Investigations of Internal Flow Characteristics of Multi-Hole Nozzle Using X-Ray Imaging Technique

Abstract

This study introduces a novel application of synchrotron X-ray phase contrast imaging to investigate the internal flow dynamics and liquid jet characteristics in a direct injection gasoline nozzle. Using optimized imaging parameters, including a 19 mm insertion gap and a 0.15 ns electron pulse (16 mA), we achieved high-quality visualization of needle motion and in-nozzle flip flow. The results show that cavitation appears rapidly with increasing needle valve lift, transitioning from unstable behavior below 40 µm to stable flip flow at higher lifts. The flip flow characteristics vary between nozzle holes due to differences in inlet angles. Internal flow velocity analysis reveals significant radial and axial gradients, with initial velocity overshoot during injection start followed by stable flow. The presence of flip flow accelerates jet breakup on the flip-contact side, leading to droplet–wall interactions in the counterbore. Different nozzle geometries, particularly hole inlet angle and length-to-diameter ratio, significantly influence jet width and velocity distributions. This comprehensive approach advances our understanding of practical nozzle internal flow dynamics and provides valuable insights for optimizing fuel injection system performance in engines.

1. Introduction

Fuel injection atomization in internal combustion engines is closely related to the fuel flow characteristics inside the nozzle and the liquid core break-up process near the nozzle exit. Due to the presence of the nozzle’s metal shell, visible light measurement techniques cannot directly observe and measure the flow process inside the nozzle. Current experimental research primarily utilizes transparent, scaled-up models of non-realistic nozzles for studies [1,2,3]. During research with non-realistic nozzles, significant differences exist between the injection pressure, the surface properties of the nozzle material, and those of actual nozzles. There is still uncertainty about whether the results obtained from non-realistic nozzle experiments can be directly applied to real nozzles. Additionally, when using non-realistic nozzles for research, the needle valve position is typically fixed, and the characteristics of the flow inside the nozzle are analyzed under a fixed valve lift [4,5,6]. This does not accurately reflect the dynamic impact of needle valve motion on the internal flow characteristics. These shortcomings are related to technical challenges such as the complexity of designing movable needle valve mechanisms and the difficulty of achieving high-pressure sealing.

Currently, visible light measurement techniques are mainly suitable for studying atomization characteristics after the liquid core breakup is complete in the downstream region of the spray, such as macroscopic characteristics like spray penetration length [7,8] and spray cone angle [9,10], and microscopic characteristics like particle velocity [11,12] and diameter [6,13]. In the liquid core break-up process near the nozzle exit, most of the fuel mass is concentrated in the liquid core. Surrounding the liquid core are a large number of smaller, discrete droplets. When visible light is used to observe the liquid core breakup, the incident light is scattered and absorbed by the surrounding droplets, which means only the outer contours of the droplet cluster can be observed, and the liquid core inside the droplet cluster cannot be imaged [2,14]. Therefore, the mechanisms of liquid core breakup are not yet fully understood in academic fields. New and effective experimental measurement techniques are needed to unravel this long-standing mystery.

In recent years, synchrotron X-ray technology has become an important tool for studying the jet liquid core break-up phenomenon. Compared to visible light-based methods like particle image velocimetry (PIV) or laser Doppler velocimetry (LDV) [15], synchrotron X-rays offer several distinct advantages. X-rays have a much shorter wavelength and are not scattered by droplets in the spray, allowing for clearer imaging of dense sprays and internal flow regions. Additionally, the high energy and brightness of synchrotron X-rays enable high temporal and spatial resolution, making them ideal for investigating small-scale, transient phenomena within practical stainless nozzles. Based on this testing technology, considerable research has already been conducted. Kastengren et al. [16] noted significant differences in near-nozzle spray density between hydroground and non-hydroground diesel injectors using X-ray radiography measurements. Moon et al. [17] revealed the dynamic structure of biodiesel and conventional fuel sprays from high-pressure diesel injectors using an X-ray phrase contrast imaging technique. Sou et al. [18] visualized the cavitation in modeling nozzles with various sizes using the X-ray phrase contrast imaging technique, and they noted that in the smaller nozzles, cavitation tends to be slightly thinner and therefore shorter due to the relatively larger radii of curvatures for the nozzle inlet edges normalized by the nozzle width. Despite many investigations using synchrotron X-rays, there has been limited widespread reporting on the use of synchrotron X-ray measurement technology to study the internal flow characteristics of actual nozzles. Especially in terms of experimental research that combines needle valve motion and internal flow characteristics with the liquid core break-up process, relevant literature is rare.

