c01052e5 lence ELSEVIER Composites: Part A 34(2003)163-170 Fabrication and testing of all-oxide cfcc tubes M.G. Holmquist& b, T C. Radsick2, O.H. Sudre F.F. Lange, k Department of Materials, University of Califomia at Sa ara. Santa Barbara. CA 93106 USA vOlvo Aero Corporation, 4618 intan, Sweden Rockwell Scientific Company, 1049 Camino Dos housand Oaks. CA 91360. USA Received 15 August 2001: revised 15 September 2002: accepted 23 October 2002 Abstract A novel, low-cost processing method was used to manufacture all-oxide ceramic composite tubes. Fibre cloths were infiltrated with a pre consolidated slurry composed of a mixture of mullite and alumina powders. The pre-consolidated slurry was first formulated at a low volume fraction of powder to produce a short-range repulsive interparticle pair potential that allowed consolidation via pressure filtration, yet produced a consolidated body that could be fluidised by vibration. Efficient infiltration of fibre tows and a homogeneous microstructure were demonstrated. The infiltrated cloths, which could be frozen for latter use, were rolled to tubular shapes. After drying, the powder matrix strengthened by cyclic infiltration and pyrolysis of an alumina precursor. When the tubes were pressurised, they delaminated at an average hoop stress of 47 MPa Delamination initiated within the porous matrix where the outer most layer terminated. A failure analysis is presented using a strain energy release rate function for this specimen/crack configuration. Values of fracture energy inferred from the analysis were in close agreement with literature data for porous ceramic Published by Elsevier Science Ltd. Keywords: E. Powder processing: E Prepreg: B. Fracture: Oxide composites 1. Introduction on the fibres. which must be isolated from matrix cracks to take advantage of their high strength. Two different All-oxide continuous fibre reinforced ceramic compo- strategies for isolating fibres from matrix cracks can be sites(CFCCs) have attracted growing interest for use distinguished; one is to develop a crack deflecting interface high temperature applications that include gas turbine between the fibres and the surrounding matrix [13-18] ngines[1-6], hot gas filters [7, 8]and thermal protection second is to produce a porous matrix [5,18,19,35].The systems [9]. These composites are stable in oxidising failure mechanisms of composites fabricated with a porous environments. However, applications are limited to 1000- matrix have been studied and described in some detail 6, 20 1200C, due to the currently available fibres [10]. More 21]. Upon loading the stress/strain behaviour is nearly conventional, SiC based, CFCCs may find applications at linear, yet the matrix continuously microcracks before the higher temperatures(=1200C), although their life may onset of fibre failure [6, 20]. The fibres are isolated from the be shorter because of their susceptibility to oxidative matrix cracks because a continuous crack front cannot exist embrittlement at stresses that exceed the matrix strength within a porous matrix. Namely, crack extension within the [11,12」 matrix occurs by the breaking of grain pairs at grain Apart from fibre development, major efforts are boundaries within the matrix. when the matrix becomes focused on the issue of making oxide CFCCs damage ufficiently dense to support a continuous crack front, tolerant and insensitive to notches. These properties depend matrix cracks extend through the fibres [36]; when this occurs, the fibres are no longer isolated from the matrix Corresponding author. Tel : +1-805-893-8248; fax: + 1-805-893-8486. crac E-mail address: flange @engineering. ucsb. edu(FF. Lange) I Present address: Advanced Engineering. SAAB. 461 80 Trollhattan Mullite is an ideal matrix material because of its high creep resistance, low modulus, low CTE and sluggish 2 Present address: USAFA/DFEM. US Air Force Academy, Colorado densification behaviour at temperatures below -1300C [21]. In the current study, mullite is used as the major matrix Published by Elsevier Science Ltd. PI:S1359-835X(02)00208-7
Fabrication and testing of all-oxide CFCC tubes M.G. Holmquista,b,1, T.C. Radsicka,2, O.H. Sudrec , F.F. Langea,* a Department of Materials, University of California at Santa Barbara, Santa Barbara, CA 93106, USA b Volvo Aero Corporation, 461 81 Trollha¨ttan, Sweden c Rockwell Scientific Company, 1049 Camino Dos Rios, Thousand Oaks, CA 91360, USA Received 15 August 2001; revised 15 September 2002; accepted 23 October 2002 Abstract A novel, low-cost processing method was used to manufacture all-oxide ceramic composite tubes. Fibre cloths were infiltrated with a preconsolidated slurry composed of a mixture of mullite and alumina powders. The pre-consolidated slurry was first formulated at a low volume fraction of powder to produce a short-range repulsive interparticle pair potential that allowed consolidation via pressure filtration, yet produced a consolidated body that could be fluidised by vibration. Efficient infiltration of fibre tows and a homogeneous microstructure were demonstrated. The infiltrated cloths, which could be frozen for latter use, were rolled to tubular shapes. After drying, the powder matrix was strengthened by cyclic infiltration and pyrolysis of an alumina precursor. When the tubes were pressurised, they delaminated at an average hoop stress of 47 MPa. Delamination