- Email: [email protected]
Effect of temperature and particle velocity on the erosion of a silicon carbide continuous fibre reinforced calcium aluminosilicate glass–ceramic matrix composite
Wear 233–235 Ž1999. 237–245 www.elsevier.comrlocaterwear
Effect of temperature and particle velocity on the erosion of a silicon carbide continuous fibre reinforced calcium aluminosilicate glass–ceramic matrix composite A.L. Ham, J.A. Yeomans ) , J.F. Watts School of Mechanical and Materials Engineering, UniÕersity of Surrey, Guildford, Surrey, GU2 5XH, UK
Abstract A unidirectional silicon carbide continuous fibre reinforced calcium aluminosilicate ŽCASrSiC. glass–ceramic matrix composite has been subjected to silica sand solid particle erosion over the temperature range 20–7268C. Initial tests were conducted with a constant mass flow rate of gas of 40 l miny1. This gave a wear rate of 0.16 mg gy1 at room temperature which increased to 0.26 " 0.02 mg gy1 at 4008C and above. When the increase in gas velocity, and hence particle velocity, was taken into account the wear rate was predicted to decrease with increasing temperature for a constant particle velocity. A wear rate of 0.06 mg gy1 was measured at 3008C, which showed excellent agreement with the prediction of 0.07 mg gy1. As the main mechanism of material removal is via lateral cracking, this decrease in wear rate is broadly consistent with the release of residual axial tensile stresses in the matrix resulting from mismatches in the coefficients of thermal expansion of the two phases on cooling down from the processing temperature. q 1999 Elsevier Science S.A. All rights reserved. Keywords: CASrSiC glass–ceramic matrix composite; Erosion; Particle velocity
1. Introduction There are potential advantages of replacing metallic components with engineering ceramics in conditions involving exposure to both high temperatures and erosive wear, yet very few experimental studies have been concerned with this topic we.g., Refs. w1–3xx. Also, over the last decade there has been considerable interest in the mechanical properties of ceramic matrix composites ŽCMCs.. Many of these materials have been targeted specifically at high temperature engine applications but few studies of the erosion resistance of these materials have been reported. As these materials are tougher than the parent matrices, an increase in wear resistance might be expected. It is, however, important to take into account the mechanisms of toughening and any microstructural instabilities. For example, it has been shown that the high temperature erosion resistance of a titanium diboride particulate toughened SiC matrix CMC can either be lower w4x or higher w5x than that of the parent material, depending on the experimental conditions. )
Corresponding author. Tel.: q44-01483-259613; fax: q44-01483259508; e-mail: [email protected]
The current work is concerned with a SiC continuous fibre reinforced calcium aluminosilicate ŽCAS. glass– ceramic matrix composite. Previous work by Powell et al. w6x showed that room temperature quasi-static indentation and single particle impact, above a critical load, generated lateral cracking. The size of the lateral cracks was, however, a function of the local microstructure. Residual tensile axial stresses in the matrix, resulting from the mismatch in the coefficients of thermal expansion of the fibres and the matrix, increase with local fibre volume fraction w7x lead to a greater propensity for lateral cracking in fibre-rich regions. Quasi-static indentation cracking data were used to make simple semi-empirical models to predict the wear rate of various potential microstructures w8x. As the residual stresses will decrease as the temperature is increased, it is hypothesised that the erosive wear rate will also decrease. This assumes that lateral cracking remains the dominant mechanism and that the reduction in residual stress is more significant than the softening of the matrix, which would be expected to lead to larger mechanical residual stresses around the impact site and hence enhanced lateral cracking, as observed in indentation studies of silicon w9x. Thus, the aim of the present work was to test this hypothesis.
