An experimental and numerical analysis of pressure pulsation effects of a Gasoline Direct Injection system

Fuel 173 (2016) 8–28

Contents lists available at ScienceDirect

Fuel journal homepage: www.elsevier.com/locate/fuel

An experimental and numerical analysis of pressure pulsation effects of a Gasoline Direct Injection system Lucio Postrioti a,⇑, Andrea Cavicchi a, Domenico Paolino b, Claudio Guido b, Marco Parotto b, Rita Di Gioia b a b

Dipartimento di Ingegneria, Università di Perugia, Italy Magneti Marelli Powertrain, Italy

h i g h l i g h t s
A complete Wet-System hydraulic bench for Gasoline Injection Systems was setup.
The actual GDI injector hydraulic behavior when operating in an engine-like system layout was assessed.
The possible influence of the injection system layout on the rail pressure pulsation and injector operation was investigated.
A 1D numerical code modeling the entire injection system was tuned and applied to further complete the analysis.
The injector hydraulic behavior with advanced injection strategies (reduced dwell time) was investigated.

a r t i c l e

i n f o

Article history: Received 10 November 2015 Received in revised form 30 December 2015 Accepted 5 January 2016 Available online 11 January 2016 Keywords: GDI injection rate GDI hydraulics modeling GDI Wet System Pressure pulsation in GDI systems

a b s t r a c t Downsized, turbocharged GDI engines are considered as the most effective system architecture car makers can implement to meet stricter CO2 production and pollutant emissions regulations. Moreover, the GDI engine is accounted to be the ideal thermal part of hybrid powertrains which will play a more and more significant role to meet future CO2 and emissions standards. Hence in the last years significant research efforts are being applied to the development of GDI technology in order to optimize its performance in terms of specific fuel consumption and emission control capabilities. These engines require an extremely reliable high pressure fuel injection system to allow advanced combustion strategies and to improve the fuel atomization process and the air–fuel mixing. Nevertheless, in these installations intense fuel pressure fluctuations may occur due to continuous pumping and injection events, possibly causing low precision in the fuel metering from cylinder to cylinder and relatively poor spray quality. For this reason the injection system design must be supported by accurate computational models able to predict the actual injector flow and the whole fuel system behavior. This paper describes a combined 1-D numerical and experimental analysis of a complete GDI injection system with a particular focus on the waves propagation phenomena and their dependence on the system geometry, such as high pressure pipe length and internal diameter, rail inlet position, flow-restrictor diameter. The numerical code was validated through the comparison of the predicted results with experimental data, mainly pressure and instantaneous injected flow rate measured by a hydraulic test bench (named Wet-System) developed at SprayLab – University of Perugia, which consists of the high pressure pump, the pipes, the fuel rail and injectors so to simulate the complete injection system operation. The injection-system mathematical model was then used to predict the system dynamic response in operating conditions beyond the test bench limits, paying specific attention to the flow-restrictor effect. Finally, the model capability in accurately predicting the waves dynamics effects on the injected fuel flow rate and mass was assessed for multi-injection strategies, when the dwell time between consecutive injections is varied. Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction

⇑ Corresponding author. http://dx.doi.org/10.1016/j.fuel.2016.01.012 0016-2361/Ó 2016 Elsevier Ltd. All rights reserved.

The current scenario for the automotive powertrain development is substantially driven by the demand for a drastic reduction

9

L. Postrioti et al. / Fuel 173 (2016) 8–28

Nomenclature Definitions/abbreviations 1-D one-dimensional AC Alternating Current BDC Bottom Dead Center cad crank angle degree (deg) CoV Coefficient of Variation (%) wall compliance (Pa1) Cwall DT Dwell Time (ls) EMI Einspritz Mengeindikator (injection quantity indicator) ET Energizing Time (ls) EVI Einspritz Verlaufindikator (injection curve indicator) FMV Fuel Metering Valve

of the specific CO2 production posed by legislation, along with the on-going decreasing trend for the allowed pollutant emission levels. In this frame, the most effective way to improve the sparkignition engine thermal efficiency is the shift from the homogeneous charge, in-direct injection scheme to the stratified charge, direct injection combustion system. Hence, Gasoline Direct Injection (GDI) engines are considerably increasing their market share, often in combination with turbo-charging of small to medium displacement units, in order to exploit the potential of such configured engines to improve thermal efficiency in key operating conditions such as low load and cold start. Further, GDI technology applied to downsized and turbocharged engines represents today the most attractive technological solution for the thermal part of hybrid powertrains. The quick development of direct injection in spark ignition engines implies a higher importance of the injection system performances in terms of combustion and emissions control. In fact, along with the theoretical advantages, the development of stratified charge, direct injection systems implies a number of issues that must be faced. The fuel metering accuracy, the rate of fuel introduction in the combustion chamber and the consequent spray evolution, fuel atomization and mixing with surrounding air directly control the combustion development. Hence the engine performance and efficiency along with its pollutant emissions and noise are directly affected by the injection system in a wide range of operating conditions. As a consequence, a deep knowledge of the actual injection rate is nowadays crucial for designing new generation injection systems, particularly in the operating conditions in which advanced injection strategies are to be applied to govern the combustion process. A typical direct injection system consists of a low pressure pump feeding a cam-actuated high pressure pump, a Fuel Metering Valve (FMV) acting as injection pressure regulation system, a pipe connecting the pump to a common rail and the injectors delivering the fuel in the combustion chamber. In the aim of increasing the injection system metering accuracy, the stand-alone injector analysis may be not adequate. In fact, the injector performances are affected by many boundary parameters, such as the pressure time-history upstream the injector, the backpressure, the fluid and injector nozzle temperature. In particular, the upstream pressure time-history is largely influenced by the injection system layout and by its actuation frequency. As well known, fast opening and closing of injectors and alternate motion of the high pressure pump piston produce considerable pressure fluctuations in the fuel system, possibly resulting in an injector behavior in terms of fuel metering and injection rate profiles that can be significantly different from what is predicted by a standard, stand-alone injector

