YANG Guoan, YIN Xin, , SONG Zheng, and HUANG Cong

1 Diagnosis & Self-recovery Engineering Research Center, Beijing University of Chemical Technology,Beijing 100029, China

2 Production Operations Center, PetroChina Guangxi Petrochemical Company, Qinzhou 535008, China

1 Introduction

The drilling pump valve is one of the most critical components in a drilling pump, having a direct impact on the efficiency and reliability of the whole machine.However, due to the complex working conditions in practical use, the service life of the valve is the shortest among all the wearing parts of a drilling pump, which restricts the development of the drilling pump in the direction of high-pressure, high-stroke and high-output.

Over the decades, a plenty of studies were carried out on the issue of valve failure, however, due to the effect of a complex system of forces on the valve, there are difficulties in the theoretical research and experimental study. Judging from the literature, it is found most of the studies on the pump valve were directly aimed at the valve itself,concentrating on exploring the mechanism of valve failure and improving valve design. Especially in the discussion of the failure mechanism, many points of view have been presented. For example, JOFFE, et al[1], believed that the main causes of the valve failure were abrasive wear, fatigue wear as well as the combined effect of cavitation and fatigue wear; in Ref. [2], analyses are made respectively on different parts of the valve, and it is believed that the causes of valve failure are different for each part of the valve; With the quantitative calculation of fault tree and an analysis of the damaged valve surface, WANG, et al[3], presented that the slurry erosion and the material of the valve had the greatest impact on the valve failure; differently, as to the work of XIA, et al[4], the study was from the perspective of impact fatigue, putting forward experimental method to measure the impact force. Researches have shown that many cases of valve failure are caused by the erosion of slurry. However, since there is no experimental study special on the flow field of the valve play, it is much lacking in the knowledge of the flow field characteristics,and thus research work in this aspect is hard to get furthered.

In this paper, a particle image velocimetry(PIV)measurement of the flow field in the model valve was performed to experimentally investigate the characteristics of the flowing fluid in the valve play, which verified and improved the previous computational fluid dynamics(CFD)simulations. This paper firstly introduces the preparation work for the measurement, including the design of the experimental setup, particle selection, and adjustment of PIV devices, such as the camera and the laser. Then,analyses are made on the obtained PIV data for different working conditions and conclusions are drawn from these results. After that, in order to get more precise data, an advanced PIV test with fluorescent particles is introduced as a supplement. Finally in this article, the measuring accuracy is discussed and the conclusions are summarized.

2 Design of Experimental Setup

Fig. 1 shows a sketch of the experimental setup. This experimental setup basically consists of a pump, a tank,pipelines and a model valve with three pairs of scaled valve discs and valve seats. Among all these parts, design for the body of the model valve and the valve discs and seats therein are keys to the whole installation, taking the 7#drilling pump valve commonly used in oilfields as a reference and scaled down with proportion of 1:4. These optically transparent models are made of acrylic glass with polished external and internal surfaces, mainly assembled with three parts which are the upper chamber, the lower chamber and the valve disc and seat in between.

Fig. 1. Sketch of the experimental setup

Considering the structure of the model and the fluid pressure under the working condition of the experiment,there are two assumptions introduced in the design of the models: (1) stress σrin the radial direction is negligible, as the body model is a thin-walled vessel and the scale of fluid pressure therein is not higher than 1 MPa; (2) the impact stress occurring with valve closing could be neglected,because the shock of the valve disc is tiny in the experimental model. On the basis of the proportion mentioned above and the design criterion for pressure vessels[5], the figure and structure of the designed model are decided as what’s shown in Fig. 2. Fig. 3 shows the model valve discs and seats. As shown in Fig. 4, the external wall of the model is designed as square, which helps to minimize the refraction at the surface.

3 Experimental Procedures

A two-dimenional PIV system was used to measure the velocity of the flow within the model. The light sheet was emitted by a integrated high power Nd-YAG duolaser, and the images of the particles were recorded by a chargecoupled device(CCD) camera, Redlake Megaplus II ES2001(resolution 1 600×1 200 pixels, frame rate 31.85 f/s). The operation of the laser and the camera was synchronized by a digital Delay/Pulse Generator, model MicroPulse710.Since the region to be photographed was as small as an area of 20 mm×20 mm and the interrogation window was even narrower and smaller of which the maximal width was less than 5 mm and the flow speed thereof was much higher, a conventional camera lens was not suitable for this shooting.So, a microfocus lens (model Nikon AF-S VR Micro-Nikkor 105 mm f/2.8G IF-ED) was used to get enlarged images.

