POAC 09 Luleå, Sweden Proceedings of the 20th International Conference on Port and Ocean Engineering under Arctic Conditions June 9-12, 2009 Luleå, Sweden POAC09-78 MODEL TESTING OF RIDGE KEEL LOADS ON STRUCTURES PART I: TEST SET UP AND MAIN RESULTS Nicolas Serré 1, Pavel Liferov 2, Karl-Ulrich Evers 3 1 Norwegian University of Technology, Trondheim, NORWAY 2 StatoilHydro, Trondheim, NORWAY 3 Hamburgische Schiffbau Versuchsanstalt HSVA, Hamburg, GERMANY ABSTRACT Experiments to investigate the loads from ridge keels on structures have been performed at Hamburgische Schiffbau-Versuchsanstalt (HSVA) in Germany in scale 1:20. Four ice ridges were built and tests were performed with ice temperature and interaction speed as variables. The present paper (Part I) reports the main interaction tests while description of ridges, testing of their mechanical properties and of the freeze bonds are given in Parts II, III and IV respectively. The structures included two 0.7 m by 0.7 m underwater cubes and one 62 -sloped cone with a water line diameter of 0.6 m. The results indicate that colder ridges exerted higher loads on the structures. However no definitive conclusions could be drawn regarding the temperature or speed effect on the effective pressure. The horizontal loads on the subsea structures reached a steady state of progressive local failure and correlated to the keel profiles. The friction loads could cause up to 25% of the vertical loads on the cubes (vertical-sided). On contrary to typical level ice action on structures, the keel action did not follow any particular frequency. The level of the keel action was dependent on the failure mode. A global failure was observed only once, for the cone interaction. The importance of the failure mode as a load release mechanism was demonstrated from videos and time series analysis. INTRODUCTION Gudmestad and Liferov (2007) discussed the use of sub-sea technology in the ice-covered waters. The design to resist ice actions requires knowledge about the loads that ice ridge keels can exert on bottom-based structures. It is often assumed that the limiting load from ridge keels is associated with either local or global failure. Liferov and Høyland (2004) reported medium scale keel interactions with a solid nearly-vertical object. They observed that the keel failure was progressive, occurring while the ice ridge moved forward. Analytical and numerical models exist, or can be developed, to calculate keel loads on structures. Ice model basin testing is believed to be the first mean to provide input for validation of these models. For a number of reasons, physico-mechanical properties of the model keels are rarely well documented. This negatively affects the confidence in the obtained load values. Therefore, the following objectives were targeted during the present research test campaign at HSVA: Investigate and document ridge behaviour and keel failure mechanisms during the interaction with structures (this paper).
Measure physical properties of the ridges with particular focus on thermo-dynamics (Part II (Repetto Llamazares et al. 2009a)). Measure mechanical properties of keels by different supplementary methods (Part III (Serré et al. 2009)). Study the freeze bond strength of model ice (Part IV (Repetto Llamazares et al. 2009b)). Analysis of test results, including derivation of rubble properties and post-modelling of the interactions will be reported in future publications. TEST SET UP Structures The HSVA ice tank is 78 m long, 10 m wide and 2.5 m deep. A 5 m deep section of 10 m by 12 m is located at the end of the tank. For the purpose of modelling shallow water conditions bottom elements (false bottom) were inserted into the tank to reduce the water depth. The structures were fixed on the false bottom which was displaced along the tank by a motor driven carriage able to provide a maximum towing load of 50 kn. The dimensions of the structures, the level ice thickness, the drift speeds and the keel depths were scaled with a Froude scaling of 1:20. A discussion on the scaling procedure for ice ridge modelling is presented in Part II. Liferov and Bonnemaire (2004) give a review of the scaling of the ice rubble strength. Two types of structures were studied: cubical shape for sub-surface keel interaction and simple conical for surface ridge interaction. A plan view of the testing procedure is shown in Figure 1 and an artistic 3D view of the ridge interaction with one subsea structure is given in Figure 2. The principal dimensions of the structures are given in Figure 3. The structures were mounted on aluminium profiles fixed on the false bottom. A wooden board mounted in front of the profiles prevented the rubble ice from sliding under the subsea structure. The water depth for the four test series was respectively 1.2, 1.06, 1.14 and 1.16 m. The friction coefficient of the ice against the painted plywood of the subsea structures was measured to be 0.11. From previous HSVA experiments, it is estimated to 0.04 on the upper part of the cone and 0.1 on the lower part. Figure 1. Test set-up, plan view. Figure 2. 3D illustration view of the subsurface interaction.