This paper utilizes synchrotron X-ray measurement technology to conduct in situ experimental research on the internal flow and jet liquid core break-up characteristics of a real direct injection gasoline engine nozzle under needle valve motion at the Advanced Photon Source (APS), 7ID-B beamline at Argonne National Laboratory in the United States. During this research, images were first captured under various X-ray source parameter conditions, and the optimal imaging technique was determined by comparing the captured images. Based on this, further experimental measurements and research were conducted, analyzing the inner-nozzle flow dynamics and their impact on liquid jet break-up. This research provides important experimental data to support the study of jet liquid core break-up mechanisms.

2. Methods and Experiment Setup

The synchrotron radiation X-ray phase contrast imaging method and the experiment system will be briefly introduced.

2.1. Methods

When electrons move through a synchrotron electron storage ring, they emit X-rays along the direction of their motion whenever the direction of their movement changes. These X-rays pass through the beamline’s focusing and adjustment processes before entering the experimental beamline. Researchers can then use these X-rays to conduct various experimental measurements.

As X-rays pass through a test sample, the incident energy is absorbed by the sample, creating an absorption contrast. The absorption contrast is related to the mass of the sample that the X-rays pass through. When studying fuel spray atomization characteristics using X-ray absorption contrast, the concentration distribution of the fuel spray can be calculated based on the absorption contrast [19]. However, the diameter of the fuel spray typically ranges from several millimeters to tens of millimeters, while X-rays have high energy and strong penetration ability, resulting in relatively weak attenuation of the X-ray energy after passing through the fuel spray. In practical experiments, it is necessary to conduct hundreds of repeated measurements and accumulate the results to achieve a better signal-to-noise ratio. Therefore, this method is not suitable for observing and studying the dynamic process of fuel spray. Many studies have been conducted using synchrotron X-ray techniques [20,21]

In addition to absorption contrast, the phase of the X-rays is also altered as they pass through the test sample, resulting in phase contrast. Wang et al. [22] reported that for less-dense materials, the phase contrast difference after the X-rays pass through the sample is about three orders of magnitude higher than the absorption contrast. Additionally, Wang et al. [22] also found that as the X-ray energy increases, both the phase contrast and absorption contrast decrease, but the decrease in phase contrast is less pronounced compared to the absorption contrast. Furthermore, phase contrast usually reaches its maximum at the interfaces between different media. Therefore, phase contrast is particularly suitable for observing and studying high-transient, multiphase flow states, such as the liquid core breakup process in fuel jets.

2.2. Experiment Setup

All experiments in this study were conducted at the 7ID-B beamline of the Advanced Photon Source (APS) at Argonne National Laboratory in the United States. The main optical parameters can be found in reference [23]. The spray measurement system built at the 7ID-B beamline is shown in Figure 1. The system mainly includes the following components: X-ray imaging and capture system, signal synchronization and control system, fuel pressurization system, and spray chamber.

The synchrotron X-ray has a high energy, high power density per unit area, and a high pulse frequency. Mechanical shutters are required to reduce the duration of X-ray exposure to the imaging system, typically down to around 10 ms, to control the thermal load on the imaging device and prevent damage. The fuel pressurization is achieved using an M-71 gas pressurization pump manufactured by HASKEL (Burbank, CA, USA). The maximum fuel pressure at the pump’s outlet can reach 60.7 MPa. X-ray phase contrast cannot be directly captured using conventional optical imaging devices, so special optical crystals called scintillators are added to the optical path. These scintillators convert the X-ray phase contrast into visible light intensity differences. The Photron (Tokyo, Japan) high-speed camera and scintillator are typically arranged at a 90° angle to each other, with a 45° mirror installed between them. This setup reduces noise in the images and prevents direct X-ray exposure to the optical lens of the high-speed camera, which could cause damage.

The spray chamber has a rectangular structure. Observation windows and window bodies are installed on both long sides of the chamber. As X-rays propagate through the air, they are absorbed by the air, leading to a decrease in brightness. To reduce this attenuation, an aluminum tube is installed along the optical path, and a vacuum is applied. Additionally, Kapton film is used as the window material for all the windows in the tubing and spray chamber. Kapton film is a thin polyimide material with a very low X-ray absorption coefficient, which helps to reduce X-ray brightness attenuation and improve the quality of the captured images. However, the pressure and thermal resistance of Kapton film is much lower than that of traditional spray chamber window materials such as quartz glass or sapphire glass. In this study, the actual size of the Kapton film window in the spray chamber is 12 × 30 mm, and its design maximum working pressure is 2 MPa (gauge pressure).

To ensure accurate calibration and alignment during the experiment, high-resolution stepper motors were used to precisely control the positions of the camera and spray chamber, maintaining the alignment accuracy required for high-resolution imaging. Thermal effects on the scintillators posed a significant challenge, as X-ray exposure could damage the scintillator material within microseconds under maloperation. To address this, strict operational procedures were implemented, although occasional damage to the scintillators still occurred, reducing experimental efficiency and increasing costs. Additionally, due to the hazardous radiation emitted by synchrotron X-rays, all experiments were conducted remotely. The remote control and observation systems were carefully designed to ensure safe operation and minimize the risk of inefficiency or system failure.

