SMART COMPOSITES APPLIED TO CONTINUOUS IN-SERVICE DAMAGE MONITORING OF GLASS/POLYMER INDUSTRIAL PARTS

Transcription

1 SMART COMPOSITES APPLIED TO CONTINUOUS IN-SERVICE DAMAGE MONITORING OF GLASS/POLYMER INDUSTRIAL PARTS Cyril TAMIATTO 1, Patricia KRAWCZAK 1, José PABIOT 1, and Fabrice LAURENT 2 1 ECOLE des MINES de DOUAI, Departent of Polyers & Coposites Technology, 941 rue Charles Bourseul, BP 838, F-5958 DOUAI, FRANCE e-ail: plastics@ens-douai.fr Fax: STRATIFORME Industries, 26 Route Nationale, BERSEE, FRANCE SUMMARY: This paper ais to develop a onitoring ethod aking it possible to control the daage of glass/resin aterials in-service in order to prevent any catastrophic failure (preventive aintenance). Carbon fibre resistive sensors are introduced directly inside the coposite aterials during anufacturing. The increase of their electric resistance, caused by the elongation and then the rupture of carbon filaents during loading of the coposite structure, is onitored. Finally the electric response of the sensors is odelled in order to develop a real tie daage onitoring syste. Statistic functions such as Weibull or sigoid functions are ipleented to describe the successive failure phenoenon of the carbon filaents. The concept is validated on a coposite pipe subjected to biaxial loading. KEYWORDS: sart aterials, daage onitoring, carbon fibre, glass fibre reinforced plastics (GFRP), dynaic fatigue, failure, coposite structures, tubes BACKGROUND Health onitoring of coposite aterials has been carried out until now by non destructive testing ethods, such as X ray, acoustic eissions, ultrasonic inspection... Nevertheless, these techniques show several drawbacks such as high onetary investents, difficulty to use the testing appliances on site in good conditions and necessity of long inactivity periods for the equipent to be controlled. For the last ten years research progras have been carried out in order to solve these probles by a direct location of soe sensors inside coposite aterials during their anufacture (ost of the tie optic fibres and piezoelectric sensors) [1-4] or by using their conductive reinforceent fibres as resistive sensors [5-7]. These new aterials which are able to continuously control soe of their properties are called sart coposites. In the present work, a different approach has been chosen. The ai is to develop a sart glass/resin coposite being able to follow the in-service daage of industrial parts for a purpose of preventive aintenance. The daage sensor used is a carbon tow. The increase of its electric resistance, caused by the elongation and then the rupture of the carbon fibres during loading of the coposite structure, is onitored.

2 A basic study [8,9] has been first carried out on test coupons in order to check the feasibility of the daage onitoring ethod, to calibrate the sensors and to investigate the possible paraeters of influence: physical characteristics of the tows (noinal elongation at break, nuber of filaents), structure of the host coposite aterial, loading ode (onotonic tension, bending or shear, repeated progressive tension, dynaic bending fatigue), environental conditions (teperature and oisture). The different ways to process the sensor responses have also been discussed. The results obtained have shown that carbon tows can be used as resistive daage integrated sensors providing their characteristics are adapted to the critical strains of the structure to be onitored. It is iportant to point out that a calibration stage is necessary in order to effectively use this technology and that the application field is restricted to non conductive coposites. Aong the potential paraeters of influence investigated, the carbon fibre elongation at break and linear density appear to be of prie iportance. The oisture content of the environent has no effect on the sensor response and the sall effect of the teperature can easily be taken into account through a correction factor. Regarding the processing of the sensor responses, it is possible to use the electric resistance increase caused either by the fracture of the conductive carbon fibres which provides partially non reversible inforation, or by the fibre elongation which provides a reversible response with a uch lower aplitude. Fro a practical point of view, prior to the integration of the sensor in a coposite structure, it is necessary to deterine both the stress ap (location and direction) and the governing laws of the aterial. The sensor integration is then possible after having chosen the characteristics of the carbon tows in order to obtain an easily easurable and reproducible response for the level of stress considered as being critical for the part to be onitored. The ai of the present paper is now to validate this onitoring concept on industrial parts under coplex loading and to develop a atheatical odel for the sensor response. In a first part, different sensors are calibrated in onotonic tension. Then the daage onitoring ethod is applied on coposite pipes under increasing and dynaic internal pressure with end effects (biaxial loading). Finally statistic functions such as Weibull or sigoid functions are ipleented to describe the electric response of the resistive carbon fibre sensors. EXPERIMENTAL MEANS Standard tensile test speciens with end tabs were ade by contact oulding of a E-glass (/9 ) non woven fabric and a vinylester (Derakane ) atrix with a fibre content of 46 % by weight. Four types of carbon tows are used as resistive sensors; their references are given in table 1. The industrial coposite parts chosen here are E-glass/epoxy filaent wound pipes of 1 diaeter and ± 55 winding angle, anufactured by Wavin Repox. In that case, the sensors are located on the pipe outer surface and ebedded in an epoxy resin: the final practice would be of course to integrate the carbon fibre sensor inside the pipe wall during anufacturing, in the sae way as for test coupons. Tensile tests were perfored on a ZWICK 1474 achine, equipped with an extensoeter, at a cross-head rate of 1 /in. The easureent of the electric resistance of the carbon tows occurs by eans of either a current divider bridge [8] or a icro oh eter (HP 3442A) if a higher accuracy is required. Acoustic eissions have also been onitored by eans of a