0043-1648r99r$ - see front matter q 1999 Elsevier Science S.A. All rights reserved. PII: S 0 0 4 3 - 1 6 4 8 Ž 9 9 . 0 0 2 2 2 - 7
238
A.L. Ham et al.r Wear 233–235 (1999) 237–245
2. Target material and specimen preparation The target material was a unidirectional composite consisting of continuous silicon carbide fibre ŽNicalone. in a calcium aluminosilicate glass–ceramic matrix. It was produced by hot pressing pre-impregnated layers of fibres drawn through a CAS slurry. A plate of composite, measuring 150 = 150 = 2.2 mm, corresponding to 12 plies, was supplied by Rolls-Royce plc. The SiC fibres had a nominal fibre diameter of 15 mm, and were present at an overall fibre volume fraction of 0.34, although local variations were evident Žsee Fig. 1.. In order to test the predictions of Powell et al. w8x, it was necessary to orient the specimens such that the lengths of the fibres were parallel to the erodent steam. This meant that the plate had to be edge-on, i.e., a specimen was only 2.2 mm wide. This caused complications in terms of locating the specimen centrally within the erodent plume and enhanced edge effects, since the erodent plume was approximately 7 mm in diameter. Thus, bars of 20 mm = 2.2 mm = 9 mm were cut and held side by side, within a purpose built sample holder Žsee Fig. 2., to give a total target area of approximately 20 mm = 17 mm. The surface was then polished by hand, finishing with 1 mm diamond paste. As there were concerns that the multiple bar target may not wear at the same rate as a monolithic target would have done, the validity of the approach was verified by testing the composite in the other direction, i.e., with the lengths of the fibres perpendicular to the erodent stream. In
this orientation it was possible to make both types of target. 3. Erosion testing A schematic diagram of the erosion apparatus used in this study is shown in Fig. 3. The air passes through an in-line filter to remove any moisture and air borne particulates and enters the erosion rig through a three way valve connected to a safety cut-out solenoid. Air flow through the rig is controlled by three flow meters representing coarse Ž0–50 l miny1 . medium Ž0–5 l miny1 . and fine Ž0–0.6 l miny1 . controls, respectively. Erodent is fed into the air stream using a screw feeder driven by a variable feed motor. From the hopper, erodent particles are fed into a flexible reinforced polyvinylchloride pipe and down to a silica glass acceleration tube. The acceleration tube is situated in a pre-heater and protrudes into the furnace below. The sample is situated within the furnace at the required impact angle. An exhaust tube protrudes from the bottom of the furnace and leads to a water cooled cyclone and on to a vacuum filter system which removes all erodent particles before the gases exit via an extraction system. Thermocouples are located within the pre-heater section to provide a temperature readout and over-temperature alarm. Similarly, thermocouples in the main furnace are connected to left- and right-hand side main furnace over-temperature alarms and furnace temperature readout. When conducting erosion tests at elevated temperatures, samples were placed in the main furnace before heating
Fig. 1. Scanning electron photomicrograph showing the microstructure of a typical region of composite, with the fibre direction normal to the surface.
A.L. Ham et al.r Wear 233–235 (1999) 237–245
239
Fig. 2. Schematic representation of the specimen holder.
began and air was allowed to pass over the sample at a flow rate lower than that used during erosion. When the set temperature had been attained the air flow rate was increased and the screw feeder activated so that erosion
could take place. In the case of experiments between room temperature and 4008C both pre-heater and furnace were set at the temperature required and a brief period allowed between increasing flow rate and commencing erosion.
Fig. 3. Schematic representation of the high temperature erosion apparatus.
240
A.L. Ham et al.r Wear 233–235 (1999) 237–245
Fig. 4. Scanning electron photomicrographs of the wear scars produced at Ža. room temperature, Žb. 4008C and Žc. 7268C.
A.L. Ham et al.r Wear 233–235 (1999) 237–245
This permitted the temperature to equilibrate after the increased cooling effect of the elevated air flow rate was introduced to the system. Beyond temperatures of 4008C the pre-heater was set higher than the required erosion temperature to aid heating of the propellant gas and thus reduce this cooling effect, which became more marked at higher temperatures. The main heater temperature was also set to be higher than that required and erosion commenced when the required temperature was reached, again so the cooling effect of the increased gas flow rate could be minimised. Even so some change in temperature occurred during the course of erosion tests at these temperatures; this was monitored throughout the test and was never more than "108C of the quoted test temperature. Thermal cycles in the absence of erodent were used to confirm that mass changes in the target material were insignificant. The erodent used in this study was RH110 grade silica sand supplied by Hepworth Minerals and Chemicals. It was supplied washed and dried. The size distribution was narrowed by sieving to give particles between 100 and 150 mm. Particle velocity measurements within the erosion rig were measured using the double disc method of Ruff and Ives w10x. Initially, all tests were conducted with a gas flow rate of 40 l miny1 , which corresponds to a particle velocity of 24 m sy1 at room temperature. Before and after testing, samples were ultrasonically cleaned in acetone and weighed so that the wear rate could be expressed as a mass loss per mass of erodent. A Hitachi S3200N scanning electron microscope ŽSEM. with variable pressure facility was used to examine wear scars. The use of variable chamber vacuum meant that samples did not need to be coated prior to examination.
4. Results of initial experiments The wear rate of unidirectional material with fibres perpendicular to the direction of particle flow was 0.17 mg gy1 at room temperature, regardless of whether the target was a monolith or comprised multiple bars. Thus, edge effects caused by the use of bars of material to make up an area of suitable size for erosion testing do not lead to significant errors. When the fibres were parallel to the erodent stream, the wear rate was found to be 0.16 mg gy1 . The fact that erosion rates seen are similar when fibres are parallel or perpendicular to particle flow is initially surprising, since the residual tensile axial matrix stress will no longer be acting to aid lateral crack propagation. It has been shown by Powell et al. w7x, however, that there is also a tensile matrix hoop stress around fibres. Thus, when impact is parallel to the fibre direction the matrix axial stress will aid lateral cracking, but when the composite is oriented such that impact is perpendicular to the fibre direction the matrix tensile hoop stress will now be parallel to the direction of impact and will aid crack propagation.