Figure 3. Dimensions of the structures. Load measurement The load components in the three directions were measured. The cubes were equipped with a load cell system able to measure the loads acting on the entire cube (roof + sides, referred as total loads) and the loads acting only on the roof of the cube. In order to measure the keel load on the cone alone, its upper and lower parts were separated by a splice. The splice location was just below the consolidated layer of the ridge. Ice conditions Four ice sheets (corresponding to test series 1000, 2000, 3000 and 4000) were produced with one ice ridge per ice sheet. The objective was to produce as wide ridges as possible. The ice ridges were confined on the sides by the tank walls. They were separated from the end of the basin by 10 m of level ice frozen to the basin side walls. The ridges 2000 and 4000 were built from colder ice and tested soon after their production, with very little consolidation (degradation) time allowed. The ridge profiles and building process are reported in Part II together with detailed information on the ice properties. The position of the roof of the subsea cubes relative to the mean keel depth and the waterline is given in Table 1. Table 1. Cubes roof position relatively to the mean keel depth and the waterline, mean keel depth. Roof - mean keel depth [m] Waterline roof [m] Mean keel depth [m] Test series 1000 0.28 0.22 0.5 Test series 2000 0.3 0.1 0.4 Test series 3000 0.22 0.18 0.4 Test series 4000 0.26 0.19 0.45 Test matrix Table 2 shows the test matrix of the experiments.
1000-test series Test set-up OPEN WATER TEST LEVEL ICE PRODUCTION Level ice h i = 0.03 m ICE RIDGE PRODUCTION Target keel depth 0.5 m TESTING : v = 0.045 m/s Table 2. Test matrix (dimensions in model scale). Description 1. Performance testing of model and installed equipment. 2. Equipment and procedure open water testing and calibration. 3. Survey of monitoring system, determination of noise. 1. Tank cooled down. 2. Freezing of an ice sheet. 3. Monitoring and testing of the ice cover 1. Loosen the sheet along tank walls, move the whole sheet out, cut for ridge. 2. Ice sheet sequentially pushed into the ridge towards the transverse boom. 3. Monitoring and testing of ice properties of level ice. 4. Profile keel and sail 2 m off starboard and portside tank and at C/L. 5. Instrument ice ridges with thermistor strings. 6. The ridge is left to consolidate until the level ice flexural strength reaches a standard value of 30 kpa (warm ridges only). 1. Moving the subsea and conical structure through the ridge. 2000-test series: similar procedure, but the ridge is tested just after its production (within approx. 5 hours instead of 18 hours, no consolidation time), to obtain harder blocks of ice. 3000-test series: similar to 1000-test series, but the interaction speed v = 0.22 m/s. 4000-test series: similar to 2000-test series, but the interaction speed v = 0.22 m/s. TEST RESULTS CUBES General behaviour and load levels Figure 4 shows the x-load time series for the test series 1000. The loads in the y-direction were not significant, below 50 N. In all tests it was generally observed that the x-load rapidly increased to reach a plateau, and decreased just before the structures exited from the ridges. Figure 4. x-total load on the starboard cube (a) and the portside (b), test series 1000. Figure 5 shows the z-load of ridge 1000 on the portside cube. For each cube, the z-roof load and z-total load followed each other qualitatively but not quantitatively. The quantitative difference was not constant along the entire interaction period, and the maximum value of the roof load and total load did not occur at the same time.
