Professional Documents
Culture Documents
www.nafems.org
performed at HSL Buxton and was recorded on high speed video. Corresponding computational simulations have been performed using MSC.Dytran, to model the dynamic response of the plate during this rapid event. A comparison of the experimental work and the simulations is made.
Experimental Setup
The gun was manufactured from three sections of stainless steel tube having a theoretical burst pressure in excess of 300 bar. The driver section, capable of being pressurised up to 16 barg using the air supply installed, is visible to the left in Figure 1. Water was retained in the centre 1 m long portion of the gun with the barrel being the remaining 1 m length of stainless steel pipe. The water column was supported at its upstream (driver) end via a double bursting disc assembly, and at its downstream end by a thin sheet of plastic-coated paper. Bursting discs are thin metallic discs, designed to rupture when the difference in pressure across the two faces of the disc exceeds a specified level. To trigger the event, a valve was opened to dump the compressed air in the void between the two bursting discs.
Page 14
January 2005
www.nafems.org
At this instant: The first bursting disc, between the driver section and void ruptures, because the differential pressure now exceeds its burst pressure. The second disc, between the void and water, also ruptures, exposing the water column to the full force of the compressed air. The paper disc at the far end of the water column rips open The 1 m column of water starts to accelerate down the barrel on the right, driven by the expanding column of compressed air from the driver section. The target was placed immediately beyond the gun barrel. The targets were steel plates, 300 mm in diameter, and of nominal thicknesses of 1, 2 and 3 mm. Two tests were performed with 1 mm target plates, the second test at a higher pressure. The arrangement of the target structure is shown in Figure 2. Holding the target plate in place around its circumference were eight M20 steel bolts, themselves mounted into a rigid steel backplate. Each bolt held the plate sandwiched between a pair of 36 mm diameter (and 2 mm thick) washers, and a pair of nuts. These bolts had their centres 27 mm from the plate rim, and held the plate a distance of 142 mm away from the backing plate. High speed video records were made of the four tests. With those for the thinnest plates it was possible to use these records to measure deformations during the early stages of the test. For the thicker plates, and during the latter stages of tests with thin plates, much of the image was obscured by water making measurements impossible.
January 2005
Page 15
www.nafems.org
the thicker plates. Firstly, deflections for the 2 mm and 3 mm plates were too small to see; and secondly water sprayed directly at the camera obscuring the view. Despite these difficulties a detailed record of deflection with time could be extracted for the thin plates, since the dishing of the plate acted to reflect water back towards the gas gun, delaying the moment at which the target could no longer be seen. TEST 4: 0.86mm Plate at 16.3 bar Here we will concentrate on the most extreme test, test 4. This test was performed using a thin plate (0.86mm) at the highest pressure possible (16.3 bar) although the gas gun itself is rated at much higher pressure than this, the limiting factor was the nylon tubing used to connect the gas gun to the air supply. For this test, the force of the Figure 4 High-Speed video still of the Test 4 water caused s i g n i f i c a n t deformation of the bolts supporting the plate (see Figure 4), and the final plastic deformation of the plate centre was measured to be 37mm. Indeed, this test was clearly on the limit of the strength of the target structure, since a tear was observed in the vicinity of one of the washer mounting points. The corresponding computer predicted shape of the plate can be seen in Figure 5, which compares very well with the actual shape observed. However, the computed permanent deformation of the plate centre was notably lower, at 27 mm. It is likely that this was caused by considerable deflection of the securing bolts, allowing greater freedom of movement than in the computational model. The peak plastic strain observed during this high pressure test was 5.75x10-2 (Figure 6) and as could be inferred from the earlier results, the peak strains and stresses occur both in the plate centre, and on lines radially out towards the mounting washers. Figure 7 presents a comparison of the deformation, with time, from the computational and experimental results for Test 4. The solid blue line shows the result from observing the high speed video noting that a 3 mm error is likely
The Fluid Region An Eulerian finite volume mesh was a d d e d , overlapping the regions of the gas gun and the target plate, in order to model the fluid. This mesh comprised 10912 hexahedral cells. As with the inside of the gas gun, the target plate acted as a barrier to the fluid flow, reflecting the water away from it. This fluid-structure interaction behaviour was modelled by using a General Coupling algorithm, which computes the fluid finite volumes and structural finite elements simultaneously. The outer faces of this region of the mesh were set to outlets, through which the reflected water passed out of the solution domain. Rather than start the computations at the instant of firing, the simulations here commenced at the point when the leading edge of the liquid slug was about to be ejected from the end of the gun barrel. Hence, the simulations began when the water would have travelled a distance of 1 m, and this would have taken approximately 45 ms from the triggering of the event. Simple spreadsheet analysis of the force and acceleration from an expanding gas volume produces an exit velocity of 52.1m.s-1 from a firing at 16.3barg.