241
For this composite, the maximum hoop stress was calculated to be 130 MPa. Given that this value will decrease with distance from the fibre–matrix interface, a mean value will not be dissimilar from the value of the axial tensile stress Ž; 80 MPa. at the same fibre volume fraction. It is, therefore, not unexpected that similar wear rates are obtained. Further, it is likely that any cross-ply composite of the same material will show similar wear rates, provided that stresses due to ply interactions are not significant. Initial studies of wear as a function of temperature were conducted at a gas flow rate of 40 l miny1 . Wear rate increased from the room temperature value of 0.16 mg gy1 , up to a value of around 26 mg gy1 at 2008C, after which it remained fairly constant. Examples of the wear scars formed at three temperatures are shown in Fig. 4. Lateral cracking appears to be the dominant mechanism of material removal in all cases but as the temperature increases the surfaces get smoother, presumably as each lateral crack is less extensive. It is also possible that some element of plastic deformation is occurring. Further, the fibre ends seem to be less distinct and level with the matrix at the highest test temperature. This could be a consequence of the carboneous interface between the fibres and the matrix being replaced by silica and the material reverting to monolithic behaviour, an effect reported by many workers looking at elevated temperature mechanical properties of this material we.g., Ref. w11xx. This would lead to lateral cracks passing through the fibres and creating smoother surfaces. It was decided that using a constant flow rate was not a useful basis from which to make a comparison of the performance of the material as the velocity of the propellant gas, and hence particle velocities, would be increasing with temperature. Thus a number of experiments were conducted to assess this effect.
5. Gas and particle velocities as a function of temperature The average velocity of the carrier gas at room temperature can be calculated simply by dividing the volume flow rate by the nozzle exit area. At room temperature the volume flow rate at the nozzle exit is the same as the measured value at the flow meters, but at other temperatures expansion of the gas must be taken into consideration. At room temperature it was also possible to measure gas velocities by means of a pitot tube connected to a digital manometer. Since there will, in fact, be a range of gas velocities across the diameter of the nozzle the position of the pitot tube was adjusted until a maximum reading was achieved. In this way average air velocities could be calculated for comparison with the average velocities calculated from the flow readings.
242
A.L. Ham et al.r Wear 233–235 (1999) 237–245
Fig. 5. Gas and particle velocities as a function of air flow at room temperature.
In order to assess the effect these predicted increases in gas velocity would have on particle velocity, double disc measurements were also carried out. It was not possible to measure particle velocity and operate the main furnace at elevated temperatures due to the need to remove and handle equipment and the temperature limitations of the apparatus itself. It was, however, possible to use the pre-heater, situated above the main furnace, to heat the particle gas mixture within the silica acceleration tube, as would occur in a normal erosion test. Pre-heater temperatures of 1008C, 2008C, 3008C and 4008C were used, with 4008C being the maximum temperature that experiments could be carried out at without damaging the apparatus.
Fig. 5 shows the results of the pitot tube readings, velocities calculated from air flow rates and particle velocities from double disc measurements vs. gas flow rate at room temperature. The agreement between calculated and measured gas flow rates is reasonably good. Differences may arise from the approximate nature of the expression for calculating average profile velocity from pitot tube readings and assumptions such as fully developed turbulent flow in the acceleration tube. It can also be seen that at all gas flow rates the particle velocity is lower than the gas velocity. Fig. 6 shows the data for calculated air velocity using volume increase and mean particle velocity measurements,
Fig. 6. Gas and particle velocities at a constant flow rate of 40 l miny1 as a function of temperature.
A.L. Ham et al.r Wear 233–235 (1999) 237–245
made using the double disc method, up to a temperature of 4008C. As expected the expansion of gas on heating leads to an increase in gas velocity in order for the mass flow rate to be maintained. This in turn leads to the increased particle velocity seen in Fig. 6. The increase in particle velocity with temperature is non-linear and appears to be reaching a plateau at around 300–4008C. The data presented for gas velocities are calculated assuming that the gas has attained the temperature indicated. In practice, however, the air needs to spend a finite amount of time in the pre-heater to reach the set temperature. At the higher temperatures the carrier gas is probably not spending enough time in the acceleration tube to reach the temperature set in the pre-heater. This suggestion is supported by the observation that as temperature is increased there is an increased cooling effect in the main furnace as the air flow is increased ready for testing Žsee above.. This tendency towards a maximum gas temperature could lead to the plateau of particle velocities seen in Fig. 6.
6. Wear rate as a function of velocity In order to assess the effect in the change of particle velocity with temperature, it was necessary to measure wear rate as a function of velocity. This was done at room temperature and the results are shown in Fig. 7. The two highest velocities were achieved by modifying the rig such that the flow meters were by-passed and the gas flow was controlled by a regulator. As reported previously w12x, there is a change in erosion mechanism at approximately 12 m sy1 , as indicated by the upturn in erosion rate after this point. SEM studies have shown that at the lower
243
velocities material removal is by means of a surface chipping mechanism with the fibres standing proud of the surrounding matrix. The edges and ends of the fibres have been chipped but there is no evidence of large scale fracture in either fibres or matrix. At the higher velocities there is more widespread, presumably lateral, cracking with fibres having fractured at the same level as the matrix. These observations are broadly consistent with the findings of Powell et al. w8x. Despite this change in mechanism, a value of 2.6 for the velocity exponent fits all the data.
