A new sliding joint to accommodate recoil of a free-piston-driven expansion tube facility

Abstract

This paper describes a new device to decouple free-piston driver recoil and its associated mechanical vibration from the acceleration tube and test section of The University of Queensland’s X3 expansion tube. A sliding joint is introduced to the acceleration tube which axially decouples the facility at this station. When the facility is fired, the upstream section of the facility, which includes the free-piston driver, can recoil upstream freely. The downstream acceleration tube remains stationary. This arrangement provides two important benefits. Firstly, it eliminates nozzle movement relative to the test section before and during the experiment. This has benefits in terms of experimental setup and alignment. Secondly, it prevents transmission of mechanical disturbances from the free-piston driver to the acceleration tube, thereby eliminating mechanically-induced transducer noise in the sensitive pressure transducers installed in this low-pressure tube. This paper details the new design, and presents experimental confirmation of its performance.

1 Introduction

The free-piston driver [1] offers a safe and efficient means by which to compressively heat a light driver gas, initially at room temperature and at low pressure, to thousands of Kelvin and tens of megapascals. This unique capability enables it to drive extremely strong shock waves for impulse facilities, and The University of Queensland (UQ) uses this device to drive its T4, X2, and X3 impulse facilities.

Fundamental to its operation is the passage of the piston down the compression tube. Initially it is accelerated from behind by compressed air, typically reaching maximum speeds of between 100 and 250 m/s. For most of the stroke, the pressure in front of the piston is relatively low, until finally the piston nears the end of the tube. At this point, the compression ratio of the driver gas, typically comprised of helium or a mixture of helium and argon, rapidly rises. Eventually the steel diaphragm ruptures, and shock tube flow is initiated.

This paper considers UQ’s X3 facility (Fig. 1), which is the world’s largest free-piston-driven expansion tube. Central to its high performance is its 15 m long, 500 mm diameter compression tube. Considering one of its higher performance driver operating conditions, the reservoir pressure behind its 200 kg piston is initially 9 MPa, which corresponds to an initial axial pressure force of 1.8 MN. When the piston is released, it accelerates downstream. Simultaneously, the axial force which was restraining the piston is removed, leaving a net equal and opposite force acting on the upstream end of the facility. This force reduces as the reservoir gas expands behind the accelerating piston. For the same condition, towards the end of its stroke, the diaphragm ruptures at a pressure of approximately 35 MPa. Immediately before this rupture event, the driver gas applies a maximum force of 6.9 MN to both the piston and the downstream end of the driver tube.

Fig. 1

X3 facility schematic showing location of the new recoil sliding joint. Manifold and supporting framework not shown. Detail views of new sliding joint and downstream stopper are provided. Drawing is to scale

The simple analysis above demonstrates that the axial loads which arise in a free-piston driver are considerable, and for this reason, the majority of such devices are supported on freely moving rollers. Allowing the facility to freely recoil avoids the problem of designing supporting hardware to restrain the device under loading; however, it is usually desirable that the test section itself, where the experiment takes place and measurements are made, is fixed relative to the laboratory frame of reference. Free-piston-driven facilities at UQ currently comprise a series of rigidly connected tubes from the driver through to the tube exit; during operation, the recoil has been accommodated by a sliding seal between the tube exit and the fixed test section.

The high axial loads also induce significant stresses in the tube itself. Both the initial recoil, and the peak load at diaphragm rupture, set in place a complex system of stress waves which traverse the length of the facility at the speed of sound of the wall material. In this case, where the material is steel, the sound speed is approximately 5 km/s, which is comparable to the shock tube flow speed for many operating conditions. It was previously shown in [2] that these stress waves induce strong vibrations in the sensitive acceleration tube wall static pressure transducers, which can lead to erroneous static pressure measurements. This is highly problematic, since these measurements are vital to establishing primary shock speed and test flow static pressure—both essential to test flow reconstruction—and [2] addressed this issue by introducing a compliant joint with a sliding seal at the shock tube-to-acceleration tube joint. This managed to significantly reduce transmitted stress waves, and therefore transducer noise, in both the X2 and X3 facilities; however, geometry constraints meant that the X3 solution required compressive preloading of the joint prior to operation (to compress o-ring seals), which made its operation more difficult to optimise in practise, as noted in [2].