Figure 5. z-total load on the portside cube (a) and z-roof load (b), test series 1000. The maximum values and average higher values for each test series are given in Table 3. The average higher value corresponds to the average of the values measured during the steady state, until the load starts the final decrease. It includes the possible valleys and peaks occurring during the steady state. Table 3 shows that the colder ridges (2000 and 4000) caused higher loads on the cubes. The loads from ridge 2000 were three times higher than from ridge 1000. Such a difference did not exist between test series 3000 and 4000. It should be noted that in test series 2000 the roof of the cube was only 0.1 m from the waterline, instead of usually 0.2 m. Table 3. Subsea interaction, maximum / average higher load levels. Portside cube Starboard cube x-load [N] z-load [N] roof z- x-load [N] z-load [N] roof z-load [N] load [N] Test series 1000 520 / 320 315 / 230 200 / 160 450 / 300 250 / 160 180 / 120 Test series 2000 1400 / 1000-400 / 315 1400 / 1000 500 / 400 400 / 300 Test series 3000 760* / 660* - 340 / 270 - - 325 / 240 Test series 4000 1000* / 750* - 400 / 300 - - 425 / 325 *Uncertain load measurements due to load cells damaged during the maintenance procedure between the test series 2000 and 3000. A calibration was performed in dry conditions at the end of the test campaign. Peaks and oscillations The x-load time plots did generally not show large peaks, except for the portside cube of test series 1000 and test series 4000. On the other hand, several peaks which seem to appear at a regular interval of 30 s were often visible in the plots of the z-loads. The z-loads presented relatively larger oscillations than the x-loads. Loads in function of the structure penetration Figure 6 shows the ridge profiles and the loads acting on the portside cube plotted against the position of the front face of the structure, for the test series 1000 and 4000 respectively. The maximum depth of the rubble accumulation in front of the cube was measured from underwater videos and is also included in Figure 6. The same plots were generated for each interaction event with the cubes. Their common characteristic was that the increase in x-load was delayed until the front face of the cube had penetrated about 0.5 m into the keel. The final load decrease always began approximately 0.5 to 1 m before the front face reached the boundary of the keel (1.5 m for the portside structure, test series 2000). The underwater videos showed that the load decreased before the rubble mound in front of the cubes started to be reduced. The z-loads were still acting on the roof even after the structure had passed the ridge. The z-loads however returned to 0 before the end of the tests.
Figure 6. Ridge profile, rubble mound and actions on the portside structure plotted against the position of the structure s front, test series 1000 (a) and 4000 (b). At a first glance, the x-loads seemed to correlate to the keel profile. They decreased before the valleys of the keel and increased in the vicinity of the peaks. This can be observed in Figure 6 (b). The x-load on the starboard cube for the interaction 1000 was showing its highest value at the beginning of the time plot. Accordingly, the ridge keel depth was slightly decreasing all along the interaction. However, some particularities contradicted the previous affirmation, i.e. the final x- load peak in Figure 6 (a). On the underwater video, this peak load occurred at the beginning of a 0.2 m increase of the depth of the rubble mound in front of the structure. Video observations For all the test series, some displacements could be observed at the surface of the ridge. Figure 7 shows the starboard cube (the square) with its front at 59 m tank position, i.e. 209s on the time series, and the corresponding deformations observed in the sail, for the test series 1000. The deformations seemed to be localized in the triangle between the two lines. Their maximum amplitude was in the shaded area, where x and z displacements of approximately 3 cm was observed. The corresponding underwater view of the starboard structure is presented in Figure 8. The graduations on the structure were separated by 10 cm. The displacements of the sail were estimated from video observations.
Figure 7. Sail deformation, starboard, 209 s, test series 1000. Figure 8. Rubble accumulation seen from behind, starboard, test series 1000, 209 s. In the test series 2000, the cube was only 0.1 m from the water line. Video observations from the surface revealed larger movements in the sail than in the other test series. At 117 s (Figure 9), the sail had a general tendency to heave around 1 cm in the triangular area between the lines. The behaviour seemed to be the same above the second cube. Larger sail displacement could also be observed in a localized area, like at 132 s (Figure 10) where the vertical and horizontal displacement could reach between 5 or 6 cm. Once the cube had been pushed through the ridge, the altitude of the sail above the wake was approximately 1 cm below its original value. Figure 9. Sail deformation, starboard, test series 2000, 117s. Figure 10. Localized larger deformation, starboard, test series 2000, 132s. Figure 11. Rubble accumulation portside, test series 2000, 139 s. Figure 11 shows the rubble accumulation on the portside structure during test series 2000. It accumulates on the whole front of the cubes with a front slope angle between 40 and 50. The front slope angle was located in this range for all the test series. The repose angle of a rubble pile in dry conditions was also between 40 and 50 (Part III, piling test).. The rubble accumulation reached the false bottom only in the test series 2000. TEST RESULTS CONE General behaviour and load levels During the cube interactions, a steady state was observed, but for the cone interactions the steady state did not really occur. Figure 12 shows the x-load on the bottom (referred as bottom x-load) and the top of the cone (referred as top x-load), and their sum (referred as cone x-load) plotted against the position of the cone axis for the test series 4000. It shared common characteristics with the three other test series: a steady state was reached for the top x-load; the bottom x-load