7. Wear rate as a function of temperature The effect of increasing particle velocity Žas a consequence of increasing gas temperature. can be taken into account, using the data from Fig. 7, to give a predicted wear rate at each temperature, assuming that the temperature has no effect other than to increase the velocity of the impacting particles. This is shown in Fig. 8. These data are above the actual experimental data, indicating that the increase in erosion temperature must be reducing the wear rate. A more useful way of analysing the data is to correct for the change in velocity by calculating what the wear rate would have been had the velocity been maintained at 24 m sy1 at all temperatures. It can be seen from the lower curve plotted in Fig. 8 that when the effect of increasing particle velocity, caused by increasing propellant gas temperature, is removed from the original wear rate vs. temperature results there is a decrease in wear rate with increasing temperature. In order to validate this approach, an erosion test was then carried
Fig. 7. Wear rate as a function of velocity at room temperature.
244
A.L. Ham et al.r Wear 233–235 (1999) 237–245
Fig. 8. Wear rate as a function of temperature.
out at a temperature of 3008C and a gas flow rate of 22 l miny1 Žparticle velocity of 24 m sy1 . which yielded a wear rate of 0.06 mg gy1 . This value is in excellent agreement with the predicted value of 0.07 mg gy1 . Unfortunately, lack of material prevented further experimentation. The closeness of the predicted and measured wear rates supports the assumption that the velocity exponent at 3008C is fairly close to the room temperature value of 2.6 which is contrary to the findings of Zhou and Bahadur w1x who reported a decrease from 2.6 at room temperature to 1.6 at 6508C when testing a range of aluminas. As discussed earlier, it was predicted that an increase in erosion temperature would lead to a decrease in wear rate due to the release of residual stresses in the composite matrix. The results of wear rate vs. temperature, after normalisation with respect to velocity, now support with this prediction and suggest that the reduction in residual stress is significant in terms of the wear behaviour of this material, from room temperature to 300–4008C. At higher temperatures, however, the wear rate is constant, despite the fact that the residual stresses will continue to decrease until the processing temperature is attained. This might indicate that softening of the matrix is increasing in significance and balancing the effect of the stress reduction.
8. Concluding remarks The erosion of a silicon carbide continuous fibre-calcium aluminosilicate glass–ceramic matrix composite has been shown to occur by a mechanism of lateral cracking and to be strongly dependent on the velocity of the erodent
particles Žwith an exponent of 2.6. but independent of the fibre orientation with respect to the erodent stream. At a constant velocity, the wear rate has been shown to decrease from room temperature to 300–4008C, which is consistent with a reduction in the driving force for lateral cracking as a result of the release of residual stresses resulting from mismatches in the coefficients of thermal expansion of the two constituents.
Acknowledgements A.L.H. is grateful to the EPSRC for financial support. We thank Rolls-Royce plc for the provision of the ceramic composite plate.
References w1x J. Zhou, S. Bahadur, Erosion characteristics of alumina ceramics at high temperatures, Wear 181r183 Ž1995. 178–188. w2x J.E. Ritter, S.R. Choi, K. Jakus, P.J. Whelan, R.G. Rateick Jr., Effect of microstructure on the erosion and impact damage of sintered silicon nitride, J. Mater. Sci. 26 Ž199. 5543–5546. w3x J.E. Ritter, L. Rosenfeld, K. Jakus, Erosion and strength degradation in alumina, Wear 111 Ž1986. 335–346. w4x A.F. Colclough, J.A. Yeomans, The elevated temperature erosion behaviour of SiC and SiC–TiB 2 , J. Hard Mater. 3 Ž1992. 53–61. w5x A.F. Colclough, J.A. Yeomans, Hard particle erosion of silicon carbide and silicon carbide–titanium diboride from room temperature to 10008C, Wear 209 Ž1997. 229–236. w6x K.L. Powell, J.A. Yeomans, P.A. Smith, Localised damage in continuous fibre reinforced ceramic matrix composites, Ceram. Eng. Sci. 14 Ž1993. 890–895.
A.L. Ham et al.r Wear 233–235 (1999) 237–245 w7x K.L. Powell, P.A. Smith, J.A. Yeomans, Aspects of residual thermal stresses in continuous-fibre-reinforced ceramic matrix composites, Compos. Sci. Technol. 47 Ž1993. 359–367. w8x K.L. Powell, J.A. Yeomans, P.A. Smith, A study of the erosive wear behaviour of continuous fibre reinforced ceramic matrix composites, Acta Metall. Mater. 45 Ž1997. 321–330. w9x M.S. Naylor, T.F. Page, Microstructural studies of the temperaturedependence of deformation structures around hardness indentations in ceramics, J. Microsc. 130 Ž1983. 345–360.
245
w10x A.W. Ruff, L.K. Ives, Measurement of solid particle velocity in erosive wear, Wear 35 Ž1975. 195–199. w11x R.F. Cooper, K. Chyung, Structure and chemistry of fibre–matrix interfaces in silicon carbide-reinforced glass–ceramic composites: an electron microscopy study, J. Mater. Sci. 22 Ž1987. 3148–3160. w12x A.L. Ham, J.A. Yeomans, J.F. Watts, Elevated temperature sold particle erosion of silicon carbide continuous fibre reinforced calcium aluminosilicate glass–ceramic composite, Wear 203r204 Ž1997. 387–392.