FPGA GDI m P PID Q_ TDC V V0

q

b

Field Programmable Gate Array Gasoline Direct Injection fluid mass (kg) pressure (bar) Proportional–Integral–Derivative flow rate (m3/s) Top Dead Center Pipe or chamber volume (m3) Injector Analyzer internal chamber volume fluid density (kg/m3) fluid Bulk Modulus (Pa)

characterization [1–9]. Further, pressure fluctuations can result in operation noise and damaging of some components [10]. Hence, the ability in evaluating (both numerically and experimentally) the actual operating conditions for the entire injection system is crucial to obtain robust results. The Wet System approach is commonly used in order to replicate, in a research test bench, the overall injection system and to control its working parameters. A GDI Wet System test bench was set-up at the SprayLab – Università di Perugia, in partnership with Magneti Marelli Powertrain, for the direct analysis of complete GDI systems and to support the development of numerical tools for the corresponding computational analysis. This paper describes a combined 1-D numerical and experimental approach applied to the analysis of a three-cylinder GDI system application. A numerical model of the complete injection system built in AMESim environment was validated through the comparison of predicted results with the evidences obtained by the SprayLab Wet System bench. Significant insight into the system layout effects on the injector hydraulic operation in terms of injection rate with single and multiple injection strategies was obtained. Peculiar attention was devoted to the influence of pressure waves propagation phenomena in different operating conditions. 2. The GDI Wet System test bench The GDI Wet System test bench assembly is show in Fig. 1, with the complete injection system composed by the high pressure pump, the pipe connecting it to the rail and the three injectors.

Pipe Sensor

GDI Pump

Common Rail

Rail Sensor

Injection Analyzer

TorqueMeter

Fig. 1. GDI Wet System test bench layout.

10

L. Postrioti et al. / Fuel 173 (2016) 8–28

Fig. 2. Magneti Marelli PHP GDI pressure pump (left); Magneti Marelli IHP1-270 GDI injector (right).

The pump shaft is driven by a 3 kW AC motor controlled by an inverter to regulate the pump speed, with the transmitted torque being measured by a dual range (0–10 Nm, 0–50 Nm, 3 kHz bandwidth) Kistler 4503 torque meter. The pump body is housed on a cam-box that ensures the coupling between the cam and the pump plunger. The high-pressure pump (Fig. 2, left) is a Magneti Marelli volumetric single-piston unit; it is fed by the low pressure circuit (4 bar) and it is actuated by a 3-lobe cam for the present application. A normally-open, solenoid actuated Fuel Metering Valve at the pump inlet is used to control the rail pressure in the range from 20 to 200 bar. To this end, a PID control system actuates the FMV actively closing the pump admission port in the compression stroke, starting the actual delivery process to the rail. Depending on FMV activation advance with respect to the pump shaft TDC, the flow delivered to the rail can be controlled thus regulating the rail pressure. The 38 cc common rail is a prototype housing three injectors for this application. The rail is designed in order to allow quick modifications of the system hydraulic configuration in terms of supply pipe insertion position, presence of flow restrictors, addition of dummy bodies to reduce the rail internal volume. Further, it offers different seats for pressure and temperature sensors. In order to analyze the pressure oscillations effect on the hydraulic system behavior, a pressure sensor (Kulite ETL-173-190) was installed along the pipe connecting the pump and the rail, while a second sensor (Kistler 4065A500) was mounted on the rail in different positions. The injectors used in the present experimental campaign are Magneti Marelli IHP1-270 direct-actuation solenoid injectors with a pressure range up to 200 bar (Fig. 2, right). The fluid temperature is acquired and regulated (from ambient to 75 °C) both at the high pressure pump inlet and in the commonrail by means of different heating elements in order to ensure the required temperature level regardless of the pump flow rate. During the test, all the three injectors are operated according to the engine actuation sequence and phased with respect to the pump shaft position. Two injectors (#1 and #3 in Fig. 8) operate with no counter-pressure as they are mounted on fixtures draining the injected liquid to the reservoir, while the analyzed injector (#2 in Fig. 8) is mounted on the UniPg Injection Analyzer, operated with a 5 bar back-pressure. The UniPg Injection Analyzer (Fig. 3) is a proprietary mass flow rate meter [11,12] that allows both the injected volume and the injection rate time-history measurements in a given operating condition, defined by the injector actuation cycle (i.e. the pulses sequence per engine cycle) and by the rail pressure level. These measurements are resolved in the injector actuation cycle, allowing a shot-to-shot analysis of both the injection rate profile and

Fig. 3. UniPg Injector Analyzer.

of the injected quantity. In case of multiple injections strategy, it is possible to evaluate the fluid volume relevant to each single event by measuring the respective injection rate curves. The injection rate measurement is based on the Zeuch’s Method, according to which the injection process takes place in a closed and constant volume chamber filled with the pressurized injected fuel. The fluid DV forced by the injector into the chamber during the injection process causes a rise DP of the vessel pressure P, which is proportional to DV according to the compressibility of the liquid b, indicated by its Bulk Modulus, and the chamber volume V0.

b ¼ V0

DP DV

ð1Þ

The injection rate equation dV/dt is derived from the Bulk equation.

dV V 0 dP ¼ dt b dt

ð2Þ

The injection rate in mass can be evaluated as

dm d dV dq ¼ ðqVÞ ¼ q þV dt dt dt dt

ð3Þ

11

L. Postrioti et al. / Fuel 173 (2016) 8–28

Pressure Rise During Engine Start-up

Table 1 Current profile settings. 65 V 12 V 12 A 700 ls 2.60 A

Considering the density variations negligible during the injection process, the final equation is:

dm V 0 dP ¼q dt b dt

ð4Þ

Therefore, the rate of injected mass can be considered proportional to fuel density, chamber volume, and rate of chamber pressure rise and inversely proportional to the fuel Bulk Modulus. The chamber volume V0 must be sized in order to limit the chamber pressure rise within appropriate limits (typically 0.015–1.0 MPa) for the actual injected volume range (e.g. 0.5–50 mm3/shot). Following each actuation cycle, a fast electro-valve opens discharging the injected fuel and bringing the chamber pressure to the original base pressure before the next injection cycle starts. As the fluid Bulk Modulus is significantly influenced by both temperature and pressure, it can be hardly evaluated especially when performing tests with non-standard fuels. To accomplish this task, the fuel discharged from the instrument chamber flows through a Coriolis mass flow meter (Siemens Sitrans CF 2100) to measure the mean injected mass and the fuel density, so to ensure a continuous system calibration in the actual operating conditions. In order to obtain both a proper thermal stabilization in each operating condition and a statistically significant data amount to evaluate the shot-to-shot dispersion, the acquisition procedure is repeated over some hundreds of consecutive injection cycles, allowing a robust evaluation of the mean injector behavior and of the dispersion in terms of injection rate and injected quantity. The Wet System test bench can operate both in time-base or in angle-base modes. The time-base mode is used when the standalone injector behavior is investigated: here the bench control system is used to generate a general trigger to start both the injector actuation and the high frequency data acquisition (50–200 kHz). Typically, in this mode the injector is fed by a static pressure generator and the basic injected mass vs. Energizing Time (ET) curves along with the standard injection rate profiles are measured. Conversely, the angle-base mode is used to analyze the complete Wet System operation, in which the pump shaft angular speed is not uniform nor perfectly repeatable. In order to account for the