initiated within the porous matrix where the outer most layer terminated. A failure analysis is presented using a strain energy release rate function for this specimen/crack configuration. Values of fracture energy inferred from the analysis were in close agreement with literature data for porous ceramics. Published by Elsevier Science Ltd. Keywords: E. Powder processing; E. Prepreg; B. Fracture; Oxide composites 1. Introduction All-oxide continuous fibre reinforced ceramic composites (CFCCs) have attracted growing interest for use in high temperature applications that include gas turbine engines [1–6], hot gas filters [7,8] and thermal protection systems [9]. These composites are stable in oxidising environments. However, applications are limited to 1000– 1200 8C, due to the currently available fibres [10]. More conventional, SiC based, CFCCs may find applications at higher temperatures ($1200 8C), although their life may be shorter because of their susceptibility to oxidative embrittlement at stresses that exceed the matrix strength [11,12]. Apart from fibre development, major efforts are focused on the issue of making oxide CFCCs damage tolerant and insensitive to notches. These properties depend on the fibres, which must be isolated from matrix cracks to take advantage of their high strength. Two different strategies for isolating fibres from matrix cracks can be distinguished; one is to develop a crack deflecting interface between the fibres and the surrounding matrix [13–18], a second is to produce a porous matrix [5,18,19,35]. The failure mechanisms of composites fabricated with a porous matrix have been studied and described in some detail [6,20, 21]. Upon loading the stress/strain behaviour is nearly linear, yet the matrix continuously microcracks before the onset of fibre failure [6,20]. The fibres are isolated from the matrix cracks because a continuous crack front cannot exist within a porous matrix. Namely, crack extension within the matrix occurs by the breaking of grain pairs at grain boundaries within the matrix. When the matrix becomes sufficiently dense to support a continuous crack front, matrix cracks extend through the fibres [36]; when this occurs, the fibres are no longer isolated from the matrix cracks. Mullite is an ideal matrix material because of its high creep resistance, low modulus, low CTE and sluggish densification behaviour at temperatures below ,1300 8C [21]. In the current study, mullite is used as the major matrix 1359-835X/03/$ - see front matter Published by Elsevier Science Ltd. PII: S1 35 9 -8 35 X( 02 )0 0 20 8 -7 Composites: Part A 34 (2003) 163–170 www.elsevier.com/locate/compositesa 1 Present address: Advanced Engineering, SAAB, 461 80 Trollha¨ttan, Sweden. 2 Present address: USAFA/DFEM, US Air Force Academy, Colorado Springs, Co.,USA. * Corresponding author. Tel.: þ1-805-893-8248; fax: þ1-805-893-8486. E-mail address: flange@engineering.ucsb.edu (F.F. Lange)
M.G. Holmquist et al./Composites: Part A 34(2003)163-170 phase, combined with a minor part alumina powder (0.30 the maximum sustainable hoop stress and energy release volume fraction) with fine particle size [23, 24, 31,38]. The rates for a propagating delamination crack. alumina can sufficiently densify at 1200C and help bind the mullite particles together without causing shrinkage Mullite will not densify at 1200C and thus ensure a stable 2. Experimental procedure matrix that will not shrink during the heat treatment, the mall volume fraction of alumina avoids braking the 2.1. materials continuity of the mullite network, which otherwise could lead to matrix densification during service [6, 21, 23). The All specimens prepared in this study were based on matrix is further strengthened by subsequent impregnation Nextel 610 alumina fibre cloth in an eight-harness satin and pyrolysis cycles with an alumina precursor, addin weave Sizing applied to the fabric by the manufacturer was more material to the bonds between the mullite particles removed through thermal treatment in air at 900C for a Pressure filtration has been used to produce this type of period of 2 h; otherwise the fabric was used in the as- composite [6, 19, 21]. In this method, a fibre preform(e.g. 3D received condition. The N610 fibres are composed of pure weave, stacked layers of cloth, etc. ) is mounted on a filter (99%)polycrystalline a-alumina and the tows in the within a die cavity. A powder slurry is introduced in th fabric contain approximately 400 filaments with diameters cavity and a pressure is applied to cause the particles to between 10 and 12 um[26, 27]. Although the N610 fibre stream through the preform and build up a consolidated shows very high room temperature strength, relative to layer within the fibre preform. The method requires N720, a mullite/alumina fibre, it is less stable and creep dispersed slurries with highly repulsive interparticle pair asistant at temperatures above 1000C 29. The potentials. This process works well for producing flat particle size and size distribution of mullite and alumina powders were selected to insure relatively high packing Vibrolntrusion is a new fabrication method for proces densities of the final filter-pressed slurry ing porous matrix composites with the advantage that it can powder was MU-107( Showa Denko KK, Tokyo Japan) be used to form complex shapes [22, 23,38) It requires a with a mean particle