Fig. 2. Designed model

Fig. 3. Photos of model valve discs & seats

Fig. 4. Cross section of the model

The particle type selection and the particle seeding are two crucial issues for PIV measurement. To select a suitable type of particles for this test, the scattering performance and the uniformity of the beads dispersing should be taken into account[6–7]. A type of hollow glass beads (density 1.05 g/cm3and diameter 1–5 μm) were chosen as tracers for this test. For the seeding of the particles, the appropriate dosage of the particles should be seeded into the circulating flow. The number of the particles within an interrogation region should be sufficient to describe the flow state, and it should be noted if the number of the particles is too large, speckles may form,which may cover the tracers[7]. However, it is not easy to figure out such a right dosage quantitatively, as many factors, such as laser power, model structure, etc, may influence the quality of PIV images. In the tests, the dosage of the particles was determined in such a way: adding particles into water bit by bit, and series of real-time images of the seeding beads were monitored and displayed on the computer screen to seed the beads, from which the right dosage was decided when the desirable picture was obtained.

In the model, positional relation between the laser and the camera is shown in Fig. 5.

Fig. 5. Positional relations among the model, laser and camera

Due to the light refraction on the conical surfaces of valve seats, if the camera shot vertically into the model,images of the valve play could not be clearly photographed,and the gap between the disc and the seat shown in the recorded pictures would be more narrow than it actually was, which might cause some difficulties in computing velocities in the next step. To avoid these, after repeated trials, the method of off-axis measurement (see Fig. 6) was finally adopted. It has been proved that when the camera and the model’s symmetry axis were placed at an angle of about 20°–25°, the influence of the refraction caused by the conical structure would get minimized, which contributed to get better and clearer images of the valve play.

Fig. 6. Vertical shot and off-axis shot

In addition, the region where the flow field is investigated is the valve play, i.e., the annular gap between the conical surface of a valve disc and that of a seat. If the camera were laid horizontally, images of the valve play would be correspondingly angled towards the horizontal position, which might also cause inconvenience to marking the interrogation area. As shown in Fig. 7, the shooting angle of the camera in the vertical plane could be adjusted to make the images horizontal.

Fig. 7. Angle relationship between the camera and particle image

4 Processing and Analysis of Experimental Results

In order to explore the velocity distribution in the flow field for different valve angles (the valve angle refers to the angle between the axis and the generatrix of the valve disc)and for different valve lifts, experiments were conducted for 9 working conditions. For each valve seat and valve disc, tests were performed at lifts of 3 mm, 4 mm and 5 mm,respectively.

When the particle images were acquired, MicroVec2 software was used to process these raw picture data. Within the software, the speed of the particles was essentially computed on the basis of cross-correlation, and additionally techniques, such as iterative algorithm, erroneous vector correction, were employed to reduce the computing error.By using the software Tecplot, vector data obtained in the computations, including velocity, velocity distributions and vorticity distributions, could be plotted as shown in Fig. 8.Fig. 8 shows the flow field state of 45° valve at the valve lift of 4 mm. As shown in Fig. 8(a), the flow direction changes in the vicinity of position A, i.e., the base angle of the valve disc. Compared with Fig. 8(b) at the same position, it is found that the maximal velocity of the entire field takes place near the corner of the disc.

Fig. 8. Flow field state of 45° valve at lift of 4 mm

As illustrated in Fig. 9 in detail, when the water flows upwards and approaches the entry to the valve play where the cross-section of the flow channel is getting narrow gradually, some of the fluids will flow in the direction of the valve play, whereas others will be forced to move straight to the bottom of the valve disc. Because of this, a fluid pressure in the area beneath the bottom of the disc arises and increases gradually, and also the internal energy thereof accumulates, which forces these impeded fluids to rush into the valve play area. As can be seen in Fig. 9, from section B-B' to A-A', width of the flow channel continuously contracts and reaches the narrowest width near the section A-A'. According to the continuity principle,it can be concluded that the maximal speed occurs in the vicinity of section A-A'. The conclusion is applicable to other working conditions.