2.3. Experiment Conditions

This research utilizes an electromagnetic valve-type injector for gasoline direct injection engines, featuring a custom-designed three-hole nozzle made from aluminum, specifically fabricated for experimental purposes. The choice of aluminum over steel is due to its superior X-ray transmission properties, which enables more photons to penetrate the nozzle’s metal casing, resulting in enhanced image quality. The aluminum nozzle is expected to replicate the internal flow characteristics of conventional steel nozzles (used in production models) for three key reasons. Firstly, the primary determinants of spray hole internal flow characteristics are the nozzle’s structure, flow conditions, and material properties, with the aluminum nozzle manufactured to match the steel nozzle’s specifications, maintaining a virtually identical structural design and dimensions. Secondly, the aluminum nozzle incorporates a steel needle valve and other steel injector components, ensuring reliable performance and maintaining stable internal flow conditions that accurately simulate the flow processes found in steel nozzles under operational conditions. Thirdly, both aluminum and steel nozzles were manufactured using the same drilling process, which typically yields similar surface finishes; thus, their surface roughness and energy characteristics are assumed to be comparable [24]. While aluminum may differ from steel in terms of thermal conductivity, expansion coefficients, and structural responses under pressure, these factors are not significant in this study, as the experiments focus on flow dynamics under sub-cooled conditions rather than thermal or mechanical effects. Based on these factors, it is reasonable to conclude that the internal flow characteristics of the spray holes in the aluminum nozzle should be essentially the same as those in the steel nozzle.

The three-hole aluminum injector nozzle features one vertical spray hole and two inclined spray holes. The two inclined spray holes have the same structure. The angles between the three spray holes are 120°, symmetrically distributed in space. The use of a three-hole structure is beneficial for experimental measurement and observation while also maintaining consistency with the internal flow state of multi-hole injectors. All spray holes are designed with a counterbore structure divided into two stages: the spray hole and the counterbore. This design is commonly used in direct-injection gasoline engine nozzles. Its main purpose is to reduce the length-to-diameter ratio of the spray hole to achieve faster jet breakup and atomization.

The structure of the nozzle is shown in Figure 2. $\beta$ represents the spray hole inclination angle, i.e., the angle between the spray hole axis and the vertical direction. $D_h$ and $D_c$ represent the diameters of the spray hole and counterbore, respectively. $l_h$ and $l_c$ represent the lengths of the spray hole and counterbore, respectively. The specific parameter values for the spray hole and counterbore structure are shown in

Table 1. Nozzle geometries.

Number	Hole Diameter ${\mathit{D}}_{\mathit{h}}$/mm	Hole Length ${\mathit{l}}_{\mathit{h}}$/mm	Counterbore Diameter ${\mathit{D}}_{\mathit{c}}$/mm	Counterbore Length ${\mathit{D}}_{\mathit{c}}$/mm	Inclination Angle $\mathit{\beta }/$°
#1	0.13	0.26	0.40	0.38	0
#2	0.13	0.18	0.40	0.42	42
#3	0.13	0.18	0.40	0.42	42

. Since the X-ray incident direction and position are fixed during the experiment, it is necessary to rotate the spray chamber to make the cross-section of the spray hole to be measured perpendicular to the X-ray incident direction before capturing the measurement.

In this experiment, the fuel injection pressures were set to 8, 15, and 20 MPa, which are typical injection pressure ranges for direct injection gasoline engines. Each test point was repeated five times. The ambient gas was nitrogen (N₂), maintained at constant temperature and pressure. The injection pulse width was fixed at 2 ms. The high-speed camera was set to capture images at a rate of 67,889 frames per second (fps). The image resolution was 512 × 512 pixels, with a single pixel size of approximately 2.54 μm. n-Heptane was used as the test fuel. n-Heptane is commonly used as a surrogate fuel to study early combustion and knocking characteristics in gasoline engines. Additionally, n-Heptane has a similar viscosity and surface tension properties to gasoline. Considering these factors, n-Heptane was chosen as a substitute for gasoline in the experimental measurements. The main experimental parameters are summarized in

Table 2. Fuel injection experiment conditions.

Parameters
Injection pressure/MPa	8, 15, 20
Injection duration/ms	2
Environmental gas	N2
Environmental pressure/MPa	0.1
Environmental temperature/℃	25

.

3. Results and Discussion

3.1. Effects of X-Ray Parameters on Inner-Nozzle Imaging

To observe the internal flow characteristics of the spray holes in a metal injector nozzle, X-rays need to pass through the metal casing to image its interior. During this process, the energy and brightness of the X-rays are absorbed by the metal shell of the nozzle, leading to an insufficient number of photons reaching the camera film, which results in imaging failure. To address this issue, one solution is to extend the camera’s shutter time. However, the internal flow pattern of the spray hole changes continuously during the injection process. If the exposure time is too long, dynamic blur may occur, making it difficult to accurately capture the instantaneous internal flow pattern. Therefore, the experiment must be conducted with short shutter times while increasing the X-ray energy and brightness to achieve effective imaging.