Another feature of the above arrangement is that the nozzle exit plane recoils relative to the test section during the experiment. This has two unfavourable consequences: firstly, it introduces the potential for mechanical vibration of the test section as the nozzle slides upstream against o-ring seals; secondly, it means that the position of the nozzle exit plane changes during the experiment. These two aspects of the sliding nozzle arrangement complicate model positioning, and make it poorly suited for alignment of optics and other aspects of experimental setup.

In order to more effectively address the stress wave disturbance issue, and to affix the position of the nozzle exit plane relative to the test section, a new sliding joint has been developed to accommodate facility recoil during operation, and this device forms the subject of the present paper.

2 The new sliding joint

The new recoil sliding joint is shown in Fig. 2, and is located immediately downstream of the tertiary diaphragm station (see Fig. 1). The joint comprises two primary fittings which screw onto the existing upstream and downstream acceleration tube sections, and replaces an existing rigid tube joint. The upstream fitting slides inside the downstream fitting, and is guided by two Nylatron alignment bands. The sliding seal is achieved with two o-rings on the upstream fitting, located between the alignment bands. The other two o-rings seal the joint fittings to each adjacent section of tube.

Fig. 2

X3 recoil sliding joint; a photograph of installed hardware; b cutaway schematic of sliding joint in closed position; c cutaway schematic of sliding joint in maximum recoil position. Schematics are to scale; the full drawing set, including 2D drawings and 3D CAD models, can be accessed at [3]

The two fittings are axially joined with eight lengths of steel shafting, which screw directly into the upstream fitting, and slide through linear bearings in a bearing holder attached to the downstream fitting. A custom designed compression spring is located over each length of shafting. Each spring is initially pre-loaded to keep the sliding joint closed. The 400 mm long springs nominally have a spring rate of 6.1 N/mm, a maximum working compression of 160 mm, and a solid length of 202 mm. The springs are initially compressed by 50 mm using M30 nuts and washers, producing a total joint preload of $P_{\mathrm{preload}}=8\times 50\times 6.1=2{,}440~\hbox {N}=2.4~\hbox {kN}$.

When the facility is fired, the upstream tube section recoils, and the sliding joint opens up, as shown in Fig. 2c. For X3 operating with its 200 kg piston, this recoil displacement is approximately 60 mm. When the recoil joint is used, the peak load across this joint depends on the size of the enforced displacement (60 mm), the initial preload displacement (50 mm), and the stiffness of the springs. At maximum recoil, the total spring displacement is $50+60=110$ mm, causing the spring load to increase to $P_{\mathrm{recoil}}=8\times 110\times 6.1=5{,}368~\hbox {N}=5.4~\hbox {kN}$.

In the original configuration, where the tube is rigidly connected at this station, the entire facility moves as one, and the rigid joint must accelerate the downstream mass to match the upstream recoil displacement. For this case, the peak load across this joint, $P_{\mathrm{rigid}}$, can be estimated as follows:

$$\begin{aligned} P_{\mathrm{rigid}}= & {} p_{\mathrm{res}}\cdot \dfrac{\pi D_p^2}{4}\cdot \dfrac{m_{\mathrm{acc}}}{m_{\mathrm{tot}}} \nonumber \\= & {} 14\times 10^6\times \dfrac{ \pi \times 0.5^2}{4}\times \dfrac{4{,}000}{50{,}000} \nonumber \\= & {} 219{,}911~\hbox {N}=220~\hbox {kN}, \end{aligned}$$

(1)

where $m_{\mathrm{tot}}$ is the (approximate) total recoiling facility mass, $m_{\mathrm{acc}}$ is the (approximate) proportion of this mass downstream of the recoil joint (comprising acceleration tube and nozzle), $p_{\mathrm{res}}$ is the maximum rated pressure for X3’s reservoir, and $D_{\mathrm{p}}$ is the diameter of the piston, which is equal to the end wall diameter upstream of the piston. This estimate is conservatively high, since it neglects the retarding effect of friction through the system of rollers supporting the facility, and assumes the assembly accelerates as a perfectly rigid body.