Effect of temperature and particle velocity on the erosion of a silicon carbide continuous fibre reinforced calcium aluminosilicate glass–ceramic matrix composite A.L. Ham, J.A. Yeomans ) , J.F. Watts School of Mechanical and Materials Engineering, UniÕersity of Surrey, Guildford, Surrey, GU2 5XH, UK
Abstract A unidirectional silicon carbide continuous fibre reinforced calcium aluminosilicate ŽCASrSiC. glass–ceramic matrix composite has been subjected to silica sand solid particle erosion over the temperature range 20–7268C. Initial tests were conducted with a constant mass flow rate of gas of 40 l miny1. This gave a wear rate of 0.16 mg gy1 at room temperature which increased to 0.26 " 0.02 mg gy1 at 4008C and above. When the increase in gas velocity, and hence particle velocity, was taken into account the wear rate was predicted to decrease with increasing temperature for a constant particle velocity. A wear rate of 0.06 mg gy1 was measured at 3008C, which showed excellent agreement with the prediction of 0.07 mg gy1. As the main mechanism of material removal is via lateral cracking, this decrease in wear rate is broadly consistent with the release of residual axial tensile stresses in the matrix resulting from mismatches in the coefficients of thermal expansion of the two phases on cooling down from the processing temperature. q 1999 Elsevier Science S.A. All rights reserved. Keywords: CASrSiC glass–ceramic matrix composite; Erosion; Particle velocity
1. Introduction There are potential advantages of replacing metallic components with engineering ceramics in conditions involving exposure to both high temperatures and erosive wear, yet very few experimental studies have been concerned with this topic we.g., Refs. w1–3xx. Also, over the last decade there has been considerable interest in the mechanical properties of ceramic matrix composites ŽCMCs.. Many of these materials have been targeted specifically at high temperature engine applications but few studies of the erosion resistance of these materials have been reported. As these materials are tougher than the parent matrices, an increase in wear resistance might be expected. It is, however, important to take into account the mechanisms of toughening and any microstructural instabilities. For example, it has been shown that the high temperature erosion resistance of a titanium diboride particulate toughened SiC matrix CMC can either be lower w4x or higher w5x than that of the parent material, depending on the experimental conditions. )
Corresponding author. Tel.: q44-01483-259613; fax: q44-01483259508; e-mail: [email protected]
The current work is concerned with a SiC continuous fibre reinforced calcium aluminosilicate ŽCAS. glass– ceramic matrix composite. Previous work by Powell et al. w6x showed that room temperature quasi-static indentation and single particle impact, above a critical load, generated lateral cracking. The size of the lateral cracks was, however, a function of the local microstructure. Residual tensile axial stresses in the matrix, resulting from the mismatch in the coefficients of thermal expansion of the fibres and the matrix, increase with local fibre volume fraction w7x lead to a greater propensity for lateral cracking in fibre-rich regions. Quasi-static indentation cracking data were used to make simple semi-empirical models to predict the wear rate of various potential microstructures w8x. As the residual stresses will decrease as the temperature is increased, it is hypothesised that the erosive wear rate will also decrease. This assumes that lateral cracking remains the dominant mechanism and that the reduction in residual stress is more significant than the softening of the matrix, which would be expected to lead to larger mechanical residual stresses around the impact site and hence enhanced lateral cracking, as observed in indentation studies of silicon w9x. Thus, the aim of the present work was to test this hypothesis.
0043-1648r99r$ - see front matter q 1999 Elsevier Science S.A. All rights reserved. PII: S 0 0 4 3 - 1 6 4 8 Ž 9 9 . 0 0 2 2 2 - 7
238
A.L. Ham et al.r Wear 233–235 (1999) 237–245
2. Target material and specimen preparation The target material was a unidirectional composite consisting of continuous silicon carbide fibre ŽNicalone. in a calcium aluminosilicate glass–ceramic matrix. It was produced by hot pressing pre-impregnated layers of fibres drawn through a CAS slurry. A plate of composite, measuring 150 = 150 = 2.2 mm, corresponding to 12 plies, was supplied by Rolls-Royce plc. The SiC fibres had a nominal fibre diameter of 15 mm, and were present at an overall fibre volume fraction of 0.34, although local variations were evident Žsee Fig. 1.. In order to test the predictions of Powell et al. w8x, it was necessary to orient the specimens such that the lengths of the fibres were parallel to the erodent steam. This meant that the plate had to be edge-on, i.e., a specimen was only 2.2 mm wide. This caused complications in terms of locating the specimen centrally within the erodent plume and enhanced edge effects, since the erodent plume was approximately 7 mm in diameter. Thus, bars of 20 mm = 2.2 mm = 9 mm were cut and held side by side, within a purpose built sample holder Žsee Fig. 2., to give a total target area of approximately 20 mm = 17 mm. The surface was then polished by hand, finishing with 1 mm diamond paste. As there were concerns that the multiple bar target may not wear at the same rate as a monolithic target would have done, the validity of the approach was verified by testing the composite in the other direction, i.e., with the lengths of the fibres perpendicular to the erodent stream. In
this orientation it was possible to make both types of target. 3. Erosion testing A schematic diagram of the erosion apparatus used in this study is shown in Fig. 3. The air passes through an in-line filter to remove any moisture and air borne particulates and enters the erosion rig through a three way valve connected to a safety cut-out solenoid. Air flow through the rig is controlled by three flow meters representing coarse Ž0–50 l miny1 . medium Ž0–5 l miny1 . and fine Ž0–0.6 l miny1 . controls, respectively. Erodent is fed into the air stream using a screw feeder driven by a variable feed motor. From the hopper, erodent particles are fed into a flexible reinforced polyvinylchloride pipe and down to a silica glass acceleration tube. The acceleration tube is situated in a pre-heater and protrudes into the furnace below. The sample is situated within the furnace at the required impact angle. An exhaust tube protrudes from the bottom of the furnace and leads to a water cooled cyclone and on to a vacuum filter system which removes all erodent particles before the gases exit via an extraction system. Thermocouples are located within the pre-heater section to provide a temperature readout and over-temperature alarm. Similarly, thermocouples in the main furnace are connected to left- and right-hand side main furnace over-temperature alarms and furnace temperature readout. When conducting erosion tests at elevated temperatures, samples were placed in the main furnace before heating
Fig. 1. Scanning electron photomicrograph showing the microstructure of a typical region of composite, with the fibre direction normal to the surface.