3 LOCAN AT acquisition unit fro PAC. Tests on pipes under increasing or dynaic internal water pressure with end effects were carried out on an hydraulic testing unit at a rate of 2 bar/in for onotonic tests and at a frequency of 25 cycles/in for dynaic tests. Table 1. References and properties of the carbon tows Fibre reference Elongation at break (%) Nuber of filaents TORAYCA T7 SC (HR) 2,1 12 Carbone HR-3K 1,5 à 1,7 3 TORAYCA M 4 (HM),6 1, 3, 6, 12 DIALEAD K13A1 (THM),36 1 EXPERIMENTAL RESULTS Calibration of the sensors in onotonic tension on test coupons Figure 1 shows the results obtained during a onotonic tensile test (parallel to the carbon fibre direction) on test speciens containing carbon tows of 2,1%, 1.5/1.7%,,6% and,36% noinal ultiate elongation respectively. The relative electric resistance versus strain curves for the four different carbon tows and the ean stress-strain curve have been plotted. The electric resistance of the sensor increases with specien elongation: at first slightly only and alost linearly due to the carbon fibre elongation, which practically is not visible for certain carbon tow types; then ore rapidly after a certain strain level, which varies according to the carbon tow type. The change of slope of the electric resistance vs. strain curves is caused by the first fractures of the conductive carbon fibres and constitutes a specific signal of the sensor which can be easily detected. It appears logically that, the higher the ultiate elongation of the carbon tow, the higher the strain for which the change of slope of the electric resistance curve appears. The end of linearity of the stress-strain curve (knee point) corresponds to the end of the elastic behaviour and can be associated to the beginning of the global non recoverable daage of the aterial. 25,5 CONTRAINTE MPa Stress (MPa) ,36 % Contrainte stress,6 % 1,5/1,7 % 2,1%,5 1 1,5 2 2,5 DEFORMATION % Strain (%),4,3,2,1 Relative RESISTANCE electric RELATIVE resistance (R-Ro)/Ro (R-R)/R Fig.1: Electric responses for different carbon fibre tows (,36%,,6%, 1,5/1,7% and 2,1% noinal elongation at break) (tensile test on test coupons) For carbon tows of,36 % and,6 % ultiate elongations, the change of slope of the electric resistance vs. strain curves occurs before the end of the elastic behaviour, and thus, before the first critical daage of the host coposite: that eans that both carbon tows are sensitive enough and suit well to act as preventive daage sensors for this particular host coposite