$P_\mathrm{{recoil}}$ is 2.5 % of $P_{\mathrm{rigid}}$, therefore axial loading is reduced to negligible levels, and it is expected that this new design should prevent transmission of stress wave disturbances associated with the initial facility recoil. At least as importantly, when the piston completes its stroke and the diaphragm ruptures, and axial loading is a maximum, the joint is open and is axially decoupled. The primary mechanism for transmission of stress waves is thus eliminated, and significant stress wave disturbances can no longer be transmitted across the joint.

3 Joint lockout feature

Per Fig. 2c, the maximum nominal recoil displacement is 100 mm. Facility recoil is a result of conservation of momentum, and increases with piston mass. Noting that the recoil is 60 mm for X3’s heaviest piston (200 kg), X3’s other piston is 102 kg [4] and will therefore have a lower associated recoil.

Considering the recoil joint in the closed position (Fig. 2b), the steel shafting flares from $\varnothing 25$ mm to $\varnothing 30$ mm at 120 mm downstream of the custom brass washers, which are 20 mm thick and have $\varnothing 25.1$ mm inner diameter. In the event that the facility recoil exceeds 120 mm, the shafting will wedge first into the brass washers, and then into the linear bearings. The recoil sliding joint will lock in this position, and any addition load will be transferred through the steel shafting into the bearing block attached to the downstream sliding joint fitting. Conservatively assuming a yield stress of 250 MPa for the steel shafting, the maximum load that can be transferred is as follows:

$$\begin{aligned} P_{\mathrm{shaft}}= & {} \dfrac{\pi \times 0.025^2}{4}\times 8\times 250\times 10^6 \nonumber \\= & {} 981{,}748~\hbox {N}=982~\hbox {kN}. \end{aligned}$$

(2)

$P_{\mathrm{shaft}}$ is over $4{\times }$ $P_{\mathrm{rigid}}$, therefore the locked sliding joint would be able to recoil the downstream acceleration tube and nozzle without catastrophic failure; the brass washers, and potentially the linear bearings, might be damaged and require replacement, which would be a trivial repair if such an unexpected event were to occur. With the recoil joint locked, the nozzle would retract from the test section in accordance with its original design.

4 Vacuum force across nozzle sliding seal

Figure 3 shows the loading which arises when a diverging nozzle with sliding seal is used. Since no axial force can be reacted between the nozzle and test section (ignoring friction across the nozzle sliding seal), an internal tube axial load, $P_v$, develops to prevent the nozzle being ‘sucked’ into the test section:

$$\begin{aligned} P_v= & {} p_{\mathrm{atm}}\cdot \dfrac{\pi (D_n^2-D_d^2)}{4} \nonumber \\= & {} 1.01\times 10^5\times \dfrac{\pi (0.44^2-0.27^2)}{4} \nonumber \\= & {} 9575~\hbox {N}=9.6~\hbox {kN}, \end{aligned}$$

(3)

where $D_n$ is the outer diameter of X3’s current Mach 10 nozzle, $D_d$ is the outer diameter of the acceleration tube, $1.01\times 10^5$ Pa is a standard atmosphere at sea level, and the initial pressure in the test section is assumed to be 0 Pa. $P_v=9.6$ kN is considerably larger than the initial recoil sliding joint preload, $P_{\mathrm{preload}}=2.4$ kN, therefore the joint will tend to open up when the facility is under vacuum.

Fig. 3

Internal axial force arising from nozzle sliding seal

Spring preloading can be increased to resist this initial vacuum force by two ways: much stiffer springs can be used, however this will result in a much larger transmitted force across the sliding joint during the facility recoil, thereby increasing undesirable mechanical disturbances to the acceleration tube. Furthermore, springs which are sufficiently stiff will have insufficient travel to accommodate the recoil. Alternatively, the existing springs can be installed with much greater initial compression, however this will then use up most or all of their available travel. It is difficult to balance spring pre-load with spring flexibility and travel, and this problem was instead therefore addressed by reacting the vacuum force at the next downstream tube joint.