A.L. Ham et al.r Wear 233–235 (1999) 237–245
239
Fig. 2. Schematic representation of the specimen holder.
began and air was allowed to pass over the sample at a flow rate lower than that used during erosion. When the set temperature had been attained the air flow rate was increased and the screw feeder activated so that erosion
could take place. In the case of experiments between room temperature and 4008C both pre-heater and furnace were set at the temperature required and a brief period allowed between increasing flow rate and commencing erosion.
Fig. 3. Schematic representation of the high temperature erosion apparatus.
240
A.L. Ham et al.r Wear 233–235 (1999) 237–245
Fig. 4. Scanning electron photomicrographs of the wear scars produced at Ža. room temperature, Žb. 4008C and Žc. 7268C.
A.L. Ham et al.r Wear 233–235 (1999) 237–245
This permitted the temperature to equilibrate after the increased cooling effect of the elevated air flow rate was introduced to the system. Beyond temperatures of 4008C the pre-heater was set higher than the required erosion temperature to aid heating of the propellant gas and thus reduce this cooling effect, which became more marked at higher temperatures. The main heater temperature was also set to be higher than that required and erosion commenced when the required temperature was reached, again so the cooling effect of the increased gas flow rate could be minimised. Even so some change in temperature occurred during the course of erosion tests at these temperatures; this was monitored throughout the test and was never more than "108C of the quoted test temperature. Thermal cycles in the absence of erodent were used to confirm that mass changes in the target material were insignificant. The erodent used in this study was RH110 grade silica sand supplied by Hepworth Minerals and Chemicals. It was supplied washed and dried. The size distribution was narrowed by sieving to give particles between 100 and 150 mm. Particle velocity measurements within the erosion rig were measured using the double disc method of Ruff and Ives w10x. Initially, all tests were conducted with a gas flow rate of 40 l miny1 , which corresponds to a particle velocity of 24 m sy1 at room temperature. Before and after testing, samples were ultrasonically cleaned in acetone and weighed so that the wear rate could be expressed as a mass loss per mass of erodent. A Hitachi S3200N scanning electron microscope ŽSEM. with variable pressure facility was used to examine wear scars. The use of variable chamber vacuum meant that samples did not need to be coated prior to examination.
4. Results of initial experiments The wear rate of unidirectional material with fibres perpendicular to the direction of particle flow was 0.17 mg gy1 at room temperature, regardless of whether the target was a monolith or comprised multiple bars. Thus, edge effects caused by the use of bars of material to make up an area of suitable size for erosion testing do not lead to significant errors. When the fibres were parallel to the erodent stream, the wear rate was found to be 0.16 mg gy1 . The fact that erosion rates seen are similar when fibres are parallel or perpendicular to particle flow is initially surprising, since the residual tensile axial matrix stress will no longer be acting to aid lateral crack propagation. It has been shown by Powell et al. w7x, however, that there is also a tensile matrix hoop stress around fibres. Thus, when impact is parallel to the fibre direction the matrix axial stress will aid lateral cracking, but when the composite is oriented such that impact is perpendicular to the fibre direction the matrix tensile hoop stress will now be parallel to the direction of impact and will aid crack propagation.