4 structure. Application to pipes under internal pressure R/ε curves corresponding to the different carbon fibre tows for a given loading ode are available fro the previous tests on standard speciens. Now in order to instruent a part in such a way that a significant electric response of the sensor occurs at the onset of the first critical daage of the host coposite structure, it is necessary to know as a preliinary the critical strain level of the part to be onitored. In our case, the first daage of the pipes subjected to an internal pressure with end effects occurs at,16% axial strain and,54% hoop strain (knee-point on stress-strain curves). As a consequence the best choice is to instruent the pipe with a carbon tow of,36% noinal ultiate elongation located in the hoop direction (its electric response due to filaents fractures occurs at,45% elongation on the basis of tensile tests on test coupons). The short-ter stress-axial & hoop strains curves are reported on figure 2, as well as the variation of the electric resistance of the hoop located carbon tow. The sudden increase of the electric resistance due to carbon fibres fracture occurs at,5% elongation, that is to say at a pressure level just preceding the aterial first daage. Indeed the sensor effectively acts as a preventive daage sensor for the instruented coposite part. Moreover it reacts well before weeping (loss of watertightness of the pipe wall), which occurs at 1,5% hoop strain: This allows the user to intervene before the final failure of the part. The reproducibility of the results, checked on three pipes, appears to be good (scattering less than 5%). Hoop stress (MPa) Axial strain First daage Hoop strain Electric resistance,1,2,3,4,5,6,7,8 Strain (%) Fig.2: Coposite pipe under increasing internal pressure - Response of a carbon fibre sensor (,36% noinal elongation at break, 1K linear density) located in the hoop direction Then, in order to siulate an intensive use of the pipe at long ter under a loading level near the critical daage point of the structure, dynaic fatigue tests at constant aplitude of stress are carried out, the upper pressure liit (115 MPa) being 2% less than the knee-point of the stress-strain curves recorded under increasing internal pressure. Figure 3 shows, as a function of the cycle nuber, the axiu and iniu electric resistances corresponding respectively to the sensor response at the upper and lower stress levels during a cycle. It appears that the sensor reacts after 82 cycles only, while the pipe weeping occurs after 14 cycles. At that oent, the pipe shows 24% axial odulus loss, 5% hoop odulus loss, and the hoop strain reaches,5% (value siilar to that easured during burst tests and close to that deterined on test saples (,45%)). Thus cycling does not affect the strain threshold for which the electric resistance increase of the sensor is noticed. An increase of the iniu electric resistance of,45 Ω is also noted after 14 cycles: Nevertheless the Electric resistance R-Ro (Ω)

5 exploitation of this signal, even if possible, would be difficult in practice because of its low aplitude. 6 6 Stiffness (GPa) Axial stiffness Hoop stiffness Max. electr. resistance failure Min. electr. resistance Nuber of cycles Fig.3: Coposite pipe under dynaic internal pressure - Response of a carbon fibre sensor (,36% noinal elongation at break) located in the hoop direction Dynaic fatigue tests on pipes have shown that the use of the carbon fibre sensor akes it possible to onitor the daage state of the structure and to produce a preventing guard signal when such an industrial structure is subitted to a cyclic loading siulating real operating conditions over a short period of tie Electric resistance R (Ω) MODELLING OF THE SENSORS ELECTRIC RESPONSE The ai here is to define the necessary atheatical treatent allowing to link the electric resistance increase of the carbon tow to its strain, and thus, to change an electric signal into echanical data related to the aterial. The design of a calculation code would then be possible, in order to follow in real tie the daage of an instruented part. Modelling of the electric resistance increase due to carbon fibre elongation The elongation of the carbon fibres induces an increase of their length, a decrease of their cross-section and thus an increase of their electric resistance, which can be calculated assuing that the carbon tow and the coposite aterial experience the sae longitudinal strain ε. Taking into account the Poisson s ratio υ effect of the carbon fibres, the electric resistance increase R e due to fibre elongation is then directly related to the applied strain ε and the initial electric resistance R of the elongated carbon fibre portion (efficient length) according to equation (1). Nevertheless due to the low aplitude of these electric resistance differences, it is very difficult to ipleent a onitoring ethod based on fibre elongation on industrial parts. R e = R ε (1+2 υ) (1) Modelling of the electric resistance increase due to carbon fibre fracture It is possible to calculate the nuber of broken carbon fibres fro the electric resistance increase of the carbon tow, considering that all the carbon fibres are arranged in parallel in the tow and have the sae electric resistance [1]. The initial electric resistance R of the carbon tow is then given by the ratio of the electric resistance R F of one single fibre to the total