Fig. 9. Sketch of flow movement near the entry to valve play

In further, the rushing flow from under the disc bottom will be squeezed by that of the lower part in the passage and its flow direction will be turned along the valve play.As a result of fluid inertia, this turning of direction occurs mostly in the middle part of the entrance. As shown in Fig.9, on the upper wall of the channel, fluids in the proximity of position C are protected by both the bottom of disc and the exterior faster flow, less affected by the subsequent flow.Hence it can be deduced that substances therein exchange much more slowly with those of the outer liquid, which contributes to the forming of vortexes. So it can be speculated here that, for the practical operating condition of a pump, fluid therein is drilling slurry containing lots of solid granules, therefore eddies at position C will probably lead to the remain of those grains, which may become one of the factors inducing the valve failure.

Looking through the graphs of flow fields for different conditions (see Fig. 10–Fig. 17), obviously, the color of the middle part of the channel is brighter than that near the two walls. The brighter color indicates the higher mean flow velocities and more uniform flow vectors. However, in the vicinity of the upper and lower walls, the color is darker,which means the lower speed and slower change in flow directions. These results are analogous to those from previous CFD simulations[8]that the highest velocity exists at the midpoint of the cross-section and gradually decreases to both sides. Moreover, from velocity profiles, it can be concluded that the flow velocities near the entry to the valve play are remarkably higher than those near the exit. In practice, this is helpful that the sealing ring shall be fixed on the upside of the valve disc to reduce the erosion.

Fig. 10. Flow field state of 30° valve at lift of 3 mm

Fig. 11. Flow field state of 35° valve at lift of 3 mm

Fig. 12. Flow field state of 45° valve at lift of 3 mm

Fig. 13. Flow field state of 30° valve at lift of 4 mm

Fig. 14. Flow field state of 35° valve at lift of 4 mm

Fig. 15. Flow field state of 30° valve at lift of 5 mm

Fig. 16. Flow field state of 35° valve at lift of 5 mm

Fig. 17. Flow field state of 45° valve at lift of 5 mm

From vorticity profiles (see Fig. 18–Fig. 26), it can be inferred that the absolute values of vorticity in the adjacent areas of both walls of the channel are much higher than those in the middle. Vorticity values near the upper wall are positive while the lower are negative. This indicates that the flow direction changes a lot in the vicinity of the both walls of the passage, usually with an anticlockwise turning near the upper wall or with a clockwise turning near the lower wall. Considering the actual working conditions of a pump, as drilling slurry passes through the valve play, both of the turnings mentioned above will cause a phenomenon that the fluids with plenty of solid granules will strike the surfaces of the valve disc and seat.

Fig. 18. Vorticity profile of 30° valve at lift of 3 mm (s–1)

Fig. 19. Vorticity profile of 35° valve at lift of 3 mm (s–1)

Fig. 20. Vorticity profile of 45° valve at lift of 3 mm (s–1)

Fig. 21. Vorticity profile of 30° valve at lift of 4 mm (s–1)

Fig. 22. Vorticity profile of 35° valve at lift of 4 mm (s–1)

Fig. 24. Vorticity profile of 30° valve at lift of 5 mm (s–1)

Fig. 25. Vorticity profile of 35° valve at lift of 5 mm (s–1)

Fig. 26. Vorticity profile of 45° valve at lift of 5 mm (s–1)

After that, the granules slow down, and some of them congregate near the walls and the amount may increase with time. When the disc impacts the seat with speed of about 20 m/s[9], some of the massing granules may not be washed out of the play immediately. The remained solids might be compressed heavily in between the disc and the seat, damaging both of the surfaces.

As summarized in Table, the maximal velocities and the mean velocities of each flow field are obtained after a classification of the data. Based on Table, trend charts for both the maximal velocities and the mean velocities are respectively plotted.