70 60

Pressure, bar

V_boost V_batt I_peak T_peak I_hold

80

50 40

air/gas content = 0.1% air/gas content = 0.3% air/gas content = 1.0% air/gas content = 2.0% air/gas content = 3.0%

30 20 10 0

Fig. 5. GDI pump start-up performance for different aeration conditions.

angular speed variability an encoder (Baumer CH-8501, 1024 pulses/rev) mounted on the pump shaft is used to derive the synchronization signals to actuate both the injectors and the Injection Analyzer discharge valve. In this operating mode, the encoder pulses are also used as clock signal for the data acquisition system. The bench control system is developed in LabVIEWTM environment and it is structured on different logical levels: FPGA, RealTime Controller, User Interface & DAQ. The FPGA and Real-Time levels are implemented on a NI Compact-RIO 9074 platform, while the User Interface level is based on a PC and a PCIe-6351 DAQ card. The input signals are typically filtered using a 5 kHz low-pass threshold. At the FPGA level, in the angle-base mode, the encoder signals are detected and the synchronization signals for the injectors, the discharge valve and the pump FMV drivers are generated according to the information received by the higher-level Real-Time Controller. In the time-base mode, an internally generated trigger is used to generate the same synchronization signals, with a global frequency up to 60 Hz. In both the operating modes, up to five injection events per actuation cycle have been implemented, with freely programmable Energizing Time (ET) and Dwell Time (DT) values. To actuate the injectors, a programmable multi-channel driver from Magneti Marelli is used, while proprietary peak-andhold drivers are used to operate the FMV and the discharge valve. At the Real-Time Controller level, the rail pressure control is implemented using a PID logic, with a 10 ms execution loop time. Further, the Real-Time software receives inputs from the User Interface & DAQ level relevant to the injection strategy and the target rail pressure level to be actuated, performs plausibility checks and safety control tasks on the bench and transmits the final

ET 0.3 ms ET 0.4 ms ET 0.6 ms ET 1.0 ms ET 2.0 ms ET 3.0 ms

Fig. 4. IHP1-270 injector current profiles.

12

L. Postrioti et al. / Fuel 173 (2016) 8–28

Fig. 6. High pressure pump validation.

14 12

Current, A

10 8

DT 4 ms DT 1 ms DT 0.3 ms

6 4 2 0 -2

of the actuation strategies in the tests, the planned number of actuation repetitions is managed into two parts: the first actuations batch is used to stabilize the system thermal conditions for the current actuation strategy (normally 200 actuations), while the remaining actuations are used for the final data acquisition and analysis (normally from 200 to 500). From all the acquired signals from the field, the mean timehistory profile is computed, allowing the dispersion analysis over the analyzed consecutive events. In particular, from the Injection Analyzer chamber pressure profile, the injected quantity and the injection rate for each of the considered events is evaluated according to Eqs. (1) and (4). The IHP1-270 injectors were actuated using the current profile settings reported in Table 1. The resulting current time-profiles for different ETs are reported in Fig. 4.

Fig. 7. Dwell time influence on current profile during multi-injections.

3. Model description settings to the FPGA level. At the higher software level, the Data Acquisition of the following signals is carried out:







Pump FMV current. Pipe and Rail pressures. Injectors Current. Injection Analyzer chamber pressure (Kistler 4075 – 100 bar). Injected mass (Siemens Sitrans CCF2100). Fuel Temperature (type-K thermocouples).

At the same level, the User interface part of the software is used for the test management. The test setup allows the system operating conditions definition in terms of rail pressure and injectors actuation strategy, defined by the sequence of up to five ETs and four DTs. The standard test is defined by a ramp of one of the nine ET or DT quantities, allowing a parametric analysis of the system behavior. The ramp can be freely defined by the user, both in terms of the values for the explored parameter and of the number of actuation repetitions for each actuation strategy. Finally, for each Table 2 System configurations description – non-standard parameters green shaded.

The fuel system was modeled following a lump parameters approach in LMS AMESim environment (Fig. A-27 in Appendix) [13]. This tool is often used to simulate hydraulic systems according to a 1-D approach, allowing a powerful analysis in time and frequency domains. Linear analysis in fact plays a crucial role during the design process detecting system’s natural frequencies and associating them to specific components [14–17]. Each sub-part of the fuel supply system has been modeled by means of an assembly of elementary components coming from the different available libraries, building the so-called supercomponents. Modeling every single part separately allowed a deep validation of the system components step by step. The advantages of this approach are several. From one side it is helpful to easily understand how to improve the entire model validation. On the other side, it offers the chance to use every validated sub-model in other future applications sensibly reducing the time required to realize a new mathematical model. Furthermore, a combined study of the fuel system using both experimental results and simulation model is even decisive, offering the chance to properly evaluate all the parameters which are difficult to be directly measured but that must be adjusted to get meaningful results for the entire system simulation. Common examples are frictions and aeration. These parameters are only rarely available for a specific system as the field measurement is complex and time consuming. Nevertheless, aeration has a great impact on a pumping system behavior both in time and frequency response. In Fig. 5, the air/gas content influence on a GDI pump performance during the engine start-up process is reported.

13

L. Postrioti et al. / Fuel 173 (2016) 8–28

Layout 1

Layout 2

Fig. 8. Analyzed common rail layouts.

Base characterization Parametric Analysis

Op. Point

Multiinjection

Energizing Time [µs]

Prail [bar]

Configuration

50

200-3000

1

100

200-3000

1

150

200-3000

1

Pump speed [rpm]

Prail [bar]

Qinj [mg/shot]

Configuration

Low Load

500

50

6

1

Mid Load

1200

120

15

1

High Load

3000

150

30

1

Prail bar] 100

Energizing Time[µs]

500/500

Dwell Time [µs] 2000

400

Configuration 1

Fig. 9. Experimental test plan.