However, the higher the X-ray energy and brightness, the greater the thermal load generated, which can damage the imaging components. Additionally, as reported by Endrizzi et al. [25], the phase contrast generated by the X-rays passing through the sample is inversely proportional to the square of the X-ray energy. In other words, the higher the X-ray energy, the lower the phase contrast that can be provided. Therefore, before conducting formal measurements, it is essential to optimize the X-ray source to achieve the best possible image quality.

During the study, imaging of the internal flow of the injector nozzle spray holes was conducted under four different X-ray source parameter conditions, as shown in

Table 3. Imaging experiment conditions.

Parameters	Test 1	Test 2	Test 3	Test 4
Insertion gap/mm	15	19	19	25
Electron bunch mode	dual	dual	single	single
Bunch pulse duration/ns	78	78	0.15	0.15
Electron current/mA	11 × 2	11 × 2	16	16
Exposure time/µs	0.35
Frequency rate/fps	67,889

. One of the key parameters in the table is the insertion gap, which is related to the control of the X-ray source. The smaller the insertion gap, the stronger the magnetic field, resulting in a higher proportion of high-energy X-rays being produced. The electron bunch mode refers to the number of electrons in the storage ring and the bunch mode, with options such as single bunch, dual bunch, and multiple bunch modes for imaging. The greater the number of bunches, the more electrons they contain, leading to higher X-ray brightness. However, there is a time gap of approximately 50 nanoseconds between electron bunches. When using dual or multiple bunch modes for imaging, it is equivalent to performing multiple exposures on the same image.

Using the simulation software Spectra [26], the energy and brightness distribution spectra of the X-rays under the four parameter conditions in Table 3 were calculated, as shown in Figure 3. The exposure time and frequency rate were kept the same across all conditions. The X-ray pulse duration is shorter than the exposure time, ensuring that the photons entering the camera vary based on the X-ray pulses rather than the exposure time. The calculation results indicate that, at the same energy, the brightness of the X-rays decreases progressively from Test 1 to Test 4, meaning that the number of photons decreases. The X-ray brightness is slightly higher under the dual-bunch condition compared to the single-bunch condition. When the insertion gap is set to 25 mm, the brightness of the high-energy portion of the X-rays (greater than 40 keV) decays rapidly.

These findings highlight the need for careful optimization of the X-ray source parameters, including the insertion gap, electron bunches, and exposure time, to balance photon intensity, image quality, and resolution when studying dynamic processes like the internal flow characteristics of injector nozzles.

To enhance the clarity of the X-ray images, a multi-step preprocessing procedure was applied to all analyzed images. Taking the results of Test 1 as an example (Figure 4), the image undergoing analysis (Figure 4b) was first divided by the corresponding background image (Figure 4a) to remove background noise, producing a processed image without gray-level adjustment (Figure 4c). However, the internal flow features were still difficult to distinguish. To further enhance visibility, gray-level adjustment was performed using histogram equalization [27], resulting in the final processed image with improved contrast and clarity, which was used for analysis. These preprocessing steps were consistently applied across all images to ensure robustness and reliability, as illustrated in Figure 4.

Figure 5 presents the inner-nozzle images of injector nozzle hole #1 obtained under different imaging conditions. The injection pressure for this set of experiments was 15 MPa. The images were captured at 1 ms after the start of injection, corresponding to the stabilized phase of the jet flow. Additionally, the images underwent preprocessing operations such as background subtraction and contrast adjustment to enhance the clarity of the internal flow observations.

In Figure 5, the internal structure of the injector nozzle, including the needle valve, pressure chamber, spray hole, and counterbore, is clearly shown. High-speed imaging techniques allow for in situ measurements of the needle valve movement, discussed in Section 3.2.1. The image also reveals the instantaneous fuel flow patterns from the nozzle hole inlet to the outside of the counterbore.

Test 3 was determined to provide the best image quality regarding the inner-nozzle flow imaging for the following reasons: the X-ray brightness was higher in Test 3 than in Test 4, allowing for better penetration of the injector nozzle’s metal casing. Compared to Test 1, Test 3 had a lower proportion of high-energy X-rays, improving phase contrast and enhancing the visibility of internal flow. Test 3 also used a single-pulse exposure with a 0.15 ns pulse width. With a jet speed of 200 m/s, the displacement during exposure was only 0.03 ms, which was smaller than the pixel size (2.54 ms), eliminating dynamic blur and ensuring a quasi-static flow representation. These factors make Test 3’s images the clearest and most accurate, as discussed in Section 3.2.2.