241
For this composite, the maximum hoop stress was calculated to be 130 MPa. Given that this value will decrease with distance from the fibre–matrix interface, a mean value will not be dissimilar from the value of the axial tensile stress Ž; 80 MPa. at the same fibre volume fraction. It is, therefore, not unexpected that similar wear rates are obtained. Further, it is likely that any cross-ply composite of the same material will show similar wear rates, provided that stresses due to ply interactions are not significant. Initial studies of wear as a function of temperature were conducted at a gas flow rate of 40 l miny1 . Wear rate increased from the room temperature value of 0.16 mg gy1 , up to a value of around 26 mg gy1 at 2008C, after which it remained fairly constant. Examples of the wear scars formed at three temperatures are shown in Fig. 4. Lateral cracking appears to be the dominant mechanism of material removal in all cases but as the temperature increases the surfaces get smoother, presumably as each lateral crack is less extensive. It is also possible that some element of plastic deformation is occurring. Further, the fibre ends seem to be less distinct and level with the matrix at the highest test temperature. This could be a consequence of the carboneous interface between the fibres and the matrix being replaced by silica and the material reverting to monolithic behaviour, an effect reported by many workers looking at elevated temperature mechanical properties of this material we.g., Ref. w11xx. This would lead to lateral cracks passing through the fibres and creating smoother surfaces. It was decided that using a constant flow rate was not a useful basis from which to make a comparison of the performance of the material as the velocity of the propellant gas, and hence particle velocities, would be increasing with temperature. Thus a number of experiments were conducted to assess this effect.
5. Gas and particle velocities as a function of temperature The average velocity of the carrier gas at room temperature can be calculated simply by dividing the volume flow rate by the nozzle exit area. At room temperature the volume flow rate at the nozzle exit is the same as the measured value at the flow meters, but at other temperatures expansion of the gas must be taken into consideration. At room temperature it was also possible to measure gas velocities by means of a pitot tube connected to a digital manometer. Since there will, in fact, be a range of gas velocities across the diameter of the nozzle the position of the pitot tube was adjusted until a maximum reading was achieved. In this way average air velocities could be calculated for comparison with the average velocities calculated from the flow readings.
242
A.L. Ham et al.r Wear 233–235 (1999) 237–245
Fig. 5. Gas and particle velocities as a function of air flow at room temperature.
In order to assess the effect these predicted increases in gas velocity would have on particle velocity, double disc measurements were also carried out. It was not possible to measure particle velocity and operate the main furnace at elevated temperatures due to the need to remove and handle equipment and the temperature limitations of the apparatus itself. It was, however, possible to use the pre-heater, situated above the main furnace, to heat the particle gas mixture within the silica acceleration tube, as would occur in a normal erosion test. Pre-heater temperatures of 1008C, 2008C, 3008C and 4008C were used, with 4008C being the maximum temperature that experiments could be carried out at without damaging the apparatus.
Fig. 5 shows the results of the pitot tube readings, velocities calculated from air flow rates and particle velocities from double disc measurements vs. gas flow rate at room temperature. The agreement between calculated and measured gas flow rates is reasonably good. Differences may arise from the approximate nature of the expression for calculating average profile velocity from pitot tube readings and assumptions such as fully developed turbulent flow in the acceleration tube. It can also be seen that at all gas flow rates the particle velocity is lower than the gas velocity. Fig. 6 shows the data for calculated air velocity using volume increase and mean particle velocity measurements,
Fig. 6. Gas and particle velocities at a constant flow rate of 40 l miny1 as a function of temperature.
A.L. Ham et al.r Wear 233–235 (1999) 237–245
made using the double disc method, up to a temperature of 4008C. As expected the expansion of gas on heating leads to an increase in gas velocity in order for the mass flow rate to be maintained. This in turn leads to the increased particle velocity seen in Fig. 6. The increase in particle velocity with temperature is non-linear and appears to be reaching a plateau at around 300–4008C. The data presented for gas velocities are calculated assuming that the gas has attained the temperature indicated. In practice, however, the air needs to spend a finite amount of time in the pre-heater to reach the set temperature. At the higher temperatures the carrier gas is probably not spending enough time in the acceleration tube to reach the temperature set in the pre-heater. This suggestion is supported by the observation that as temperature is increased there is an increased cooling effect in the main furnace as the air flow is increased ready for testing Žsee above.. This tendency towards a maximum gas temperature could lead to the plateau of particle velocities seen in Fig. 6.
6. Wear rate as a function of velocity In order to assess the effect in the change of particle velocity with temperature, it was necessary to measure wear rate as a function of velocity. This was done at room temperature and the results are shown in Fig. 7. The two highest velocities were achieved by modifying the rig such that the flow meters were by-passed and the gas flow was controlled by a regulator. As reported previously w12x, there is a change in erosion mechanism at approximately 12 m sy1 , as indicated by the upturn in erosion rate after this point. SEM studies have shown that at the lower
243
velocities material removal is by means of a surface chipping mechanism with the fibres standing proud of the surrounding matrix. The edges and ends of the fibres have been chipped but there is no evidence of large scale fracture in either fibres or matrix. At the higher velocities there is more widespread, presumably lateral, cracking with fibres having fractured at the same level as the matrix. These observations are broadly consistent with the findings of Powell et al. w8x. Despite this change in mechanism, a value of 2.6 for the velocity exponent fits all the data.