was higher than the top x-load; the loads did not return to zero at the end of the tests due to a rubble accumulation resting on the cone when full stop was reached. Figure 12. x-load on the cone plotted against the position of the cone axis, test series 4000. Figure 13. x-load on the cone plotted against the position of the cone axis, test series 2000. Figure 14. z-load on the cone plotted against the position of the cone axis, test series 2000. The z-loads on the cone for the test series 2000 are given in Figure 14. They showed the same behaviour as the x-loads of test series 2000. For the other test series, the z-loads showed a behaviour similar to the standard x-load behaviour shown in Figure 12. The maximum values for each test series are presented in Table 4. The colder ridges (2000 and 4000) caused higher loads on the cone. Table 4. Maximum loads on the cone. Cone top Cone bottom Total x-load / z-load [N] x-load / z-load [N] x-load / z-load [N] Test series 1000 584 / 450 902 / 590 1369 / 1000 Test series 2000 309 / 210 1521 / 1000 1744 / 1180 Test series 3000 549 / 380 1158 / 700 1410 / 930 Test series 4000 566 / 500 1290 / 700 1720 / 1140 Peaks and oscillations The x and z-loads of the test series 2000 (Figure 13 and Figure 14) showed a particular behaviour, with two large peaks observed at 58.5 m and 61 m tank position. The oscillations in the top x- load after 61 m for the test series 4000 could also be observed in the time series of the top x-load of test series 3000. Both these series had a high interaction speed (0.22 m/s). These oscillations occurred when the top of the cone encountered the level ice covering the ridge. Another type of oscillations with a 30 s period could also be observed clearly in the z-bottom load at the end of each test, when the cone was immobile and rubble was accumulated on it. These oscillations were also present in the x-bottom load with smaller amplitude. Video observations The videos showed that in all the test series except series 2000, the ridges always failed locally. The extent of the disturbed zone in the sail in front of the cone was oscillating during each interaction. In the test series 2000 a global failure could be observed. It occurred when the first peak in the x-load was reached (Figure 13), after the ridge had previously failed locally. The
failure pattern with three zones is represented in Figure 15. The videos revealed that the ridge was starting to bend in a three point bending mode, with the cone as the load application point and the tank walls as the supporting points. Then the bending continued only in zone 1, and the two other zones appeared. Zone 2 corresponded to a failure zone and zone 3 to a rubble plug pushed forward by the cone. On the portside of the ridge zone 2 was also observed but zone 1 was less noticeable. The ice sheet behind the ridge was neither moving nor failing. These observations were made from above water videos only, but zone 2 and 3 were also visible underwater (Figure 16). Zone 1 was not within the coverage of the underwater video cameras. After the cone axis had passed the 59 m position, the plug jammed and a local failure developed again. It corresponded in the time series of Figure 13 to the new increase of the load just after the valley. Figure 15. Side view of the global failure pattern, cone interaction, test series 2000; zone1- bending movement, zone 2 - fracture, zone 3 - plug. Figure 16. Global failure pattern seen from portside underwater, test series 2000; zone 2 -fracture, zone 3 - plug. In all the test series, the underwater videos showed a dead (stationary) wedge of rubble accumulating in front of the cone. It never reached the false bottom and its slope angle (to the horizontal) was between 30 and 45. It was smaller for the high speed test series (3000 and 4000). The altitude and circumference of the cone above the water line were not high enough to study the slope angle of the rubble accumulation above the water. DISCUSSION Friction loads on the subsea structures Figure 17 and Figure 18 show the z-loads and friction loads acting on the sides of the portside cube (test series 1000) and the starboard cube (test series 2000). The comparison was also performed for the starboard cube, test series 1000. The z-loads acting on all the sides were determined by equation (1) and the friction loads by equation (2): F S z F F (1) T z R z F T (2) f F x S where Fz is the z-load on the sides, friction load, T F z is the total z-load, R Fz is the z-load on the roof, Ff is the T F x is the total x-load and is the friction coefficient. A more accurate estimate of
the friction loads would require subtracting the x-roof load from the x-total load in equation (2). However the x-roof load was not measured. It was estimated to be negligible in comparison with the x-total load. Figure 17. z-side loads on the portside structure and estimated friction loads, test series 1000. Figure 18. z-side loads on the starboard structure and estimated friction loads, test series 2000. The estimated friction load followed the general trend of the measured load. However, some load peaks were underestimated (Figure 17). For the starboard cube of test series 2000, the frictional vertical loading represented approximately 25% of the total vertical loading. An increase of the vertical load on a bottom-based structure could present a stabilizing effect (Gürtner et al, 2008) Vertical load oscillations All the time plots of the z-load on the cubes seemed to present a regular oscillatory pattern of three oscillations per 100 s, as can be seen in Figure 5. These oscillations could also be observed when the cubes were at rest (no false bottom movement) under the level ice sheet at the end of the test 2000 (Figure 19). In the 2000 test series the cubes where only 10 cm below the water line. The oscillation period was 31 s. The same oscillations were also present in the time series of the vertical load on the cone, when the cone was at rest at the end of the tests and rubble was accumulated on it. A spectrum density analysis performed on the signal given in Figure 19 revealed an energy peak at the frequency 0.033 Hz, corresponding to a period of 30.3 s. An energy peak at the same frequency was found in the spectrum density plots of the time series issued from the z-load on each cube during their penetration into the ridges. Since these oscillations were present not only for each test series during the interaction phase but also during the resting period on the cone and on the cubes of test series 2000, they probably do not result from a particular behaviour of the ice ridge but rather from the test set up. Figure 19. z-total load oscillation on the portside cube at rest, end of test series 2000.