In Test 1 and Test 2, double-pulse exposure creates regular striped patterns resulting from jet surface displacement between pulses. Using a method similar to particle image velocimetry (PIV), in situ jet velocity measurements can also be performed, as outlined in Section 3.2.3.

3.2. Results of Inner-Nozzle Dynamics

3.2.1. Inner-Nozzle Needle Motion

By combining synchrotron X-rays with high-speed imaging, continuous images of the injector nozzle interior can be captured, allowing for calculation of the needle valve’s motion. The Test 3 conditions were used for simultaneous in situ measurement of needle valve motion and internal flow. Figure 6 shows the calculation method: First, a query domain is defined on an image before injection, including part of the needle valve at rest (Figure 6a). In the moved image, a larger scanning domain is defined, covering the range of needle valve displacement. A matching domain, equal in size to the query domain, is extracted from the scanning domain (Figure 6b). The covariance of the grayscale matrices of the matching and query domains is calculated to obtain the correlation coefficient. This process is repeated step by step, and the maximum correlation coefficient indicates the needle valve’s position (Figure 6c). Based on the matching domain coordinates, axial displacement and radial offset can be calculated. For a detailed explanation of the method, please refer to the relevant references [28,29].

Figure 7a,b show the variation in the needle valve lift and its lift speed over time under different injection pressures. Figure 7a reveals a noticeable overshoot during the needle valve lift, likely due to the deformation of the needle valve rod. When the solenoid valve is energized, the armature drives the needle valve upward until it contacts the armature stopper. At this point, inertia causes the needle valve rod to bend, resulting in overshoot. Subsequently, the needle valve gradually recovers its shape, ultimately settling at its set position. Under different injection pressures, the lift curves during the lifting and closing stages of the needle valve remain consistent, as shown in Figure 7a. The needle valve’s motion is primarily influenced by the force from the solenoid valve and is independent of the fuel pressure in the injector. Once the needle valve is fully opened, fuel injection enters a steady state. At this point, as the injection pressure increases, the needle valve’s lift slightly increases. This phenomenon may be related to changes in the contact between the armature and the armature stopper. Experiments from our other investigations have shown that the contact between the armature and the stopper is not completely sufficient. The increase in fuel pressure improves this contact, making it tighter, which slightly raises the armature’s opening height and consequently increases the needle valve’s opening. These observations were obtained from commercial experimental investigations conducted as part of this study; however, due to proprietary constraints, specific details of these experiments are not publicly available. The results in Figure 6b show that the needle speed remains nearly unchanged under different injection pressures.

3.2.2. Inner-Nozzle Hydraulic Flip

Figure 8a–f illustrates the variation in the fuel flow state within hole #2 at a fuel injection pressure of 15 MPa over time. As shown in Figure 8a, with the needle valve opening, a gas–liquid phase separation interface is observed on the left side of the nozzle hole, indicating cavitation formation inside the nozzle hole. At this stage, the tail of the cavitation is not clearly defined. The cavitation will reattach near the nozzle outlet. Once the fuel exits the nozzle and enters the counterbore, it expands rapidly and almost covers the entire counterbore exit, as shown in Figure 8b. A fuel wall impact phenomenon is observed near the left-side exit of the counterbore. It should be noted that the unclear gas–liquid phase separation interface in Figure 8a,b may be due to high turbulence intensity in the nozzle flow at low needle valve openings, which causes unstable cavitation and complex deformations in the nozzle. In the current imaging setup using the line-of-sight method, when cavitation is unstable and the nozzle flow has a complex twisted structure, part of the emitted light overlaps, affecting the resolution of the cavitation gas–liquid interface.

As the needle valve opening increases further, the gas–liquid phase separation interface on the right side of the nozzle hole extends from the nozzle inlet to the nozzle outlet, and the interface becomes clearer, indicating a stable internal flow state, as shown in Figure 8c,d. These features indicate the formation of hydraulic flip, abbreviated as “flip flow”. Sou et al. [30] noted that in short nozzles (with a length-to-diameter ratio of 2), initial cavitation composed of tiny gas bubbles can reach the nozzle outlet. Even at low flow rates, hydraulic flip can form inside the nozzle, rather than cavitation. This analysis explains the rapid formation of flip flow in nozzle #1 (with a length-to-diameter ratio of 2). Furthermore, we observed that once fuel injection enters the steady-state stage, the internal flow and jet state inside the nozzle remain stable. Bode et al. [31] pointed out that when flip flow forms inside the nozzle, the resistance and initial disturbance to the fuel flow are reduced, leading to a smaller jet angle and improved stability.

In the end-of-injection stage, as the needle valve closes, the flip flow inside the nozzle gradually loses its stability, as shown in Figure 8e. The flip flow reattaches near the nozzle outlet, and the jet in the counterbore expands significantly. When the needle valve is fully closed, as shown in Figure 8f, the high-pressure rail fuel supply is cut off, and the jet flow rate rapidly decreases until it completely halts.