size(as determined by the manufac- slurry formulated with a short range repulsive potential turer) of 1 um and a particle size distribution of 0.5 between the particles that will create a weakly attractive 2.5 um. Its chemical composition is reported by the particle network [24, 25]. This interparticle pair potential manufacturer as 75.5%0 Al2O3 and 24% Sio2(by weight) with only trace amounts of TiO2, Fe,O3 and NayO. An o allows a powder compact that has been previously alumina powder (AKP-50, Sumitomo Chemicals, Tokyo, consolidated by pressure filtration, to be fluidised. The Japan)was used with a mean particle size of 0.2 um and a fluidised powder compact, called pre-consolidated slurry, is vibrated into a fibre cloth(or any other fibre preform). The particle size distribution of 0.1-0.3 um. Its chemistry essentially pure a-Al2O3(99.995%). An alumina precursor. shear rate thinning behaviour of the pre-consolidated slurry aluminium Ill sec-butoxide, C12H27O3Al (Gelest Inc. causes vibration to reduce its viscosity, which allows rapid Tullytown, PA, USA) was used to impregnate and thus d efficient intrusion of the fibre tows. Prepregs made in this manner can be frozen and stored. Once thawed they are by volume. strengthen the matrix. The precursor yielded 4% of alumina flexible and can be bent cut and formed much like an epoxy/fibre prepreg. Prepregs can be stacked and formed 2.2. Composite tube manufacturing into complex geometries like T-joints, doubly curve and tubes [38. Drying causes only minimal shrinkage due Aqueous slurries were prepared with a 20% volume to the high volume fraction of solid present in the pre- content of solids, comprising 70 vol% mullite and 30 vol% consolidated slurry alumina. Tetramethylammonium hydroxide (TMA-OH The purpose of this study was to demonstrate the was used to maintain the pH at about 11. Zeta-poter usefulness of the VibroIntrusion process in the fabrication measurements on mullite and alumina powders showed that of tubular structures from prepregs. Several potential pH 1l was sufficient to insure a dispersed slurry through the applications for ceramic matrix composites, including development of electrostatic repulsive interactions between high-temperature power piping and hot gas filters in the particles. two weight percentage(relative to the solids) power generation, and tubular combustors or curved of poly-ethylene oxide urethane silane(PEG-silane, Gelest combustor liners in gas turbine engines, are based on this Inc ) was added to induce a steric dispersing effect due to the simple geometry. A mechanical evaluation of tubular- adsorption of molecules by a reaction with -OH surface shaped specimens is also conveniently performed using a sites on both powders [25]. The slurry was contained in a simple pressure test, giving insights on the failure plastic jar that was placed on a mechanical roller for mechanisms and their relation to the fuid containment approximately 12 h and after mixing, tetramethylam- aptitudes of these materials. The failure behaviour monium nitrate(TMA-N)salt was added to make a 0.25M of pressurised tubes was analysed in terms of both solution. The addition of TMa counterions produce
phase, combined with a minor part alumina powder (0.30 volume fraction) with fine particle size [23,24,31,38]. The alumina can sufficiently densify at 1200 8C and help bind the mullite particles together without causing shrinkage. Mullite will not densify at 1200 8C and thus ensure a stable matrix that will not shrink during the heat treatment. The small volume fraction of alumina avoids braking the continuity of the mullite network, which otherwise could lead to matrix densification during service [6,21,23]. The matrix is further strengthened by subsequent impregnation and pyrolysis cycles with an alumina precursor, adding more material to the bonds between the mullite particles. Pressure filtration has been used to produce this type of composite [6,19,21]. In this method, a fibre preform (e.g. 3D weave, stacked layers of cloth, etc.) is mounted on a filter within a die cavity. A powder slurry is introduced in the cavity and a pressure is applied to cause the particles to stream through the preform and build up a consolidated layer within the fibre preform. The method requires dispersed slurries with highly repulsive interparticle pair potentials. This process works well for producing flat panels. VibroIntrusion is a new fabrication method for processing porous matrix composites with the advantage that it can be used to form complex shapes [22,23,38]. It requires a slurry formulated with a short range repulsive potential between the particles that will create a weakly attractive particle network [24,25]. This interparticle pair potential allows a powder compact that has been previously consolidated by pressure filtration, to be fluidised. The fluidised powder compact, called pre-consolidated slurry, is vibrated into a fibre cloth (or any other fibre preform). The shear rate thinning behaviour of the pre-consolidated slurry causes vibration to reduce its viscosity, which allows rapid and efficient intrusion of the fibre tows. Prepregs made in this manner can be frozen and stored. Once thawed they are flexible and can be bent, cut and formed much like an epoxy/fibre prepreg. Prepregs can be stacked and formed into complex geometries like T-joints, doubly curved shapes and