A standing wave could have been created in the basin due to the motion of the false bottom and the structures during the open water test before each test series. The time left between the open water test and the interaction test (approx. 10 min) was not enough to dampen the wave. This wave would have lifted and lowered the water line and the ridge by few millimeters. The oscillatory part of the vertical load signal would have therefore resulted from the rubble weight above the structures changing due to a gain / loss of buoyancy. This explanation is supported by the standing wave equation in a shallow water closed basin (Dean and Dalrymple, 1991): T 2l / gh (3) where T is the wave period, l is the length of the basin, g the gravity acceleration and h the water depth. For the 72 m long and 2.5 m deep HSVA basin, the estimated wave period is 29 s. Furthermore, the video recordings showed a 10 to 20 mm vertical oscillatory movement of the ice when the structures were at rest at the end of the test series. Load level in relation to the temperature and interaction speed. Table 3 and Table 4 show the load exerted by the ridges on the cubes and on the cone, for the four test series. The loads from the cold ridges (2000 and 4000) were higher, but no conclusion could be drawn regarding the effect of the interaction speed. However the effect of the ridge temperature was less obvious when looking at the effective pressure (Table 5). The effective pressure was here defined as the x-keel load divided by the area of interaction. The x-keel load is equal to the x-total load for the cubes, and to the x-bottom load for the cone. For the cubes, the area of interaction is computed as the product of the cube width and the distance from the roof to the average keel depth (Table 1). For the cone, the area of interaction is the projected area of the cone interacting with the keel (average keel depth). Table 5. Effective pressure. Portside cube (average) Starboard Cube (average) Cone bottom (maximum) Test series 1000 1630 Pa 1530 Pa 2840 Pa Test series 2000 4760 Pa 4760 Pa 4210 Pa Test series 3000 4285 Pa / 4210 Pa Test series 4000 4120 Pa / 4207 Pa The x-effective pressure on the cubes in series 3000 was slightly higher than in test series 4000. The load cells of the cubes were damaged during the maintenance procedure between the test series 2000 and 3000. It was therefore not possible to directly compare the loads on the cubes between the two first test series and the two last ones. The effective pressures on the cone bottom were identical for the three last test series. However the air temperature was negative between the ridge building and the testing for the test series 2000 and 4000 (Part II). Therefore a higher amount of freeze bonds could have developed in the keel close to the water surface during these two test series. Since the rubble interacting with the cubes and the cone bottom for the test series 2000 was close to the water line, it could explain why the load and effective pressure were high during these test series. Without this effect, a higher effective pressure might have been observed for the high speed interactions than for the slow speed.