Figure 9a–f show the variation in the fuel flow state within hole #2 at a fuel injection pressure of 15 MPa over time. The results indicate that the internal flow pattern of hole #2 follows a similar trend to that of hole #1. The difference lies in the fact that the flip flow in hole #2 occurs on the side farther from the nozzle sac, and the width of the flip flow is narrower than that in hole #1. This phenomenon is likely related to the smaller angle of hole #2 relative to the pressure chamber. As a result of the flow separation location, the liquid core of the jet in hole #2 is slightly shifted toward the nozzle sac side.

Additionally, from Figure 9e, it can be seen that when the needle valve is nearly closed, cavitation occurs on the side of the nozzle near the pressure chamber. This clearly demonstrates the change in the direction of the fuel flow inside the nozzle. During the needle valve opening, fuel primarily flows from upstream of the needle valve, and cavitation or flip flow occurs on the left side of the nozzle (the sharp-edged inlet side). When the needle valve closes, the upstream flow is blocked by the needle valve, and fuel enters the nozzle from the nozzle sac side, leading to cavitation near the nozzle sac side.

Figure 10 illustrates the variation in the flip flow widths within hole #1 and hole #2 over time. Figure 10a,c define the flip flow widths within hole #1 and hole #2, respectively. Figure 10b shows how the flip flow width in hole #1 changes over time under different injection pressures. The results indicate that as the injection pressure increases from 8 MPa to 15 MPa, the flip flow width at the hole outlet slightly decreases. At injection pressures of 15 MPa and 20 MPa, the flip flow width is nearly identical. Pratama et al. [32] reported that increasing the injection pressure can gradually increase the length of cavitation inside the nozzle until flip flow forms. However, this study found that with a short nozzle (length-to-diameter ratio of 2), the nozzle hole structure significantly influences the internal flow, causing it to quickly reach saturation. Even at a lower injection pressure (8 MPa), flip flow forms rapidly within the nozzle hole, and increasing the injection pressure does not significantly alter the internal flow characteristics.

Figure 10d compares the variation in flip flow width over time between hole #1 and hole #2 under 15 MPa conditions. As shown, the variation in flip flow width for hole #2 follows a similar trend to that of hole #1, but the flip flow width in hole #2 is smaller than that in hole #1.

Figure 11 illustrates how the flip flow width in hole #1 and hole #2 changes with needle valve lift at an injection pressure of 15 MPa. It can be seen that at partial needle valve openings, the measurement variance is high across multiple tests, indicating that the flip flow state inside the nozzle is unstable at low needle valve lifts. This instability continues until the lift reaches about 40 μm. Beyond this point, the flip flow width stabilizes, and the measurement variance reduces. Even during needle valve overshoot, the flip flow width remains largely unchanged, with minimal variance. During the closing process, when the needle valve lift drops below around 20 μm, the flip flow width decreases rapidly, and the measurement variance increases again. This instability may be due to the gradual reduction in fuel pressure within the nozzle’s sac chamber during needle valve closure.

Additionally, a comparison of the variation in flip flow width at the exits of hole #1 and hole #2 reveals that, under the same needle valve lift, the formation and variation of flip flow in hole #2 are slightly delayed compared to hole #1, and the maximum flip flow width is smaller. This is closely related to the inlet direction and angle of the orifice. Hole #1 is a vertical orifice, allowing fuel to enter more easily from the upstream side, which leads to quicker formation of flip flow. Meanwhile, the inlet angle of hole #1 is sharper, resulting in a more significant change in the fuel flow direction. This causes stronger gas–liquid phase separation inside hole #1, leading to a larger flip flow width. These findings suggest that adjusting the orifice inlet angle is an effective method for controlling the internal flow state within the nozzle, thereby influencing the breakup behavior of the jet liquid core.

3.2.3. Inner-Nozzle Jet Velocity

Under the experimental conditions of Test 2, in situ measurements of the injection velocity inside the injector’s orifice were conducted using synchrotron X-ray. However, during the analysis, it was found that the results had poor signal-to-noise ratios and insufficient valid data. To address this issue, this study first analyzes the impact of the internal flow velocity algorithm on the calculation results.

The dual-pulse synchrotron X-ray velocity measurement does not require tracer particles but instead uses the fuel jet’s structure to calculate the velocity. During the injection or spray breakup, the spray surface becomes unstable due to shear forces and fuel turbulence, causing liquid core breakup and forming discrete droplets. Using X-rays in dual-exposure mode, these surface distortions or droplets are captured twice on the same film (Figure 12a), providing dual-exposure images.