tubes [38]. Drying causes only minimal shrinkage due to the high volume fraction of solid present in the preconsolidated slurry. The purpose of this study was to demonstrate the usefulness of the VibroIntrusion process in the fabrication of tubular structures from prepregs. Several potential applications for ceramic matrix composites, including high-temperature power piping and hot gas filters in power generation, and tubular combustors or curved combustor liners in gas turbine engines, are based on this simple geometry. A mechanical evaluation of tubularshaped specimens is also conveniently performed using a simple pressure test, giving insights on the failure mechanisms and their relation to the fluid containment aptitudes of these materials. The failure behaviour of pressurised tubes was analysed in terms of both the maximum sustainable hoop stress and energy release rates for a propagating delamination crack. 2. Experimental procedure 2.1. Materials All specimens prepared in this study were based on Nextel 610e alumina fibre cloth in an eight-harness satin weave. Sizing applied to the fabric by the manufacturer was removed through thermal treatment in air at 900 8C for a period of 2 h; otherwise the fabric was used in the asreceived condition. The N610 fibres are composed of pure (.99%) polycrystalline a-alumina and the tows in the fabric contain approximately 400 filaments with diameters between 10 and 12 mm [26,27]. Although the N610 fibre shows very high room temperature strength, relative to N720, a mullite/alumina fibre, it is less stable and creep resistant at temperatures above ,1000 8C [27–29]. The particle size and size distribution of mullite and alumina powders were selected to insure relatively high packing densities of the final filter-pressed slurry. The mullite powder was MU-107 (Showa Denko KK, Tokyo Japan) with a mean particle size (as determined by the manufacturer) of 1 mm and a particle size distribution of 0.5– 2.5 mm. Its chemical composition is reported by the manufacturer as 75.5% Al2O3 and 24% SiO2 (by weight) with only trace amounts of TiO2, Fe2O3 and Na2O. An aalumina powder (AKP-50, Sumitomo Chemicals, Tokyo, Japan) was used with a mean particle size of 0.2 mm and a particle size distribution of 0.1–0.3 mm. Its chemistry is essentially pure a-Al2O3 (99.995%). An alumina precursor, aluminium III sec-butoxide, C12H27O3Al (Gelest Inc., Tullytown, PA, USA) was used to impregnate and thus strengthen the matrix. The precursor yielded 4% of alumina by volume. 2.2. Composite tube manufacturing Aqueous slurries were prepared with a 20% volume content of solids, comprising 70 vol% mullite and 30 vol% alumina. Tetramethylammonium hydroxide (TMA-OH) was used to maintain the pH at about 11. Zeta-potential measurements on mullite and alumina powders showed that pH 11 was sufficient to insure a dispersed slurry through the development of electrostatic repulsive interactions between the particles. Two weight percentage (relative to the solids) of poly-ethylene oxide urethane silane (PEG-silane, Gelest Inc.) was added to induce a steric dispersing effect due to the adsorption of molecules by a reaction with –OH surface sites on both powders [25]. The slurry was contained in a plastic jar that was placed on a mechanical roller for approximately 12 h and after mixing, tetramethylammonium nitrate (TMA-N) salt was added to make a 0.25M solution. The addition of TMAþ counterions produce 164 M.G. Holmquist et al. / Composites: Part A 34 (2003) 163–170
M.G. Holmquist et al./Composites: Part A 34(2003)163-170 short-range repulsive potential, and thus a weakly attractive tubes were subsequently impregnated with the alumina particle network [25]. Slurries were returned to the precursor solution under vacuum. The impregnation step mechanical roller for another 12 h. Pressure filtration was performed in a dry nitrogen atmosphere to prevent (4 MPa)was used to consolidate the slurry; the consolidated premature gelation of the precursor due to atmospheri body was subsequently fluidised again via vibration after it water vapour. The composites were left in the precursor was placed in a plastic bag to prevent drying. The solution for 2 h at atmospheric pressure and transferred into consolidation pressure of 4 MPa was below the critical ammoniated water(pH 10) to gel the precursor throughout pressure where a large number of particles are pushed into the body and prevent it from re-distributing to the surface ontact. which would obviate fluidisation after consolida- during the evaporation of the solvent [32, 33]. After 4 h the tion [25, 30). Using the weight difference method, the composites were removed, dried and heated to 900"C in volume fraction of solids within the water saturated order to pyrolyze the precursor. This was repeated three consolidated bodies was determined to 56.4+ 0.4 times and following the last cycle, the composites were Cloth, cut into-102x 76 mm? rectangles with the two given a final sintering treatment at 1200"C for 2h, which outer most tows along all four edges removed, were put in served to crystallise the precursor to the corundum (a) separate plastic bags; an excess of the pre-consolidated structure. In this manner, the strength of the connection slurry was dispensed on both sides of each cloth. The plastic between mullite particles could be increased without any bag served to prevent the