3.1. Fuel pump For this work, the high pressure pump (Fig. 2, left) has been modeled in detail to investigate its operation effects on the global injection system. In particular, all the key components have been modeled separately in order to achieve a good level of reliability, including the inlet valve, the electromechanical actuator, the

pumping element connected to the cam and the outlet valve. Moreover, also an overpressure valve and a damper at the inlet of the pump were included in the pump model to limit pressure fluctuations on the supply circuit (reproducing the experimental Wet System bench setup). As regards the PID controller, a simplified model replicating the real logical architecture of both the complete ECU control and the Wet System bench was created in

14

L. Postrioti et al. / Fuel 173 (2016) 8–28

(a)

(b)

Fig. 10. EMI curves (a) and injected volume Coefficient of Variation (b) for Prail 50, 120 and 150 bar.

AMESim. In particular, it consists in a closed loop control continuously comparing the measured rail pressure and the target value and updating the opening and closing timing of the electromechanical actuator connected to the inlet valve (FMV). A first validation of the 1-D model of the MM GDI Pump was performed and its results were compared with the experimental data coming from dedicated efficiency tests, at different working pressures, performed at Magneti Marelli hydraulic test bench (Fig. 6). A further and deeper validation of the model has been performed as regards all the internal volumes and main natural frequencies thanks to the pressure pulsations measurements obtained by the Wet System bench. 3.2. Injectors The aim of this component is to ensure the amount of fuel delivery requested with the right timing and spray quality (Fig. 2, right). In this study, the injectors have been modeled in detail to analyze their behavior depending on rail pressure fluctuations. To this aim, peculiar attention was devoted to mechanical and hydraulic

Table 3 n-Heptane physical properties. Density (kg/m3)

0.681–0.687

Boiling point (°C)

@ 40 kPa @ 100 kPa @ 300 kPa

69.65 98.25 140.35

Speed of sound (m/s)

@ 298.15 K

1130

modeling, since evaluating the correct volume inside the injector is crucial in order to obtain a meaningful response for the pressure pulsations. For the electro-magnetic behavior of the injector, two distinct modeling approaches were used in the different steps of the present research for different injector actuation strategies. In a first step and for single-event actuation strategies, basing on previous FEM simulations, only the driving force acting on the needle was considered without a direct simulation of the electro-magnetic phenomena. This approach was used to validate the mechanical and hydraulic modeling of the injector and it brought to satisfying results as regards all the tests concerning a single-injection per engine cycle strategy, for both the so-called EMI and EVI curves for all the examined working pressures. A slightly more complex approach was needed for tests concerning multiple injections. In this case the influence of coil residual magnetization is not negligible for very close injections (usually concerning a dwell time lower than 1 ms). This behavior is highlighted in Fig. 7, showing how the current slope increases for very close injections (smaller dwell time), significantly affecting the opening force applied to the injector needle. The introduction of the dependency of the force as a function of the dwell time was mandatory in order to enhance the model behavior for multiple-injection tests. Finally, in order to obtain an accurate validation of the injector operation on the Wet System bench, a simple model of the Injection Analyzer chamber was included in the AMESim model. By this approach, the actual pressure boundary condition time-history was accounted for during the injection process, resulting in a more accurate evaluation in terms of instantaneous flow rate.

15

L. Postrioti et al. / Fuel 173 (2016) 8–28

(a)

30 28 26 24 22 20 18 16 14 12 10 8 6 4 2 0 -2 0.0

Prail 50 bar

0.5

ET 0.3 ms ET 0.4 ms ET 0.6 ms ET 1.0 ms ET 2.0 ms ET 3.0 ms

1.0

1.5

2.0

2.5

3.0

3.5

4.0

Time from ET start, ms

Injection Rate, mm3/ms

(b)

30 28 26 24 22 20 18 16 14 12 10 8 6 4 2 0 -2

0.0

Prail 100 bar

0.5

ET 0.34 ms ET 0.4 ms ET 0.6 ms ET 1.0 ms ET 2.0 ms ET 3.0 ms

1.0

1.5

2.0

2.5

3.0

3.5

4.0

Time from ET start, ms

Injection Rate, mm3/ms

(c)

30 28 26 24 22 20 18 16 14 12 10 8 6 4 2 0 -2

Prail 150 bar

ET 0.34 ms ET 0.4 ms ET 0.6 ms ET 1.0 ms ET 2.0 ms ET 3.0 ms

Fig. 11. EVI curves for Prail 50, 120 and 150 bar.

Fig. 12. Injector model validation: EMI curves.

16

L. Postrioti et al. / Fuel 173 (2016) 8–28

Fig. 13. Injector model validation: EVI curves.

3.3. Common rail and pipes Both rail and high pressure pipes were modeled using complex lumped parameters sub-models in which resistive, capacitance and inertia effects were taken into account in order to correctly analyze wave effects. Hydraulic sub-models are governed by a basic equation concerning the mass conservation for a generic volume:

V dp dV X _ Qi  þ ¼ b dt dt i

ð5Þ

where V is the volume of each line or chamber, dV is the volume varidt ation due to pumping effect of the plunger displacement; Q_ i are the

flow rates coming in and out of the volumes and b is the effective Bulk Modulus; In detail, b is evaluated as the effective Bulk Modulus, resulting from the combination of both fluid and wall properties according to the following equation:

1 1 ¼ þ C wallðpÞ bðpÞ bfluidðpÞ

ð6Þ

where bfluid is the fluid Bulk Modulus, computed at the pressure p and Cwall is the wall compliance. Anyway, dealing with steel pipes in the fuel system model, the effective Bulk Modulus appears as

L. Postrioti et al. / Fuel 173 (2016) 8–28

17

(a)

(b)

(c)

Fig. 14. Experimental results for rail and pipe pressure in base configuration at low (a), medium (b) and high (c) load.

equal to fluid Bulk Modulus as the lines are assumed as extremely rigid. Common experience demonstrates that the modeling phase of all the high pressure and low pressure lines is crucial in the study of a complete injection system. In this research, every connection was studied separately in accordance to its position and design, in order to find the right trade-off between model accuracy and acceptable computational times.