Referring to Fig. 1, a mechanical stopper was installed at the next downstream tube joint, and prevents the nozzle being sucked into the test section. A photograph of the stopper is shown in Fig. 5. Prior to the shot, if the recoil joint remains shut under spring preload, it can be assumed that most of $P_v$ is reacted by the mechanical stopper. When the facility is fired, the upstream section recoils, and the recoil joint opens up. As shown previously, for a 60 mm recoil displacement with 50 mm spring pre-load, $P_{\mathrm{recoil}}=5.4$ kN, which is approximately 56 % of $P_v$. Therefore, the downstream acceleration tube and nozzle assembly should remain stationary relative to the test section.

It is noted that the same function of the stopper could be achieved by installing a shoulder on the nozzle, which would press against the test section front plate and resist the vacuum force, and this approach may be adopted in the design of future nozzles.

5 Experimental results

The new sliding joint was commissioned using the Mach 10 scramjet flow condition detailed in Table 1, and first described in [5], and its performance was established across eight shots of the facility (shots x3s451 to x3s458). Figure 4 shows still images extracted from high-speed colour video of the joint during the first experiment (x3s451), which was recorded at 125 frames per second. In these images, $t=0$ ms refers to the image just prior to initiation of joint movement. The joint operated as expected; the upstream fitting moved freely with the recoiling facility; the downstream fitting and attached acceleration tube remained static.

Table 1 Initial configuration of X3 for Mach 10 flow condition [5]

Fig. 4

Operation of recoil sliding joint for shot x3s451. Motion was recorded at 125 frames per second with a Photron MX100 high-speed colour camera. The plot shows recoil displacement measured from the images, with an associated uncertainty of ${\pm }1$ mm. Points on the plot represent the recoil displacement at the times shown in the images. The associated high-speed video, at $0.2{\times }$ playback speed, is provided in Online Resource 2

For these experiments, the sliding joint was operated with an initial opening of 12 mm (see top of Fig. 4), in order to account for any ‘over-shoot’ from the downstream fitting. Referring to the plot in Fig. 4, there was indeed a small amount (6 mm) of overshoot, and therefore this arrangement will be retained. This gap size is established by adjusting the position of the stopper relative to the adjacent tube joint prior to pumping the facility down to vacuum. Once the facility is under vacuum, the tube is sucked downstream until the tube joint contacts the stopper.

5.1 High-speed imaging

Online Resource 1 shows actual ($1{\times }$) speed video and audio of the sliding joint in operation, filmed using a Canon EOS Kiss X5 digital SLR at 60 frames per second (fps), for shot x3s451. Online Resource 2 shows the same experiment filmed using a Photron MX100 at 125 fps, but played back at 25 fps ($0.2{\times }$ speed). Reviewing both videos, a small amount of vibration is apparent in the downstream fitting attached to the acceleration tube. The source and timing of this vibration was further examined in subsequent experiments.

Referring to Fig. 5, 1000 fps high-speed video was recorded of the downstream stopper using the Photron camera, for experiment x3s455. The video is provided in Online Resource 3. The video begins approximately 100 ms before the piston is fired. It can be seen that there is first a very slight upstream movement of the acceleration tube (approximately $0.1{-}0.2$ mm), soon followed by a more pronounced deflection of approximately 1 mm downstream (in the direction of the test section).

Fig. 5

High-speed imaging of recoil sliding joint stopper. See Online Resource 3 for a video of the stopper during experiment x3s455

The initial upstream deflection is attributed to the rapid increase in spring force developed as the sliding joint opens during recoil, and is a consequence of the upstream tube being mechanically unrestrained. The magnitude of this deflection, and the associated tube inertial loading, is considered to be negligible, which is confirmed in the next sub-section.

The 1 mm downstream deflection is more significant, and its timing with relation to the useful experimental test time therefore needed to be established. Figure 6 shows high-speed imaging of the nozzle and test section performed during shot x3s456, at a frame rate of 2000 fps. The red dashed line has been added to the images, is aligned with the initial position of the test section flange, and is fixed relative to the laboratory. X3’s Mach 10 nozzle is filament-wound from fibreglass and as a consequence is opaque. For the purposes of this analysis, flow luminosity is sufficient to establish the timing of test section flow processes in relation to the tube motion currently being examined.