slurry from drying and from shrinkage of the mullite network adhering to tools used in subsequent manufacture. Assisted by vibration, the slurry was manually rolled across the 23. Characterisation surface of the fibre cloth with a piece of aluminium rod until the cloth was fully infiltrated. Since the fluidised body The porosity was measured using the Archimedes exhibits shear-rate thinning, the vibration reduces the method in water. Fibre volume fraction was then calculated viscosity and allows rapid intrusion of the particles into from knowledge of the fibre cloth volume and total tube the fibre cloth [22, 24). Prepregs were frozen to aid removal volume. Microstructures were studied by optical from the plastic bag, i.e. they popped apart due to the microscopy and scanning electron microscopy of cross differential thermal expansion sections of as processed and tested tubes. Damage evolution To produce the composite tube, the prepreg was clamped during testing was monitored by visual observations of the into a three-part collapsible mandrel(diameter 12.7 mm) tube surfaces. In one case, a thin red lacquer coating was made of stainless steel using a plastic extension tab and applied on the outer surface of a tube to ascertain the failure placed atop a vibrating aluminium table at room tempera ture. as the matrix material thawed and became fluid once ain,the prepreg section was slowly wound around the 2. 4. Testing procedure mandrel as shown in Fig. 1. Gentle pressure was applied to Brass testing fixtures were prepared to enable pressure the prepreg in order to squeeze out a small portion of the testing of the CFCC tubes. Tubes were joined to the test fluid powder matrix and create a wake of matrix just in fror of the composite tube that was forming around the mandrel. fixtures using a rapid curing epoxy resin. The fixture This wake of matrix slurry served to fill in air gaps as the included a central rod that prevented axial loading of the tubes and premature failure at the epoxy joint. Internal prepreg layer wrapped on itself, thus preventing large-scale porosity in the final composite. Once the cloth was fully pressure was applied to the composite tube from a nitrogen source(2500 bar). The gas pressure was controlled by wound around the mandrel, the strip of protective plastic was tightly wrapped around the prepreg tube and placed in a The absolute pressure near the inlet of the tube and drying oven at 70C. The mandrel was collapsed and volumetric flow rate at the exhaust port were monitored emoved and the tubes were heated to 900"C for 2 h to The permeability, K, of the composite material was partially sinter the alumina particles to the mullite particles calculated at a low pressure difference(138 kPa)using the and give the porous matrix some strength. The composite following equation, which is applicable to laminar flow of compressible fluids through porous media [ 39] =/s (P1-P) (1) where Pi is the internal pressure, Po, the surrounding Fig 1 Schematic illustration showing how the prepreg is wrapped around a pressure(atmospheric), t, the tube wall thickness, A, the collapsible mandrel. viscosity of the gas, S, the surface area of the tube wall and
short-range repulsive potential, and thus a weakly attractive particle network [25]. Slurries were returned to the mechanical roller for another 12 h. Pressure filtration (4 MPa) was used to consolidate the slurry; the consolidated body was subsequently fluidised again via vibration after it was placed in a plastic bag to prevent drying. The consolidation pressure of 4 MPa was below the critical pressure where a large number of particles are pushed into contact, which would obviate fluidisation after consolidation [25,30]. Using the weight difference method, the volume fraction of solids within the water saturated consolidated bodies was determined to 56.4 ^ 0.4%. Cloth, cut into ,102 £ 76 mm2 rectangles with the two outer most tows along all four edges removed, were put in separate plastic bags; an excess of the pre-consolidated slurry was dispensed on both sides of each cloth. The plastic bag served to prevent the slurry from drying and from adhering to tools used in subsequent manufacture. Assisted by vibration, the slurry was manually rolled across the surface of the fibre cloth with a piece of aluminium rod until the cloth was fully infiltrated. Since the fluidised body exhibits shear-rate thinning, the vibration reduces the viscosity and allows rapid intrusion of the particles into the fibre cloth [22,24]. Prepregs were frozen to aid removal from the plastic bag, i.e. they popped apart due to the differential thermal expansion. To produce the composite tube, the prepreg was clamped into a three-part collapsible mandrel (diameter 12.7 mm) made of stainless steel using a plastic extension tab and placed atop a vibrating aluminium table at room temperature. As the matrix material thawed and became fluid once again, the prepreg section was slowly wound around the mandrel as shown in Fig. 1. Gentle pressure was applied to the prepreg in order to squeeze out a small portion of the fluid powder matrix and create a wake of matrix just in front of the composite tube that was forming around the mandrel. This wake of matrix slurry served to fill in air gaps as the prepreg layer wrapped on itself, thus preventing large-scale porosity in the final composite. Once the cloth