7. Wear rate as a function of temperature The effect of increasing particle velocity Žas a consequence of increasing gas temperature. can be taken into account, using the data from Fig. 7, to give a predicted wear rate at each temperature, assuming that the temperature has no effect other than to increase the velocity of the impacting particles. This is shown in Fig. 8. These data are above the actual experimental data, indicating that the increase in erosion temperature must be reducing the wear rate. A more useful way of analysing the data is to correct for the change in velocity by calculating what the wear rate would have been had the velocity been maintained at 24 m sy1 at all temperatures. It can be seen from the lower curve plotted in Fig. 8 that when the effect of increasing particle velocity, caused by increasing propellant gas temperature, is removed from the original wear rate vs. temperature results there is a decrease in wear rate with increasing temperature. In order to validate this approach, an erosion test was then carried
Fig. 7. Wear rate as a function of velocity at room temperature.
244
A.L. Ham et al.r Wear 233–235 (1999) 237–245
Fig. 8. Wear rate as a function of temperature.
out at a temperature of 3008C and a gas flow rate of 22 l miny1 Žparticle velocity of 24 m sy1 . which yielded a wear rate of 0.06 mg gy1 . This value is in excellent agreement with the predicted value of 0.07 mg gy1 . Unfortunately, lack of material prevented further experimentation. The closeness of the predicted and measured wear rates supports the assumption that the velocity exponent at 3008C is fairly close to the room temperature value of 2.6 which is contrary to the findings of Zhou and Bahadur w1x who reported a decrease from 2.6 at room temperature to 1.6 at 6508C when testing a range of aluminas. As discussed earlier, it was predicted that an increase in erosion temperature would lead to a decrease in wear rate due to the release of residual stresses in the composite matrix. The results of wear rate vs. temperature, after normalisation with respect to velocity, now support with this prediction and suggest that the reduction in residual stress is significant in terms of the wear behaviour of this material, from room temperature to 300–4008C. At higher temperatures, however, the wear rate is constant, despite the fact that the residual stresses will continue to decrease until the processing temperature is attained. This might indicate that softening of the matrix is increasing in significance and balancing the effect of the stress reduction.
8. Concluding remarks The erosion of a silicon carbide continuous fibre-calcium aluminosilicate glass–ceramic matrix composite has been shown to occur by a mechanism of lateral cracking and to be strongly dependent on the velocity of the erodent
particles Žwith an exponent of 2.6. but independent of the fibre orientation with respect to the erodent stream. At a constant velocity, the wear rate has been shown to decrease from room temperature to 300–4008C, which is consistent with a reduction in the driving force for lateral cracking as a result of the release of residual stresses resulting from mismatches in the coefficients of thermal expansion of the two constituents.
Acknowledgements A.L.H. is grateful to the EPSRC for financial support. We thank Rolls-Royce plc for the provision of the ceramic composite plate.
References w1x J. Zhou, S. Bahadur, Erosion characteristics of alumina ceramics at high temperatures, Wear 181r183 Ž1995. 178–188. w2x J.E. Ritter, S.R. Choi, K. Jakus, P.J. Whelan, R.G. Rateick Jr., Effect of microstructure on the erosion and impact damage of sintered silicon nitride, J. Mater. Sci. 26 Ž199. 5543–5546. w3x J.E. Ritter, L. Rosenfeld, K. Jakus, Erosion and strength degradation in alumina, Wear 111 Ž1986. 335–346. w4x A.F. Colclough, J.A. Yeomans, The elevated temperature erosion behaviour of SiC and SiC–TiB 2 , J. Hard Mater. 3 Ž1992. 53–61. w5x A.F. Colclough, J.A. Yeomans, Hard particle erosion of silicon carbide and silicon carbide–titanium diboride from room temperature to 10008C, Wear 209 Ž1997. 229–236. w6x K.L. Powell, J.A. Yeomans, P.A. Smith, Localised damage in continuous fibre reinforced ceramic matrix composites, Ceram. Eng. Sci. 14 Ž1993. 890–895.
A.L. Ham et al.r Wear 233–235 (1999) 237–245 w7x K.L. Powell, P.A. Smith, J.A. Yeomans, Aspects of residual thermal stresses in continuous-fibre-reinforced ceramic matrix composites, Compos. Sci. Technol. 47 Ž1993. 359–367. w8x K.L. Powell, J.A. Yeomans, P.A. Smith, A study of the erosive wear behaviour of continuous fibre reinforced ceramic matrix composites, Acta Metall. Mater. 45 Ž1997. 321–330. w9x M.S. Naylor, T.F. Page, Microstructural studies of the temperaturedependence of deformation structures around hardness indentations in ceramics, J. Microsc. 130 Ž1983. 345–360.
245
w10x A.W. Ruff, L.K. Ives, Measurement of solid particle velocity in erosive wear, Wear 35 Ž1975. 195–199. w11x R.F. Cooper, K. Chyung, Structure and chemistry of fibre–matrix interfaces in silicon carbide-reinforced glass–ceramic composites: an electron microscopy study, J. Mater. Sci. 22 Ž1987. 3148–3160. w12x A.L. Ham, J.A. Yeomans, J.F. Watts, Elevated temperature sold particle erosion of silicon carbide continuous fibre reinforced calcium aluminosilicate glass–ceramic composite, Wear 203r204 Ž1997. 387–392.