4. Test plan Being the present research focused on the analysis of the rail pressure fluctuations effects on the injector operation, a parametric investigation of different hydraulic circuit designs was carried out. Four system configurations (Table 2) were used in order to deepen the effect of several parameters on the complete injection system hydraulic response. In detail, the influence exerted by the introduction of a flow restrictor at the rail inlet was investigated; to this end the system was tested with and without a 1.5 mm diameter calibrated hole. Further, the effect of different length and internal diameter for the pump–rail pipe was analyzed. Finally, two different layouts

for the fuel rail were used; the first one with a top connection for the pump–rail pipe (Fig. 8, Layout 1), the second with a side connection (Fig. 8, Layout 2). The above described system layouts were used for the Wet System analysis according to the experimental test plan reported in Fig. 9, carried out on the Wet System GDI bench at SprayLab. In the first part of the research activity, the base injector characterization was performed using Layout 2 in order to obtain the mean injected volume per shot (EMI curve) and the injection rate timeprofiles (EVI curves) for 50 bar, 100 bar and 150 bar rail pressure levels. In this base analysis, for both the injected volume and the injection rate, the shot-to-shot dispersion was also determined. After the basic injector characterization, for each different layout, the system was tested at three load levels and pump speeds, to simulate three different engine operating conditions. As final step of the research activity, the system hydraulic behavior with multi-injection strategies was investigated; these experimental data were used to further assess the reliability of the developed 1D model of the complete Wet System. In the complete system hydraulic analysis, only the injector in the central position was admitted to the injection rate measurement, while the other two mounted at the sides of the fuel rail, injected in an ambient maintained at constant atmospheric

18

L. Postrioti et al. / Fuel 173 (2016) 8–28

Fig. 15. Injected flowrate and rail pressure for base configuration at low load.

pressure. Conversely, for the injector base characterization, Layout 2 was used with only the central injector under test being actuated, so to excluded reciprocal influences among consecutive injections. The fluid used during the present test campaign was n-heptane, a hydrocarbon used in research applications as an alternative to gasoline. It has a high chemical stability and well known chemical and physical properties (Table 3). 5. Results and discussion 5.1. Stand-alone injector characterization: tests and simulation The injector base characterization was carried out by means of the UniPg Injector Analyzer at three different rail pressure levels. The focus of these tests was the stand-alone injector analysis, hence just the injector under test was actuated. The EMI and EVI curves allowed the characterization of the actual injector operation when installed on the Wet System. Further, these data were used to evaluate the ETs required for the following injection quantityoriented analysis. The time-base acquisition mode was used with an acquisition time interval of 20 ls. The system configuration

used in this set of acquisitions was layout 2 (6 mm diameter and 420 mm long pipe) and 25 °C fuel temperature; this was assumed as the base system configuration. The EMI curves obtained for different pressure levels (50–100– 150 bar) are reported in Fig. 10. The EMI curves evidence the typical GDI two-zone operation differing each other for a net slope change: a ballistic zone, for small ETs, where the needle is not fully raised and a linear zone where the injected quantity becomes proportional to the command duration. In Fig. 10, increasing the rail pressure from 50 up to 150 bar, the ballistic zone and the minimum ET required to open the injector are shifted toward higher ETs because the injector needs a longer energizing time to overcome the higher pressure force on the injector control valve. On the other hand, the linear zone slope increases with the rail pressure because, once the needle is completely raised up, higher upstream pressure determines a higher injection rate. The injected volume dispersion, reported in terms of Coefficient of Variation CoV (i.e. the percentage standard deviation to mean value ratio) is noticeably reduced from the ballistic to the linear zone. The corresponding EVI curves are depicted in Fig. 11 for the three analyzed pressure levels and for different ETs. The injector

(a)

(b)

Fig. 16. Experimental comparison among base configuration and layout-1 configuration at medium load for rail pressure (a) and pipe pressure (b).

L. Postrioti et al. / Fuel 173 (2016) 8–28

19

(a)

(b)

(c)

(d)

Fig. 17. Experimental comparison among base configuration and no-flow restrictor configuration at medium load for rail pressure (a) and pipe pressure (b) and at high load for rail pressure (c) and pipe pressure (d).

operation in the ballistic zone can be appreciated in terms of injection rate for ET = 300 ls with Prail = 50 bar and 100 bar while for higher injection pressure values the linear zone is not yet achieved for ET = 400 ls. During ballistic operation the injection rate doesn’t attain the steady flow level because of the incomplete needle

opening; in these conditions the obtained flow rate level decreases at higher rail pressure for a given ET. For ETs in the linearoperation zone the steady flow is achieved and the injection rate curve duration depends only on the energizing time. The oscillations following the fast injector opening could be an effect of

(a)

(b)

Fig. 18. Experimental comparison among base configuration and longer-pipe configuration at medium load for rail pressure (a) and pipe pressure (b).

20

L. Postrioti et al. / Fuel 173 (2016) 8–28

(a)

(b)

(c)

(d)

(e)

(f)

Fig. 19. Experimental vs. Numerical results for rail and pipe pressure at low load ((a) and (b)), medium load ((c) and (d)), high load ((e) and (f)).

mechanical vibrations triggered by the stroke of the needle on its stop. The noise after the end of injection seems to be attributable to residual pressure wave reflections into the measuring chamber.

The developed numerical model was validated by means of the obtained experimental results both in terms of EMI and EVI curves. In Fig. 12 the EMI curves comparison among numerical and experimental results is depicted for the analyzed pressure levels.

L. Postrioti et al. / Fuel 173 (2016) 8–28

21

Fig. 20. Summary of natural and forcing frequencies of the system.

In the linear zone a good agreement is achieved, with the experimental curve being slightly lower for high ETs, high injection pressure operating conditions. In the injection rate measurement, the pressure increase in the test chamber during each injection process causes a decrease of the actual pressure delta between the rail and the downstream ambient. This differential pressure decrease can become significant for high injected quantity operating conditions and hence has been taken into account by the simulation of the Injection Analyzer in the model. In Fig. 13 the validation of EVI curves for three pressure levels and three ETs is reported. The agreement among numerical and experimental results is appreciable for many of the analyzed conditions. The model resulted to be very accurate for the linear zone operating conditions both in terms of injection start and end timings and in terms of flowrate level. In the ballistic zone the injector is strongly sensitive to small changes of the injector electromechanical parameters, hence generally the model accuracy is lower than in the linear zone. The injection rate at ET = 400 ls is precisely predicted at 50 bar, while at 150 bar, the numerical injection rate overestimates the experimental evidences. In fact, moving from Prail = 50 bar to 150 bar, the injector operating mode at ET = 400 ls shifts from the linear to the ballistic zone, evidencing the need for a deeper model development. 5.2. System dynamics experimental analysis During the dynamic analysis tests, the fluid-dynamic behavior of the overall Wet System was investigated, mainly in terms of pressure waveforms; angle-based acquisitions were performed recording data over an entire pump shaft revolution to better appreciate the pressure waveforms phasing related to the pumping system operation. A combined approach coupling both experimental tests and simulations was adopted in order to fully understand the influence of every single system parameter and accurately validate the numerical model. The Wet System was experimentally tested following the plan reported in Fig. 9. Each hydraulic configuration was analyzed in three different load conditions (low, medium and high) in order to investigate a wide engine operating range. Load conditions are characterized by the pump rotational speed, the rail pressure and the fuel injected quantity per cycle as reported in Fig. 9. The hydraulic layout relating to Config#1 (see Table 2) was set as base configuration and considered as reference for the following parametric analysis. All the experimental plots in this section are the result of an averaging procedure over 200 raw records; further, no filters were applied to raw data before the averaging procedure in order to preserve the actual phasing among the different signals. In the present tests, the injector was actuated with a null delay with respect to the pump TDC, assumed as pump crank angle reference position in the following plots. In Fig. 14 experimental