The previous results are only indicative, and the rubble accumulation should be considered. Further work will focus on integrating the size of the rubble accumulation into the computation of the area of interaction. Failure mode The global failure is assumed to represent a situation where the failure surfaces reach the side of the ridge opposite to the structure (plug failure). It is opposed to the local failure, where the ridge fails only in the area close to the structure. In the cube interactions, the global failure of the keel was never observed. The load rather decreased only when the local failure zone reached the free boundary. A global failure occurred only for the cone interaction in the test series 2000, where the ice ridge and the ice sheet behind it were not frozen to the side walls of the tank. The ridge was also more damaged by the subsea structures than in the other test series, since the they were only 10 cm from the water level in test series 2000 (20 cm otherwise). This is most probably the main reason for the global failure to occurr. The failure zones seemed to develop diagonally until they reached the cubes wakes and then followed the wakes as shown in the sky view of the interaction 2000 (Figure 20). Figure 20. Sky view of the global failure, test series 2000. After the global failure had occurred, the ridge exerted lower load on the cone as long as the plug could be pushed forward by the structure. After some displacement the plug jammed and local failure took place again thus increasing the ridge action on the cone. The jamming could have been caused by the increase in the contact area between the moving plug and the overlaying stationary ice sheet. This event showed the incidence of the ridge failure mode on the keel action on a seabed structure. The global failure appeared as a load release mechanism. CONCLUSION The experimental set up and the preliminary results of the ridge keel structure interaction tests performed at HSVA in November 2008 are presented in this paper. Two 0.7 m-side cubical subsea structures and one 62 sloped upward breaking cone interacted with four ice ridges. Two different interaction velocities and two different ice ridge temperatures were used. The time series from the subsea structure interactions showed that the horizontal loads tended to be correlated with the ridge profile and that steady state was reached in all the interactions. Besides,
25% of their vertical loading could be caused by friction loads. The friction load could therefore increase the stability of vertical sided seabed structures. No particular frequency could be observed in the spectrum density functions, apart from a 30 s oscillation in the vertical load, caused by a standing water wave in the basin. The video observations revealed a general local failure mode of the rubble, apart from one cone interaction. This global failure event was well documented by time series and video observations. It illustrates the influence of the failure mechanism on the ridge action. It appears that colder ridges exerted higher loads but no definitive conclusions could be drawn regarding the effect of interaction speed or temperature on the keel action. Further analysis of the test results will be reported in later publications. The present paper is completed by a study of thermo-dynamics of model ice ridges and their mechanical properties in Part II: Ridge building and physical properties, part III: Investigation of model ice rubble mechanical properties and part IV: Preliminary results of freeze bond shear strength experiments. ACKNOWLEDGEMENTS The authors would like to thank NTNU and StatoilHydro for providing the means to realize these experiments. They also express their gratitude to Dr. Knut Høyland and Ada Repetto for their help and scientific advises. Andrea Haase, Hanne Hagen and Christian Lonøy are also particularly acknowledged for the precious help they provided during the test period. Prof. Sveinung Løset is much acknowledged for creating an opportunity so this research project became a reality. The work described in this report/publication was supported by the European Community s Sixth Framework Programme through the grant to the budget of the Integrated Infrastructure Initiative HYDRALAB III, Contract no. 022441(RII3). The authors would like to thank the Hamburg Ship Model Basin (HSVA) and the ice tank crew, for the hospitality, technical and scientific support and the professional execution of the test programme in the Research Infrastructure ARCTECLAB. REFERENCES Dean, R. G. and Dalrymple, R. A., 1991. Water Wave Mechanics for Engineers and Scientists, World Scientific, 353 pp. Gudmestad, O.T. and Liferov, P., 2007. Design of subsea equipment to withstand loads from ice, POAC 07. Gürtner, A. and Gudmestad, O.T., 2008. Innovative ice protection for shallow water drilling, part II, SIB model testing in ice, OMAE 2008 Liferov, P. and Bonnemaire, B., 2004. Ice rubble behaviour and strength: Part I. Review of testing and interpretation of results. Journal of Cold regions Science and Technology, 41/2 135-151. Liferov, P. and Høyland, K.V., 2004. In-situ ice ridge scour tests: Experimental set-up and basic results. Journal of Cold Regions Science and Technology, 40/1-2: 97-110. Repetto-Llamazares, A., Serré, N., Evers, K-U., Jochmann, P., Høyland, K. 2009a. Model testing of ridge keel loads on structures Part II: Ridge building and physical properties, POAC 2009.
Repetto-Llamazares, A., Jochmann, P., Evers, K-U., Høyland, K. 2009b. Model testing of ridge keel loads on structures Part IV: Preliminary results of freeze bond shear strength experiments, POAC 2009. Serré, N., Liferov, P., Jochmann, P., 2009. Model testing of ridge keel loads on structures Part III: Investigations of rubble ice mechanical properties, POAC 2009.