Results
Tests Performed Four tests were performed with the results recorded by high-speed video. The key parameters for these tests are shown below in Table 1.
Table 1 Test Parameters Test Target plate thickness (mm) 0.86 2.05 3.00 0.86 Driving pressure (bar gauge) 9.7 10.0 ~ 9.0 16.3 Peak elastic deformation (mm) 39 45 Permanent plastic deformation (mm) 27 2 1 37
1 2 3 4
For all the tests a graduated paper scale was placed behind the plate to allow measurement of elastic response with time. This presented several problems, however, with Page 16
January 2005
www.nafems.org
In light of these comments, postulating that the simple elastic modelling of the bolts may have affected the results, a further run was performed, with a defined yield stress for the bolt material. Each M20 bolt was assumed to have a yield stress of 6.4x108Pa (the minimum for grade 8.8 bolts). As before, each bolt was modelled as a single beam element, but it was recognised that the use of more elements could have captured the deformation in more detail. The difference in the results was quite striking. Figure 8 reproduces Figure 7, and includes an additional line (in purple) Figure 5 MSC.Dytran-computed shape of Figure 6 Final Plastic strains on the 0.89 mm showing the effects of including more realistic bolt 0.86 mm plate used in Test 4 plate from 16.3 bar firing (Test 4) characteristics on the displacement of the plate. The predicted peak displacement is now 43.5mm, from observing the paper scale. An initial rapid compared with the experimental peak displacement of deformation occurs over the first 4 ms, and then the plate 45mm (an error of 3%). The computed ultimate centre remains steady about this level until a second event deformation at plate centre is 35.4mm, the experimentally occurs, producing the peak displacement at about 13ms, measured value was 37mm (an error of 4%). from which the plate starts to recover elastically. The plate continued to oscillate for a while (but water spray prevents quantification of this), before eventually exhibiting the plastic deformation shown in the dotted cyan line
Figure 8 The effect of allowing elastic-plastic behaviour of the bolts in the computer model.
Summary
Figure 7 A comparison of the dynamic response from computer and experiment (Test 4)
The corresponding MSC.Dytran simulation is shown in the yellow line. Although the shape of this curve is very similar: an initial rapid deformation, with a secondary event occurring later; the peak of the displacement was about 5 mm less than experiment. It was unclear exactly why this discrepancy occurred, however a likely source of error was the way in which the supporting bolts were modelled. Subsequent Modelling of Elastic-Plastic Bolts Initially, it had been assumed for all the modelling work that the bolts were rigid, and hence the plate was fixed in space at its mounting washers. When it became apparent that the bolts did bend during the test, they were included in the model as simple beam elements, with purely elastic material properties. Scrutiny of the high speed video, however, shows that there was substantial deformation of these bolts. During the final test the bolts had been bent plastically, making it very difficult to remove the plate afterwards.
This work has produced both computational and experimental results studying the phenomena of condensation-induced waterhammer, as can occur in steam systems. This work clearly illustrates the potential damage that could occur within a steam system should a slug of liquid water be allowed to accumulate in the main. The results from the computational and experimental programmes agree very well. For the most extreme test, the modelled centreline displacement agreed within 4% of the experimentally obtained value. The computed shape of the deformed structure also agreed very well with experiment. A final test case showed that altering the material model for the supporting bolts to include an element of plastic response made a significant improvement to the results for the most severe test. As a consequence of this finding it is expected that the correlation between experiment and prediction can be improved by more accurate modelling of the bolt material characteristics and refining the bolt mesh.
Contact
Dr Aubrey Thyer, HSL, aubrey.thyer@hsl.gov.uk Page 17
January 2005