results for the base configuration at low (a), medium (b) and high (c) loads are plotted in terms of rail and pipe pressure waveform over a complete pumping period, i.e. 120 cad for the 3-lobe cam. In Fig. 14, the pressure time-history during the pumpinginjection cycle can be observed along with the FMV actuation current profile. Increasing the pump speed and load, requires a greater FMV actuation advance with respect to the pump TDC. In fact, the FMV operation during the pump compression stroke generates a pressure pulse at the pump delivery port with a delay depending on the FMV electrical and mechanical inertia and on the pump rotational speed. The generated pressure wave then travels through the connecting pipe toward the rail, being detected by the pipe pressure sensor with a further delay depending on the sensor position along the pipe. Finally, the flow rate and pressure pulses propagate in the rail where the rail pressure sensor detects it. The FMV advance relative to pump TDC is bigger for higher load/speed condition (17 cad for low-load, 31 cad for mediumload and 52.5 cad for high-load). The corresponding rail pressure peak delay from the FMV actuation in the considered operating condition is a consequence of the increase of both pump speed and injected volume. The oscillations amplitude for the pipe pressure is always higher than for rail pressure, due to the bigger rail internal volume and to the flow restrictor effect (1.5 mm diameter in base configuration). Phase displacement between rail and pipe pressure waveforms is determined by the time the pressure wave spends to cover the distance between these two pressure sensors. In medium-load condition (Fig. 14(b)), the computed phase shift is 1.41 cad that, at 1200 rpm, corresponds to 0.196 ms. Being the actual distance between rail and pipe pressure sensors 222 mm, the calculated speed of sound is v = d/t = 1132 ms, in good agreement with the actual value for the n-heptane at ambient temperature reported in Table 3. The injection event, phased with the pump shaft TDC, determines a negative pressure wave which travels back from the injector toward the pump and is reflected by the pump check valve. In Fig. 15 it is possible to appreciate the injector flowrate and the consequent pressure decrease into the rail. Taking the base configuration (Config#1) as a reference, the effect of the high pressure pipe design, in terms of internal diameter and length, the flow-restrictor and the rail inlet position were studied. In this section, not all results are reported for the sake of brevity. In Fig. 16 the experimental results for Config#4 at medium load, compared to the base configuration (Config#1), are depicted. The pipe diameter reduction from 6 to 2 mm, i.e. nine times in terms of area, increases the pipe pressure amplitude from 8 to 18 bar due to the lower internal volume. The pressure oscillations are then reduced in the rail due to the relevant fluid volume stored and hence they are quite similar to the larger pipe Config#4. The actual effect of the flow-restrictor on the injection process was analyzed removing it from the inlet pipe connection to the rail

22

L. Postrioti et al. / Fuel 173 (2016) 8–28

(a)

(b)

(c)

(d)

(e)

(f)

Fig. 21. Experimental and numerical spectrum analysis at low load ((a) and (b)), medium load ((c) and (d)), high load ((e) and (f)) for every tested condition.

L. Postrioti et al. / Fuel 173 (2016) 8–28

23

Fig. 22. Rail pressure during 800–7000 rpm sweep (a) and 3-dimensional display of the spectral map (b). No restrictor architecture.

in the reference Config#1, obtaining Config#2. As can be seen in Fig. 17, both rail and pipe pressure first impulses, located at about 110 cad, are synchronized with the reference curve because the hydraulic configuration is the same, except for the removal of the flow-restrictor. The pressure pulsation main frequency is slightly increased without the flow-restrictor and the pipe pressure amplitude is slightly reduced in medium load conditions (Fig. 17 (a) and (b)). The experimental results obtained in high speedhigh load conditions evidence the flow restrictor effectiveness in damping intense pressure pulses, that in this application are significantly attenuated in the connecting pipe. As consequence, also the pressure time-history in the rail is significantly smoothed (Fig. 17 (c) and (d)). The effect of the high pressure pipe length was investigated in order to comprehend its influence on the Wet System performances. A 520 mm long pipe was tested (Config#3) and compared with the base configuration (Config#1, 420 mm pipe). The other parameters were kept constant. As shown in Fig. 18(a), at about 110 cad a slightly delay in the rail pressure pulse detection determined by the increased pipe length can be appreciated; conversely, in Fig. 18(b), the pipe pressure waveform does not evidence a delay because the distance between pump and pipe pressure sensor is the same. The increment in length produces a decrease in the system main frequency from 500 to 420 Hz. Both the rail and the pipe pressure amplitude are similar and also the pressure pulsations dampening rate does not change. 5.3. Numerical results and model validation For each experiment, a parallel simulation was run with the purpose of tuning the 1-D model and enhancing its accuracy. The

validation was achieved using both pressure acquisitions coming from the rail and the high pressure pipe sensors. The results were compared either in time or in frequency domain, achieving consistent outcome in all test conditions. In Fig. 19 numerical results (solid curves) are depicted overlapped to the correspondent experimental plots (dot curves) for three cases at different loads. It is interesting to point out that the experimental data used for these comparisons are related to a pump single-period raw acquisition. The reason is that a not negligible dampening effect on the pressure profile was noticed analyzing the mean signal of several consecutive pressure acquisitions, especially at low pump rotational speed. This determines a misleading behavior of the pressure signal which would be impossible to get by the 1-D model. As can be seen, despite the unfiltered experimental signals used for the present analysis, a good agreement among experimental and numerical results has been obtained in terms of pressure peaks intensity, frequency content and phase with respect to the forcing factors, i.e. the actuation of the FMV and the injectors.