was fully wound around the mandrel, the strip of protective plastic was tightly wrapped around the prepreg tube and placed in a drying oven at 70 8C. The mandrel was collapsed and removed and the tubes were heated to 900 8C for 2 h to partially sinter the alumina particles to the mullite particles and give the porous matrix some strength. The composite tubes were subsequently impregnated with the alumina precursor solution under vacuum. The impregnation step was performed in a dry nitrogen atmosphere to prevent premature gelation of the precursor due to atmospheric water vapour. The composites were left in the precursor solution for 2 h at atmospheric pressure and transferred into ammoniated water (pH 10) to gel the precursor throughout the body and prevent it from re-distributing to the surface during the evaporation of the solvent [32,33]. After 4 h the composites were removed, dried and heated to 900 8C in order to pyrolyze the precursor. This was repeated three times and following the last cycle, the composites were given a final sintering treatment at 1200 8C for 2 h, which served to crystallise the precursor to the corundum (a) structure. In this manner, the strength of the connection between mullite particles could be increased without any shrinkage of the mullite network. 2.3. Characterisation The porosity was measured using the Archimedes method in water. Fibre volume fraction was then calculated from knowledge of the fibre cloth volume and total tube volume. Microstructures were studied by optical microscopy and scanning electron microscopy of cross sections of as processed and tested tubes. Damage evolution during testing was monitored by visual observations of the tube surfaces. In one case, a thin red lacquer coating was applied on the outer surface of a tube to ascertain the failure location. 2.4. Testing procedure Brass testing fixtures were prepared to enable pressure testing of the CFCC tubes. Tubes were joined to the test fixtures using a rapid curing epoxy resin. The fixture included a central rod that prevented axial loading of the tubes and premature failure at the epoxy joint. Internal pressure was applied to the composite tube from a nitrogen source (2500 bar). The gas pressure was controlled by manually opening or closing the regulator on the gas supply. The absolute pressure near the inlet of the tube and volumetric flow rate at the exhaust port were monitored. The permeability, K; of the composite material was calculated at a low pressure difference (138 kPa) using the following equation, which is applicable to laminar flow of compressible fluids through porous media [39] Q ¼ KS mt ðPi 2 PoÞ ð1Þ where Pi is the internal pressure, Po; the surrounding pressure (atmospheric), t; the tube wall thickness, m; the viscosity of the gas, S; the surface area of the tube wall and Fig. 1. Schematic illustration showing how the prepreg is wrapped around a collapsible mandrel. M.G. Holmquist et al. / Composites: Part A 34 (2003) 163–170 165
M.G. Holmquist et al./ Composites: Part A 34(2003)163-170 Table 1 Summary of composite tubes Tube Thickness Fibre volume fraction, V Composite porosity, P Matrix porosity, Pm Pressure at failure Tangential stress at failure 4 32.8 1234 34.2 354 53.4 45.7 789 390 >81.3 ued with epoxy along the line formed by the edge of the outer ply. Q is given by contained three layers of cloth. All thickness measurements were made in the two-layer region of the tube Based on weight measurements the mullite-to-alumina (2) volume ratio decreased from 70/30 to 64.6/35.4 after three cycles of precursor infiltration and pyrolysis. Average fibre where @o is the volumetric flow rate of the gas measured on volume fraction in the final composite tubes was determined the atmospheric pressure side of the test set-up and P is the to be 37% and the average matrix porosity was 44%.As mean of internal and surrounding gas pressure. A viscosity shown in Fig. 2, the fibre bundles appeared to be well of 17.5 x 10-N s/m was assumed for nitrogen at room infiltrated by the matrix slurry. The matrix within the fibre temperature [37] bundles has a relatively uniform distribution of pores, Pressure was increased manually at approximately 25 I um in diameter. This observation is consistent with 50 kPa/s. The flow meter was restricted to a maximum flow reports from others made on similar materials [18, 21]. rate of 1.67 Is(100 I/min), but by using a by-pass valve Occasional large pores(-50X -100 um")were observed twice as high flow rates could be evaluated. The pressure at between the cloth plies. These pores are certainly related to the specimen inlet and the flow rate at the outlet were air bubbles trapped between the plies during the wrapping monitored as the pressure at the tank regulator procedure. Some cracklike faws, perpendicular to the fibres increased. Several pressure cycles were performed during and with regular spacing, were observed in matrix rich the test, in particular when a pressure drop occurred at the regions. These cracks were more than likely due to the pecimen inlet Such a pressure drop indicated an increase in constraint the fibres impose on the matrix shrinkage during flow rate in the line, the pressure at the tank outlet being drying and heating [6, 23]. Because the openings of these fixed. This pressure drop was used as an indication of a cracks are relatively large it was not possible