Table. Max. velocities and mean velocities of each flow field

From Fig. 27(a), it can be seen that when the valve lift is constant, the maximal speed in the flow field shows a tendency of decreasing with the angle of the valve disc.This experimental result proves the CFD software computed result in Ref. [8] which indicates that, when the diameter of the disc bottom is constant (and so is the diameter of the valve hole), velocity in the flow field decreases with the increase in the valve angle. The scale of average velocities could be used to evaluate the rapidness of the flow. As shown in Fig. 28(a), when the valve angle increases, the mean speed also exhibits a downward trend.

Similarly, in Fig. 27(b), the plot of maximal speed exhibits a trend of reducing when the valve disc rises. It is interesting to note here that, as shown in this chart, the curve of the velocities is somewhat approximate to that of a negative exponential function, with a larger decreasing rate of velocity at shorter valve lifts but a smaller one when the lifts get higher. According to Fig. 28(b), the variation in the average velocity is still consistent with the general trend of decreasing, and the speed of the flow reduces when the valve lifts increase. Hence a conclusion could be drawn that,the flow velocity in the valve play is declining continuously while the valve disc is rising, and the rate of the decreasing is getting smaller gradually. Combined with the valve failure mechanisms, it can be inferred that when a working valve is about to open or to close, its lift is tiny and the flow velocity is extremely high. With such high-speed drilling slurry containing abrasive solid particles, serious erosion damage to valves, especially to sealing rings is ineluctable and the injuries may accelerate the failure of the valve.

Fig. 27. Variations of maximal velocity

Fig. 28. Variations of mean velocity

In order to get more precise data, an improved measurement, in which fluorescent particles were introduced, was carried out for the 30° valve at the lift of 4 mm. The fluorescent particles, of which the main component is Rhodamine B, are usually seeded in Micro-PIV experiments. Unlike those ordinary tracer particles, in the experimental tests with fluorescent particles,when these tracers are illuminated by laser, the fluorescence with a wavelength of around 590 nm can be excited. Meanwhile, when recording the particle images,the camera shoots into the flow with a bandpass color filter fixed in front of its lens, therefore visible light of other wavelengths could be effectively filtered out and its harmful influences on the images would get minimized.

Results of this measurement are presented in Fig. 29, in which A-A' denotes the section of the entrance to the valve play.

By analyzing these graphs, the following conclusions can be drawn.

(1) With the velocity distribution (see Fig. 29(b)), it is evident that when the water flows into the area of the channel(above Section A-A'), its velocity increases rapidly and approximately shows a parabolic layered distribution,which experimentally verifies the CFD result computed in Ref. [8];

Fig. 29. Results of measurement with fluorescent particles

(2) Fig. 29(d) presents a distribution of the velocity fluctuation. Obviously the values of velocity fluctuation in the channel area (above section A-A') are higher than those of the outer (below section A-A'), and generally the velocity fluctuation in the area near the wall is higher than that in the middle. Besides, there exists the highest velocity fluctuation in the proximity of position B, and that is the site where the actual valve failure initially appears.

Compared with experiments using hollow glass beads, it is found that results from this test are consistent with the previous conclusions.

5 Discussion of Measuring Accuracy

The accuracy of PIV may be affected by three factors:the precision of the PIV system, the following behavior of tracing particles, and the refraction of light, which are detailed as follows.

(1) Precision of the PIV system. The precision of the PIV system here mainly refers to the measuring accuracy of the computed results by processing particle images with the PIV software. With an experimental assessing approach presented in Ref. [10], the precision of the MicroVec2 software has been calibrated, and the average relative error is 1%.

(2) The following behavior of tracing particles. Results computed by the PIV system are actually the movement of those tracing particles, and that is why the following behavior must be considered when selecting the particles.In the work of LI, et al[11], an analysis of the following behavior was presented. And judging from their results, the particles used in this measurement, whose density is 1.05 g/cm3and diameter is 1–5 μm, have very good following feature, which is suitable for the test.

(3) Light refraction. During the collection of particle images, what the camera records were not the original images of the particles but the refracted ones penetrating the water and the acrylic glass, and that directly affects the measurement results. Since the refractive index of acrylic glass is about 1.49 and that of the water is about 1.33 (after the particles being seeded, the index will be a little higher),the angle of refraction from water to acrylic glass will be very small. However, the angle will be larger when light travels from acrylic glass to the air, and thus an error will be caused. As a solution of this, the external wall of the model is designed as square to minimize this negative influence, which has already been mentioned in Fig. 4.