Fig. 6

Images 1–7 High-speed imaging of X3 nozzle and test section for shot x3s456. Motion was recorded at 2000 fps with a Photron MX100 high-speed colour camera. The associated high-speed video, at $0.025{\times }$ playback speed, is provided in Online Resource 4. Image 8 2000 fps High-speed imaging of test section flange detail, for shot x3s458. The associated high-speed video, at $0.05{\times }$ playback speed, is provided in Online Resource 5

Time is referenced in Fig. 6 to the first camera frame where nozzle glow is observed. The initial nozzle glow ($t=0$ ms) is due to light emitted from the hot shock-processed accelerator gas, approaching from the upstream direction. The nozzle glow intensifies when the shock-processed accelerator gas passes through the nozzle ($t=0.5$ ms), whereby light is transmitted directly from the hot gas. The test and driver gases, which arrive in the test section approximately 0.5 ms after the initial arrival of the shock-processed accelerator gas, are not hot enough to emit visible light. However, the nozzle is observed to continue glowing for tens of milliseconds. This is due to light emitted from the downstream test section, from hot shock layers produced by test and driver gas flow over the model; for these experiments, a 2-D scramjet was located at the exit of the nozzle.

For this flow condition, which has a test time of order 1 ms [5], the useful test time has ended a couple of milliseconds after the nozzle glow is first observed. Referring to Fig. 6, there is a downstream movement of the test section, which begins at $t=13.5$ ms, but this occurs well after the useful test time has ended. This motion is primarily due to the primary shock wave impacting the downstream end of the test section, therefore imparting a large axial force to the structure. The result is a vibration characterised by an initial 2 mm downstream deflection. This deflection causes a force to be applied to the nozzle through the sliding seal, and is the source of the 1 mm acceleration tube deflection observed at the stopper in Online Resource 3.

Image 8 from Fig. 6 shows a detail view of the test section front flanges at the time of maximum test section displacement (2 mm). This image is from a video recorded during experiment x3s458, which is provided in Online Resource 5. This detail view high-speed video confirms the above observations; before and during the test time, there is a minor vibration caused by an initial downstream deflection, which is associated with the sliding joint spring deflection during recoil; after the test time, there is a larger vibration associated with shock impact into the downstream end of the test section.

It is finally noted that $1{\times }$ and $0.125{\times }$ playback speed video and audio of the nozzle and test section during a test are provided in Online Resources 6 and 7. These are included in order to provide an overview of these phenomena on real time scales.

5.2 Acceleration tube static pressure traces

Figure 7 shows three sets of tube wall static pressure traces from X3’s acceleration tube. Figure 7a shows results for shot x3s400. These results were originally presented in [2, Fig. 10a] and are for a similar free-piston driver condition, but a completely rigid acceleration tube. Shock arrival, characterised by the sharp rise in static pressure, is preceded in each case by irregular and large amplitude disturbances. These are attributed to mechanical disturbances in the tube, originating from free-piston driver operation, and are seen to significantly degrade the static pressure measurements.

Fig. 7

Acceleration tube wall static pressure traces for a Mach 10 flow condition in X3, a with rigid tube connection; b after installation of the diaphragm holder and buffer arrangement described in [2]; c with the new recoil sliding joint reported in this paper. Transducers are all located downstream from the tertiary diaphragm (TD) plane, by an amount $\hbox {TD}+x$ as indicated in the plots

Figure 7b shows results for shot x3s401, which were also originally presented in [2, Fig. 10b]. For x3s401, a compliant diaphragm holder/buffer mechanism was installed at the tertiary diaphragm station of X3 (Fig. 8), which was able to compressively deflect by several millimetres during facility operation. Observing Fig. 7b, it can be seen that large amplitude noise has been considerably reduced at transducer locations AT4 to AT8. However, upstream transducers AT2 and AT3 (blue and red curves) are observed to have significantly increased high-frequency noise.

Fig. 8

X3 diaphragm buffer/holder assembly; the tube connection hardware is not shown. Originally presented in [2, Fig. 8]

Reference [2] attributed the high-frequency noise in Fig. 7b to the transient response of the diaphragm holder within the holder/buffer assembly, which could potentially ‘rattle’ around in the joint after undergoing impact from the primary shock wave, as well as disturbances from subsequent diaphragm rupture and tube mechanical vibration. While the actual cause of these vibrations was not confirmed, the results in Fig. 7b were considered a significant improvement compared to the original configuration, especially since the downstream transducers are far more useful for actual test flow characterisation.