to eliminate damage event in the composite tube and it was usually also them in the subsequent impregnation/pyrolysis steps. The accompanied by an audible sound. When this occurred, the transition from the section containing three layers of cloth to pressure was decreased and the tube was visually inspected two layers was generally smooth, although some specimens to determine type and location of failure. The pressure was had a more abrupt step-like transition then increased again to study the progression of damage Permeability of the as-processed tubes was measured to Nine tubes were tested in this manner be about 5.1 x 10-6m at a differential pressure of 138 kPa(20 psi), which is in reasonable agreement with 3. Result The tubes were composed of two layers of cloth and were produced by winding an initial single strip of prepreg cloth around the mandrel two times. A summary of the tubes evaluated is given in Table 1. The length of the tubes fabricated by the described method varied between 57 and 68 mm and their inner diameter averaged 12. 8 mm thickness of all the tubes was 0.49+ 0.02 mm. The sm 0.5mm section of the tube(approximately one-fifth of its circ ference or 8 mm) where ends of the cloth overlapped Fig. 2. Microstructure of as processed tube (cross-section)
Q is given by Q ¼ QoPo P ð2Þ where Qo is the volumetric flow rate of the gas measured on the atmospheric pressure side of the test set-up and P is the mean of internal and surrounding gas pressure. A viscosity of 17.5 £ 1026 N s/m2 was assumed for nitrogen at room temperature [37]. Pressure was increased manually at approximately 25– 50 kPa/s. The flow meter was restricted to a maximum flow rate of 1.67 l/s (100 l/min), but by using a by-pass valve twice as high flow rates could be evaluated. The pressure at the specimen inlet and the flow rate at the outlet were monitored as the pressure at the tank regulator was increased. Several pressure cycles were performed during the test, in particular when a pressure drop occurred at the specimen inlet. Such a pressure drop indicated an increase in flow rate in the line, the pressure at the tank outlet being fixed. This pressure drop was used as an indication of a damage event in the composite tube and it was usually also accompanied by an audible sound. When this occurred, the pressure was decreased and the tube was visually inspected to determine type and location of failure. The pressure was then increased again to study the progression of damage. Nine tubes were tested in this manner. 3. Results The tubes were composed of two layers of cloth and were produced by winding an initial single strip of prepreg cloth around the mandrel two times. A summary of the tubes evaluated is given in Table 1. The length of the tubes fabricated by the described method varied between 57 and 68 mm and their inner diameter averaged 12.8 mm. The thickness of all the tubes was 0.49 ^ 0.02 mm. The small section of the tube (approximately one-fifth of its circumference or 8 mm) where ends of the cloth overlapped contained three layers of cloth. All thickness measurements were made in the two-layer region of the tube. Based on weight measurements, the mullite-to-alumina volume ratio decreased from 70/30 to 64.6/35.4 after three cycles of precursor infiltration and pyrolysis. Average fibre volume fraction in the final composite tubes was determined to be 37% and the average matrix porosity was 44%. As shown in Fig. 2, the fibre bundles appeared to be well infiltrated by the matrix slurry. The matrix within the fibre bundles has a relatively uniform distribution of pores, ,1 mm in diameter. This observation is consistent with reports from others made on similar materials [18,21]. Occasional large pores (,50 £ , 100 mm2 ) were observed between the cloth plies. These pores are certainly related to air bubbles trapped between the plies during the wrapping procedure. Some cracklike flaws, perpendicular to the fibres and with regular spacing, were observed in matrix rich regions. These cracks were more than likely due to the constraint the fibres impose on the matrix shrinkage during drying and heating [6,23]. Because the openings of these cracks are relatively large it was not possible to eliminate them in the subsequent impregnation/pyrolysis steps. The transition from the section containing three layers of cloth to two layers was generally smooth, although some specimens had a more abrupt step-like transition. Permeability of the as-processed tubes was measured to be about 5.1 £ 10216 m2 at a differential pressure of 138 kPa (20 psi), which is in reasonable agreement with Table 1 Summary of composite tubes Tube Thickness (mm) Fibre volume fraction, Vf (%) Composite porosity, pc (%) Matrix porosity, pm (%) Pressure at failure (MPa) Tangential stress at failure (MPa) 1 0.48 38.9 28.0 45.8 2.48 32.8 2 0.48 34.2 28.5 43.3 3.38 44.7 3 0.50 35.4 27.5 42.5 4.21 53.4 4 0.52 35.0 29.2 44.9 3.81 45.7 5 0.48 36.7 27.0 42.6 4.21 55.6 6 0.48 36.8 27.4 43.3 3.93 52.0 7 0.51 37.8 27.6 44.4 3.96 49.4 8 0.52 36.0 28.8 45.0 3.45 42.1 9 0.49 39.0 28.0 45.6 .6.27 .81.3 Tube 9 was glued with epoxy along the line formed by the edge of the outer ply. Fig. 2. Microstructure of as processed tube (cross-section). 166 M.G. Holmquist et al. / Composites: Part A 34 (2003) 163–170