5.4. Spectral analysis According to the common experience gained during engine applications simulations, the linear analysis of the system was performed to get proper knowledge of the system natural frequencies. The linear Analysis Tool from LMS AMESim was used in this part of the activity. By using this approach, the natural frequencies for the different system layouts were determined (Fig. 20), to be compared with the forcing phenomena frequencies, represented in this case by the pumping effect and by the three injectors actuation.

24

L. Postrioti et al. / Fuel 173 (2016) 8–28

Fig. 23. Rail pressure during 800–7000 rpm sweep (a) and 3-dimensional display of the spectral map (b). 1.5 mm restrictor architecture.

Fig. 24. Multi-injection EVI curve.

A frequency domain analysis relevant to the system first mode, using FFT procedure, was carried out for the whole test plan both for experimental and numerical data; the spectral results are reported in Fig. 21. Globally, an appreciable agreement between the spectral analysis of experimental and numerical pressure traces was obtained. In particular, the FFT procedure results seem to confirm the Linear Analysis outcome, highlighting how natural frequencies move according to system layout and high pressure pipe length. All the tested operating conditions confirm Config#4 as the fuel system layout with the lowest natural frequency

(300 Hz, red plots in Fig. 21); further, for this configuration also the amplitude of the pressure oscillations is significantly increased with respect to the other configurations. Moreover, the increase of the pipe length of about 100 mm in layout 2 (Config#3 vs. Config#1) yields to a total decrease of about 100 Hz for the first natural frequency of the whole system for low and medium load conditions. The effect exerted by the flow restrictor removal (Config#2 vs. Config#1) on the pressure time-history in terms of spectrum analysis is mainly a growth of a high frequency content (around 600 Hz) for all the tested operating conditions.

25

L. Postrioti et al. / Fuel 173 (2016) 8–28

Fig. 25. Comparison between measured EVI curves and simulation results for different dwell time.

*** RAV ***

20

simulation results experimental results

tot inj volume [mm3]

19.5

19

18.5

18

17.5

17

0

500

1000

1500

2000

2500

3000

3500

4000

[us] Fig. 26. Comparison between measured injected mass and simulation results for different dwell time.

5.5. Model extrapolations The validated 1-D model of the complete injection system can be used to investigate its dynamic behavior in a wide range of operating conditions. As an example, the code has been applied to the analysis of the efficacy of a flow restrictor at the rail inlet to minimize the rail pressure pulsation over a sweep of engine speed, covering all usual engine working conditions (from 800 to 7000 rpm). The results are shown in spectral maps where X axis

is the signal frequency in Hertz, Y axis is the engine speed in rpm and Z axis is the pressure amplitude (Figs. 22 and 23). Comparing the results of two different configurations (Config#2 vs. Config#1) shows that high levels of pressure pulsation corresponding to 600 Hz frequency are present in the system without flow restrictors. These oscillations occur at all engine speeds and become more relevant increasing the engine speed. The explanation of this behavior is that 600 Hz is a natural frequency of the system with no flow restrictor and it is excited by the forcing fre-

26

L. Postrioti et al. / Fuel 173 (2016) 8–28

Fig. A-27. AMESim model of the three-cylinder Wet System – main components.

quency of injection and pumping. The 3-D representation of the spectral map (Fig. 22(b)) reveals the interaction between the first natural frequency of the system and the different orders harmonics relative to the forcing phenomena connected with consecutive injection and pumping events. These harmonics are represented by the lines diagonally cutting the XY plane. The lowest natural frequency of the system appears to be excited by high order harmonics even in low speed conditions. Higher pressure pulsations occur with lower order harmonics running into the system natural

frequency. These results confirm and explain the data collected during high speed tests (3000 pump rpm) where oscillations of remarkable amplitude where noticed for 600 Hz frequency both in experimental and simulations results. Conversely, including the flow restrictor in the system layout, the 600 Hz natural frequency is no longer present resulting in significantly smoother operation in terms of rail pressure (Fig. 23(b)) and confirming the flow restrictor to be a proper solution for high frequency oscillations dampening.

L. Postrioti et al. / Fuel 173 (2016) 8–28

5.6. Multi-injection strategy analysis The developed and tuned 1-D model of the analyzed GDI injector was also used to investigate the system behavior with multievent strategies in terms of both injected volume and injection rate. To this end Config#1 (flow restrictor at the rail supply connection, pump–rail pipe 420 mm) was tested and modeled. An operating condition of 100 bar injection pressure and a double ET = 470 ls with variable dwell time DT was used; in particular, both the ETs for the first and second event were chosen in order to avoid a possible injector operation in ballistic mode that, as shown in the basic injector characterization section, is typically affected by a high shot-to-shot dispersion in terms of injected volume and instantaneous mass flowrate. In this analysis, DT was varied between 2 ms and 0.3 ms, covering the entire spectrum of operating conditions from completely separated events to merged injection events. The obtained experimental curves show injections melting (see Fig. 24) for dwell time shorter than 400 ls and a consistent increase in the total injected mass for DTs lower than 2 ms (Fig. 26). The phenomenon, as well known, is mainly determined by residual magnetization effects for the injector coil bringing to a faster opening during the second injection and a consequent injected mass increase. A change in the injected volume relevant to the injector actuation in a multiple event strategy may also be due to the pressure dynamics in the rail, depending on the system geometry, as evidenced by the injected quantity vs. dwell time curve reported in Fig. 26. Here, monotonically decreasing DT values results in an injected volume trend with local maximum and minimum conditions, which can be explained as the result of rail pressure oscillations differently phased with the injector actuation. In order to model multi-injection strategies, as discussed in the Model Description Section, a slight enhancement of the 1-D model was needed. During multi-injections in fact, eddy currents strongly affect the time-profile of the solenoid current and hence of the magnetic force during the opening phase of the injector. This detail is not negligible in this kind of analysis and so a dedicated modeling of the electro-magnetic circuit was added to the existing 1-D injector model. This approach gave a satisfactory degree of confidence in the model used for predicting the overall system behavior. The model offers meaningful results both regarding the global injected volume and instantaneous mass flow rate for all the tested conditions (see Figs. 25 and 26). As can be seen, the experimental injection rate is well predicted in both reduced dwell time operation and conditions in which two consecutive actuation produce fused injection events. Further the used 1-D model seems to be able to predict a small post injection event possibly due to hydraulic phenomena, as also evidenced by the experimental analysis. In terms of global injection quantity as a function of the reducing dwell time, the raising experimental trend seems to be captured by the model (Fig. 26) with a slight underestimation for dwell times in the range 2–1 ms. Further improvements in the 1-D tool accuracy could be attained with a more detailed modeling of the electro-magnetic phenomena involved in the injector operation, coupled with a dedicated experimental campaign aimed at characterize the magnetic force as function of the injection strategy. This would certainly bring to more reliable simulations for close injections; these improved capabilities would be probably paid with a consistent increase in the computational time.