After acquiring the dual-exposure images, preprocessing steps such as background subtraction and normalization are first performed. Then, a computational domain is defined in the image, and the autocorrelation algorithm is used to extract the velocity of the characteristic structures within the domain. There are several algorithms for autocorrelation calculation, and in this study, a fast Fourier transform ($FFT$)-based algorithm is used to obtain the correlation matrix for the dual-exposure images, as shown in Equation (1):

$$m\times m={FFT}^{-1}\left[FFT\right(m)\times FFT(m,rot180\left)\right]$$

(1)

where $m$ represents the computational domain, and m × m refers to the autocorrelation calculation within the domain. $FFT$ and ${FFT}^{-1}$ represent the forward and inverse Fourier transforms, respectively. $(m,rot180)$ indicates that the matrix of the computational domain is inverted. The results of the autocorrelation calculation are shown in Figure 12b.

According to the basic principle of autocorrelation calculation, the peak values in the results matrix represent the initial position of the observed object. The relative position between the secondary peak and the primary peak indicates the displacement and direction of the observed object during the dual-exposure time interval, as shown in Figure 11c. Once the secondary peak location is found, the velocity and direction of the observed object can be calculated using the method in Equation (2):

$$u=S_a/∆t;v=S_r/∆t;∆t=68ns$$

(2)

where $S_a$ and $S_r$ represent the displacement of the observed object in the axial and radial directions, respectively, and $u$ and $v$ represent the axial and radial velocities of the observed object. By moving the computational domain across the entire image and repeating the calculation process, the velocity distribution of the observed objects throughout the field can be obtained.

Previous studies have highlighted issues such as poor signal-to-noise ratios and insufficient valid data when calculating internal flow velocity from acquired images. To address these, this study analyzes the impact of the computational domain area on velocity calculation accuracy. The maximum injection pressure in this study is 20 MPa, leading to a theoretical jet velocity of 241 m/s at the nozzle exit, which corresponds to a maximum internal flow velocity of 241 m/s. In the X-ray dual-pulse exposure mode, the time interval between the two pulses is 68 ns, resulting in a maximum displacement of 16.4 µm (approximately 6.5 pixels). This defines the lower bound for the computational domain length. The radial displacement, typically no more than one-third of the axial velocity, is limited to 5.5 µm (about 2 pixels), setting the lower bound for the computational domain width. Therefore, the computational domain should be at least 6×2 pixels.

Using hole #1 as an example, this study examines the effects of domain area on the calculation results (Figure 13). Figure 13a shows the measurement positions, including the center ($r/R=-0.29$) and outer edge ($r/R=0.95$) of the flow. As shown in Figure 13b, increasing the domain width slightly reduces both axial and radial velocities and increases the variance. This is due to the significant velocity gradient in the radial direction, where low-speed areas are included as the domain width increases, lowering the calculated velocity. Reducing domain width increases the computational load but improves efficiency. Thus, a domain width of 10 µm is optimal. Figure 12c shows that increasing the domain length slightly increases the axial velocity and reduces variance. This improves calculation reliability, and a domain length of 40 µm is recommended.

In conclusion, this study suggests using a computational domain size of 10×40 pixels for optimal velocity calculation results. The algorithm optimization improves calculation accuracy, signal-to-noise ratio, and valid data extraction.

Figure 14 shows the variation in internal flow velocity in hole #1 with respect to the fuel injection duration. The measurement location of this internal flow velocity is at the center of the flow (r/R = −0.29) shown in Figure 13a. As seen in Figure 14, after the injection begins, the internal flow velocity rises rapidly and quickly reaches a stable state. During the initial phase of the velocity rise, which occurs between 0.5 ms and 1.0 ms of injection duration, there is a slight overshoot in the internal flow velocity. After this, the flow enters a steady state, with the internal flow velocity remaining almost constant, corresponding to an injection duration between 1.0 ms and 2.0 ms. The overshoot in internal flow velocity may be attributed to the needle valve’s own opening overshoot.

Figure 15 shows the radial distribution of fuel flow at different positions inside the nozzle and the counterbore: 200 µm (inside the counterbore), 400 µm (counterbore exit), and 600 µm (outside the counterbore). For convenience, these flow velocities are collectively referred to as internal flow velocities of the nozzle. The measurement locations for the internal flow velocities are shown in Figure 15a.

From Figure 15b, it can be seen that the axial velocity of the internal flow is relatively higher on the negative side of the radial position. On the positive side, the internal flow velocity gradually decreases along the radial direction. As the flow propagates downstream, the velocity at the center of the flow remains nearly constant, while the velocity at the outer edge increases slightly. This result is likely related to the formation of flip flow and jet contraction (vena contracta) inside the nozzle. Due to jet contraction, the fuel is compressed inside the nozzle, and upon leaving the nozzle, it experiences elastic recovery, causing a slight increase in the velocity at the outer edge of the flow. From the radial velocity data, at a location close to the nozzle exit (200 µm), the outer edge has a higher radial velocity. This result further suggests that the elastic recovery of the fuel as it leaves the nozzle leads to a partial increase in internal flow velocity. As the fuel continues to propagate downstream, the spray expands, and the velocity decreases again.