6 Conclusions

(1) Near the base angle of the valve disc, flow direction changes distinctively, with a rapid velocity and serious erosion at this position. However，in the particular region which is marked as C in Fig. 9, an area of slower flow exists there, which may be easy to form eddies and cause the remaining of solid granules.

(2) The central region of the flow field has larger average flow velocities and more uniform flow vectors.Contrastively, in the vicinity of the both walls of the valve play, flow speeds thereof slow down and the direction changes greatly. In addition, the experimental test has proved that the velocities near the entrance to the play are significantly higher than that near the exit, thus sealing rings of the pump valve should be fixed on the upper part of the valve disc to reduce the erosion.

(3) The absolute values of the vorticity near the walls are much higher than those in the middle area, and the upper has positive values whilst the lower has negative values.Therefore, in practice, some of solid particles in drilling slurry may gather near both of the walls, and when the pump valve closes, they may be compressed, leading to the failure of the valve.

(4) A conclusion of previous CFD simulations (in Ref. [8]) has been verified that when the diameter of the disc bottom is constant (and so is the diameter of the valve hole), velocity in the flow field decreases with the increase in the valve angle. Besides, a group of plots have been presented in Fig. 27(b), from which it could be speculated that serious erosion would happen when the disc gets a smaller lift.

(5) With the results from the improved measurement with fluorescent particles, a parabolic layered distribution of velocities has been plotted, which experimentally verifies the previous CFD results (in Ref. [8]); it has been measured that values of velocity fluctuation in the valve play are higher than that out of it, and the highest fluctuation occurs on the site where the actual valve failure initially appears.

[1] JOFFE S H D, ALLEN C. The wear of pump valves in fine particle quartzite slurries[J]. Wear, 1990, 135(2): 279–291.

[2] YE Yongbiao, ZHU Yongyou, WANG Longlong. The failure analysis of the valve in reciprocating pump[J]. Water Conservancy& Electric Power Machinery, 2006, 28(1): 37–38. (in Chinese)

[3] WANG Zhonghui, LU Yongming, ZHANG Jinzhong. Fault tree analysis for failure of drilling pump valve[J]. Oil Field Equipment,1995, 24(1): 26–27. (in Chinese)

[4] XIA Yuanbai, LUO Hui. Testing method of the impacting force of drilling pump valve in the course of valve closing[J]. Oil Field Equipment, 2001, 30(6): 30–32. (in Chinese)

[5] ZHENG Jinyang, DONG Qiwu, SANG Zhifu. Process equipment design[M]. Beijing: Chemical Industry Press, 2001. (in Chinese)

[6] MEINHART C D, WERELEY S T, SANTIAGO J G. PIV measurements of a microchannel flow[J]. Experiments in Fluids,1999, 27: 414–419.

[7] WEN Jian, LI Yanzhong, WANG Simin, et al. Investigation of flow patterns in the entrance of heat exchanger by PIV technology[J].Journal of Harbin Institute of Technology, 2008, 40(1): 113–117. (in Chinese)

[8] YANG Guoan, HUANG Cong, YU Li. Compute and analysis of the flow field in the play of the drilling pump valve based on the simulation by FLUENT[J]. Oil Field Equipment, 2008, 38(3):41–44. (in Chinese)

[9] YANG Guoan, ZHANG Dong, HUANG Cong. The impact characteristic analysis of drilling pump valve[J]. Journal of Vibration and Shock, 2008, 27(12): 18–22. (in Chinese)

[10] DONG Mingzhe, WANG Yang, JIANG Ningtao. An experimental method for evaluating the PIV measurement error and its application in optimizing PIV experiment parameters[J]. Journal of Experiments in Fluid Mechanics, 2005, 19(2): 79–83. (in Chinese)

[11] LI Enbang, LI Zhiping, LI Chun, et al. Numerical analysis of following behaviors of particle tracers in turbulent[J]. Chinese Journal of Scientific Instrument, 2009, 30(2): 225–231. (in Chinese)