As noted in [2], a limitation of the device detailed in Fig. 8 was that the joint had to be compressively pre-loaded to ensure pre-compression of o-rings. Subject to this requirement, the joint then had to be assembled as ‘loosely’ as possible, to minimise compression of the rubber bumper. Furthermore, referring to Fig. 8, there was the potential for transient venting past o-rings ‘1’ and ‘2’ depending on the transient response of the diaphragm holder inside the joint assembly. It was stated in [2], therefore, that a future upgrade would incorporate a design where pre-compression of the joint was not required.

The recoil sliding joint presented in this paper fully realises the concept proposed in [2]. Figure 7c shows acceleration tube static pressure traces for shot x3s456, which are typical of the eight shots performed with the new sliding joint. Compared to Fig. 7a and b, a transducer was installed at location ‘AT1’, which is just upstream of the sliding joint location. This transducer was not originally used for the [2] study, since pressure traces close to rupturing diaphragms typically yield noisy pressure data of little value to test flow characterisation. However, for these experiments, the AT1 trace was included to enable examination of noise upstream of the joint. It is also noted that Fig. 7c lacks an AT2 trace; this is because the sliding joint replaces the tube joint where this transducer was previously located.

Figure 7c has two important features:

• 1. Mechanical noise is still evident in AT1, which is upstream of the sliding joint, and this noise precedes the arrival of the shock wave itself.

• 2. Mechanical noise has been entirely eliminated from AT3 to AT8, and traces are also improved in relation to Fig. 7b. Furthermore, the high-frequency noise at AT3 has now been eliminated.

Figure 7 shows that the new recoil sliding joint effectively decouples the acceleration tube from upstream mechanical disturbances. Unlike the Fig. 8 design from [2], the sliding joint requires no special configuration or adjustment each shot, and does not introduce any potential leak paths during operation. Furthermore, it achieves improved tube wall static pressure measurements.

6 Operational considerations

There is a moderate amount of static friction associated with the sliding seal (a few thousand Newtons of force). Under vacuum there is ample force to overcome this friction, and the joint opens until the downstream tube joint contacts the stopper. Once the experiment is complete, and the facility has been vented to atmospheric pressure, the springs are relied upon to close the joint. Establishing an appropriate spring tension was an iterative exercise; the facility is pumped down to vacuum so that the joint opens up, and it is then vented. If the joint does not close under spring force, the springs are tightened until it closes. This cycle is repeated until reliable closing after venting is achieved.

The thin inner sleeve (Item 6 in Fig. 2) ensures that only a small gap is introduced between tube sections during recoil. The sleeve is not sealed, and gas can therefore vent to the main o-rings (Item 4), however the gaps are small and this venting is expected to have negligible effect on critical flow processes. The sleeve is thin in order to reduce the radial step size, and has therefore been made replaceable.

The stopper needs to be opened during facility turnaround to allow movement of the tubes. Interlocks have been installed to prevent operation of carriage tracking mechanisms when the stopper is closed, and to prevent facility pump-down when the stopper is open.

It is finally noted that the recoil sliding joint introduces several pinch hazards (closing and opening gaps at the actual joint and the stopper, as well as the springs). Images in this paper show these gaps exposed, and they were left uncovered for these initial commissioning experiments. However, perspex covers have since been installed for both the sliding joint and stopper. The springs have been left exposed; standard operating procedures, signage, and laboratory access restrictions address this hazard.

7 Conclusions

A new sliding joint has been developed for UQ’s large X3 expansion tube facility, which axially decouples the facility at its acceleration tube. The operating principles of the new device have been detailed in this paper, and its performance has been experimentally verified. It has been shown that the new joint achieves two important functions: firstly, it eliminates nozzle movement relative to the test section before and during the experiment, which has benefits in terms of experimental setup and alignment. Secondly, it reduces, to negligible levels, the transmission of mechanical disturbances from the free-piston driver to the acceleration tube. This eliminates mechanically-induced transducer noise in the low-pressure acceleration tube transducers. The device can be retro-fitted to an existing facility, and once initial commissioning has been accomplished, the system is essentially ‘set-and-forget’. Given its clear performance advantages, the new device presents a practical upgrade option to enhance both existing and future free-piston-driven expansion tubes.