6. Conclusions In the present paper, an integrated numerical and experimental approach for the analysis and design of a complete GDI injection

27

system is described. This combined approach is mainly focused on the system behavior optimization in terms of injection metering accuracy and injection rate control capabilities, both significantly influenced by the system hydraulic layout. The experimental activity was carried out by a hydraulic test bench (Wet System) composed of a cam driven, single piston GDI pump, a research common rail feeding three injectors and a pipe connecting the pump and the rail. The three injectors were actuated with an engine-like operation sequence. During the system operation, in different speed and load conditions, one of the injectors was used to measure the instantaneous flow rate by means of a Zeuch Method based Injection Analyzer; further, the pressure time-history in the rail and along the pump–rail pipe were acquired. The effects exerted by the system hydraulic layout (pump–rail pipe length and diameter, flow restrictor presence) on the pressure pulsation and injection rate were determined. In parallel, 1-D numerical model of the complete experimental Wet System was developed in LMS AMESim environment, simulating the high pressure pump, the pump to rail pipe, the rail and the injectors. The aim was obtaining a detailed validation of all the included component modeling, enabling the use of the complete numerical Wet System to investigate in detail the hydraulic system behavior. Globally, an appreciable agreement between experimental and numerical results was obtained during the validation phase, enabling a parametric analysis to investigate the effect of the system geometry and configuration on the fluid-dynamic performances. The effect of the high pressure pipe length and diameter, of the flow-restrictor at the rail inlet and of the rail inlet position was studied in terms of pressure waves signal with the aim of spectral analysis over a wide range of operating conditions (from 800 to 7000 rpm, with injection pressures up to 150 bar). Finally, the effect of the dwell time variation in multi-injection strategies was investigated in terms of injection rate ranging from isolated to completely fused injection events. The analysis was carried out both experimentally and numerically, obtaining a satisfactory agreement. The mutual influences between contiguous injections, resulted in a significant increase of the injected volume, mainly interpreted as the consequence of residual magnetic effects in the dwell time. Acknowledgments The Authors wish to thank Magneti Marelli Powertrain for having supported this research (R.C. 4000043785). Appendix A See Fig. A-27. References [1] Baumgarten C. Mixture Formation in Internal Combustion engines. SpringerVerlag; 2006. ISBN-13 978-3-540-30855-5. [2] Li J, Treusch C, Honel B, Neyrat S. Simulation of pressure pulsations in a gasoline injection system and development of an effective damping technology. SAE technical paper 2005-01-1149; 2005. http://dx.doi.org/10. 4271/2005-01-1149. [3] Spegar T. Minimizing gasoline direct injection (GDi) fuel system pressure pulsations by robust fuel rail design. SAE technical paper 2011-01-1225; 2011. http://dx.doi.org/10.4271/2011-01-1225. [4] Baur R, Blath J, Bohn C, Kallage F, et al. Modeling and identification of a gasoline common rail injection system. SAE technical paper 2014-01-0196; 2014. http://dx.doi.org/10.4271/2014-01-0196. [5] Lee J, Liu H, Noh Y, Shin H, et al. A model based design analysis for a gasoline direct injection pump. SAE technical paper 2015-01-1267; 2015. http://dx.doi. org/10.4271/2015-01-1267. [6] Spegar T, Chang S, Das S, Norkin E, et al. An analytical and experimental study of a high pressure single piston pump for gasoline direct injection (GDi) engine

28

[7]

[8]

[9]

[10]

[11]

L. Postrioti et al. / Fuel 173 (2016) 8–28 applications. SAE technical paper 2009-01-1504; 2009. http://dx.doi.org/10. 4271/2009-01-1504. Hiraku K, Tokuo K, Yamada H. Development of high pressure fuel pump by using hydraulic simulator. SAE technical paper 2005-01-0099; 2005. http:// dx.doi.org/10.4271/2005-01-0099. Bianchi G, Forte C, Negro S, Pelloni P. A 1D model for the prediction of flash atomization in GDI multi-hole injectors: preliminary results. SAE Int J Engines 2009;1(1):1278–93. http://dx.doi.org/10.4271/2008-01-2516. Hu Q, Wu S, Lai M, Stottler S, et al. Prediction of pressure fluctuations inside an automotive fuel rail system. SAE technical paper 1999-01-0561; 1999. http:// dx.doi.org/10.4271/1999-01-0561. Orand N, Rhote-Vaney R. Modeling, validation and analysis of the fuel supply and injection system for NVH improvement. SAE technical paper 2009-012055; 2009. http://dx.doi.org/10.4271/2009-01-2055. Postrioti L, Ubertini S. An integrated experimental–numerical study of HSDI diesel injection system and spray dynamics. SAE technical paper 2006-011389; 2006. http://dx.doi.org/10.4271/2006-01-1389.

[12] Postrioti L, Buitoni G, Pesce FC, Ciaravino C. Zeuch method-based injection rate analysis of a common-rail system operated with advanced injection strategies. Fuel 2014;128:188–98. http://dx.doi.org/10.1016/j.fuel.2014.03.006. [13] AMESim user manual. LMS Imagine Lab, Version 13 SL2. [14] Bianchi G, Falfari S, Brusiani F, Pelloni P, et al. Advanced modelling of a new diesel fast solenoid injector and comparison with experiments. SAE technical paper 2004-01-0019; 2004. http://dx.doi.org/10.4271/2004-01-0019. [15] Bianchi G, Falfari S, Brusiani F, Pelloni P, et al. Numerical investigation of critical issues in multiple-injection strategy operated by a new C.R. fastactuation solenoid injector. SAE technical paper 2005-01-1236; 2005. http:// dx.doi.org/10.4271/2005-01-1236. [16] Cortese M. A study of rail pressure variation for various fuel injectors on a simple rail design over the engine operation range. SAE technical paper 200401-2937; 2004. http://dx.doi.org/10.4271/2004-01-2937. [17] Keskinen K, Kaario O, Tilli A, Hulkkonen T, et al. Improving the accuracy of 1-D fuel injection modeling. SAE technical paper 2012-01-1256; 2012. http:// dx.doi.org/10.4271/2012-01-1256.