M.G. Holmquist et al./Composites: Part A 34(2003)163-170 the specimen inlet occurred and was associated with an audible sound. This was interpreted as the onset of damage A hysteresis could then be observed during unloading. For all specimens, the failure initiated along the line formed by Final fracture he edge of the outer ply as shown on the lacquer-coated specimen in Fig. 4. Delamination of the wound fibre cloth was the main failure mode. The tube could be pressurised Progress of again, although fow rates increased relative to the first two cycles for a given applied pressure, suggesting the presence of larger flow short-circuits through the composite wall. The -o- Cycle 5 tube still held pressures above that of the previous maximum, indicating that some mechanical integrity was still retained. The crack front on the specimen surface Flow(W/s) gradually propagated towards the ends of the tube causing new pressure drops(cycle 4 in Fig. 3). At still higher Fig. 3. Internal absolute pressure in the tube versus flow rate through the tube surface. The tube failed by delamination(end of cycle 3 pressures(about 4.5 MPa), the crack front on oated specimen was observed to deviate at an angle, measurements on other porous ceramics [39]. A typical breaking fibres parallel to the tube axis in a combination of curve of flow rate as a function of pressure is shown in Fig 3. shear and tensile loading (location 2 in Fig. 4). The new In the first two cycles, the pressure was increased and crack trajectory possibly resulted from the constraint on the released without any sign of hysteresis. Upon the third cycle deformation imposed by the glued ends of the tubes. The at a maximum pressure of 3.5 MPa, a pressure drop at test was terminated when the pressure could no longer be increased and the gas tank was emptying too rapidly. In some cases, possibly depending on the rate of the pressur increase, the pressure caused the tube to blow apart as shown in Fig. 5. Cross-sections of tested tubes confirmed that delamina- tion was the major mode of failure, the crack propagating inwards with a spiral trajectory between the cloth layers (Fig. 6). Upon completion of the test, the crack propagated beyond the location of the inner termination of the wrap, making the tube wall only one layer thick. Another delamination crack was observed at the termination of 2 Fig 4. The tubes glued with epoxy resin to brass fixtures. The brass xtures were connected to each other by a central rod, running inside the tube. Failure sequence of the tubes were: (1)delamination starting at the ermination of the wound prepreg followed by (2) deviation of the crack front resulting in failure of fibres parallel to tube axis. Fig. 5. Tube blown up from
measurements on other porous ceramics [39]. A typical curve of flow rate as a function of pressure is shown in Fig. 3. In the first two cycles, the pressure was increased and released without any sign of hysteresis. Upon the third cycle at a maximum pressure of 3.5 MPa, a pressure drop at the specimen inlet occurred and was associated with an audible sound. This was interpreted as the onset of damage. A hysteresis could then be observed during unloading. For all specimens, the failure initiated along the line formed by the edge of the outer ply as shown on the lacquer-coated specimen in Fig. 4. Delamination of the wound fibre cloth was the main failure mode. The tube could be pressurised again, although flow rates increased relative to the first two cycles for a given applied pressure, suggesting the presence of larger flow short-circuits through the composite wall. The tube still held pressures above that of the previous maximum, indicating that some mechanical integrity was still retained. The crack front on the specimen surface gradually propagated towards the ends of the tube causing new pressure drops (cycle 4 in Fig. 3). At still higher pressures (about 4.5 MPa), the crack front on the lacquercoated specimen was observed to deviate at an angle, breaking fibres parallel to the tube axis in a combination of shear and tensile loading (location 2 in Fig. 4). The new crack trajectory possibly resulted from the constraint on the deformation imposed by the glued ends of the tubes. The test was terminated when the pressure could no longer be increased and the gas tank was emptying too rapidly. In some cases, possibly depending on the rate of the pressure increase, the pressure caused the tube to blow apart as shown in Fig. 5. Cross-sections of tested tubes confirmed that delamination was the major mode of failure, the crack propagating inwards with a spiral trajectory between the cloth layers (Fig. 6). Upon completion of the test, the crack propagated beyond the location of the inner termination of the wrap, making the tube wall only one layer thick. Another delamination crack was observed at the termination of Fig. 5. Tube blown up from escaping gas. Fig. 3. Internal absolute pressure in the tube versus flow rate through the tube surface. The tube failed by delamination (end of cycle 3). Fig. 4. The tubes were glued with epoxy resin to brass fixtures. The brass fixtures were connected to each other by a central rod, running inside the tube. Failure sequence of the tubes were: (1) delamination starting at the termination of the wound prepreg followed by (2) deviation of the crack front resulting in failure of fibres parallel to tube axis. M.G. Holmquist et al. / Composites: Part A 34 (2003) 163–170 167