly egraing the engine spee response. Conventional approaches to