3.3. Effects of Inner-Nozzle Flow Dynamics on Liquid Jet Characteristics

Figure 16 shows an analysis of the effects of the internal flow on the jet characteristics under an injection pressure of 15 MPa. As shown in Figure 15a, the flip flow width in hole #1 is wider than in hole #4. Additionally, due to the longer nozzle length of hole #1, the flip flow length inside hole #1 is also greater than that in hole #4. From Figure 16b, it can be seen that the jet width from hole #1 is slightly larger than that from hole #4. Both hole #1 and hole #4 show flip flow on their left sides. After exiting the nozzle, the fuel rapidly forms small droplets, which collide with the right wall of the counterbore. Consequently, the left-side boundary of the jets from both hole #1 and hole #4 aligns with the counterbore boundary. On the right side of the jets, away from the flip flow, the jet does not collide with the counterbore wall. Therefore, the jet boundary on the right side of hole #1 is farther out than that of hole #4, making the jet width from hole #1 larger than from hole #4. Fuel–wall collisions affect emissions and nozzle durability. Preliminary results show that counterbore diameter and initial conditions, like fuel and nozzle temperatures, play key roles. Further details will be shared in future studies.

The fuel closer to the flip flow side is more prone to breakup, likely due to the strong interaction between the fuel in the flip flow and the surrounding air. Simulations performed by Itaya et al. [33] suggest that air entering the nozzle on the flip flow side accelerates the breakup of the liquid core surface, leading to the formation of discrete droplets. Once these droplets detach from the main jet, they further expand radially and collide with the counterbore. The speed, direction, and size of the droplets after impact with the counterbore wall may be significantly altered. During actual engine operation, such phenomena could potentially contribute to increased emissions and nozzle carbon buildup, which warrants further investigation. Optimization of nozzle structure design, such as adjusting the length-to-diameter ratio and inlet angle, could help mitigate the issue of fuel–wall collisions in the counterbore.

On the side away from the flip flow, the fuel is pushed toward the nozzle wall by the flip flow. The larger the flip flow width, the stronger the pushing force. After exiting the nozzle, the fuel, no longer constrained by the nozzle wall, expands outward under internal pressure, or it undergoes elastic recovery. The wider the flip flow inside the nozzle, the greater the compression of the fuel, leading to a more pronounced elastic recovery once the fuel exits the nozzle. This likely explains why the left-side boundary of the jet from hole #1 is farther out than from hole #4.

Figure 16c shows the velocity distribution of the jets from hole #1 and hole #4 from the nozzle exit to the outer side of the counterbore. The internal flow velocity in hole #1 exhibits more noticeable spatial variation, likely due to the wider flip flow width in hole #1. In contrast, the internal flow velocity in hole #4 changes more gradually. However, it is worth noting that in the 100–200 µm range from the nozzle exit, the internal flow velocity in hole #4 is relatively lower. This could be due to the shorter length-to-diameter ratio of hole #4, which causes jet contraction to complete farther from the nozzle.

4. Conclusions

This paper utilizes synchrotron X-ray phase contrast imaging technology to conduct a in situ investigation of the inner-nozzle flow dynamics and liquid jet characteristics in a direct injection gasoline nozzle. The main conclusions of the research can be summarized as follows:

• The quality of internal flow imaging in metal fuel injectors depends on X-ray source energy, brightness, and pulse width. The best image quality was achieved with a 19 mm insertion gap and a single 0.15 ns electron pulse (16 mA), contributing to simultaneous observation of needle motion and in-nozzle flip flow. Dual-pulse X-ray imaging allows for in situ measurement of jet velocity within the nozzle.
• As the needle valve opening increases, cavitation appears within the nozzle hole rapidly. Below 40 µm of needle lift, cavitation is unstable, but once the needle valve opens further, a stable flip flow forms. The flip flow width in hole #2 is smaller than that in hole #1, likely due to the sharper inlet angle of hole #1. Increasing the injection pressure does not significantly alter the internal flow characteristics.
• Internal flow velocity has significant radial and axial gradients. The optimal calculation domain size is 10 × 40 pixels. During the initial injection stage, the internal flow velocity exhibits some overshoot, likely due to needle valve overshoot. In the stable phase of injection, the internal flow velocity remains almost constant. Radial velocity distributions are asymmetric, with velocity initially increasing and then decreasing as the flow progresses downstream.
• Flip flow accelerates jet breakup on the flip-contact side, causing detached droplets to collide with the counterbore wall. The jet width from hole #1 is larger than from hole #4. Hole #1, with a wider flip flow, shows more significant spatial variation in jet velocity. In contrast, hole #4, with a shorter length-to-diameter ratio, experiences delayed jet contraction, resulting in relatively lower internal flow velocity within 100–200 µm of the nozzle exit.