6 nuber of carbon fibres N (equation (2)). If n carbon fibres are broken, the increase of electric resistance R r due to fibre fracture is thus expressed by equation (3). R r = R n R ( nn) ( ) 1 n N = R F N i.e. n = R r N R + R r (2) (3) Now it is well known that the echanical resistance and ultiate elongation of a carbon fibre are directly related to the probability of defects existing along it. According to the weakest link theory, the fracture probability of a single filaent corresponds to the fracture probability of the weakest link containing the ost critical defect and is given by equation (4) for a fibre of length L, where σ r is the ultiate strength, σ the applied stress and θ(σ) an epirical function (ost of the tie, Weibull or sigoid functions are chosen to describe the carbon fibre fracture). This equation is also valid for strains, providing a substitution of σ into ε. Weibull odel: Pr (σ r < σ) = 1 - exp (-θ(σ) L) (4) Soe authors [11,12] have odelled the defect population, estiated fro tensile tests on single filaents or dry carbon tow, thanks to Weibull functions assuing that: 1: Fro an electrical point of view, all the fibres are arranged in parallel in the carbon tow. The transverse electric contacts between different fibres are neglected: Each single filaent is ideally electrically isolated fro its neighbours by a resin layer. 2: The strains of the coposite test saple and of the ebedded carbon tow are equal. Effects due to differences in initial tension of the filaents are neglected. 3: The influences of the atrix and of the glass-resin host aterial on the fibre fracture are neglected; only one unique fracture per fibre is taken into account. The fracture probability of the carbon fibres is then given by equation (5), where ε r is the ultiate strain, N the total nuber of fibres and n the nuber of broken fibres for the applied strain ε. Writing equation (4) in ter of strains, it is possible to express the function as follows (equation (6)). n P ( ε r < ε ) = = N ln ( 1 P ( ε < ε)) θε ( ) = r 1 = L L R R + ln (5) R 1- R R + R Now, the 3 paraeters Weibull distribution being given by equation (7), a cobination with (6) leads to the final relationship (8) between strain ε and electric resistance R. (6) θε ( ) = ε εs ε R ε = ε[ θ( ε) ] + εs = ε ln 1- + εs L R + R (7) (8)

7 with ε : scale (or width) paraeter : shape paraeter (diensionless exponent) ε s : onset paraeter (threshold strain for which the fracture probability still equals ) ε s corresponds to the strain at which the first fracture of the carbon tow occurs. It is experientally deterined fro an analysis of the aplitude of acoustic eission signals, assuing that carbon fibre failures induce signals of aplitudes higher than 95 db, while signals of lower aplitudes coe fro atrix and interface cracking. The coefficients ε and are fitted on the basis of a ean square regression. The values of the coefficients of equation (8) are given on figure 4 for a carbon sensor of,6% noinal ultiate elongation under onoaxial tension. Sigoid odel: Other authors [13] have chosen sigoid functions to odel the fracture distribution during ultifragentation tests (tensile tests on ipregnated fibres). In that case, the test coupons are ade of one single carbon filaent ebedded in a ductile atrix. After the first fibre break, interfacial shearing stresses enable to load the fibre again: The carbon filaent still keeps its reinforceent role. The final failure of the test saple thus occurs after successive fractures of the filaent, which ultiply until the ean length of the fragents reachs the critical length representative of the fibre/atrix syste, for which the filaent is no longer able to reinforce the atrix. Assuptions 1 and 2 previously ade for Weibull odel are still valid here, but this new odel now takes into account ultiple fractures for a sae fibre. Thus it suits better fro a echanical point of view. For this odel, the fracture probability is expressed as a function of the nuber of fractures and no longer as a function of the nuber of broken fibres (any breaks per fibre are possible). Nevertheless, assuing that the nuber (a) of possible fractures is the sae for each carbon fibre and that the fractures preferentially occur on the sae fibre, the expression of fracture probability which is finally obtained is the sae as previously (equation (5)) as shows equation (9). n P ( r < ) = r a. n ε ε = = N rt a. N n N (9) with n r : nuber of fractures n : nuber of broken fibres N rt : total nuber of possible fractures N : total nuber of fibres a : nuber of fractures per fibre As a consequence, equation (6) reaining unchanged, its cobination with the expression (1) of the sigoid function leads to the final relationship (11). A is an epirical coefficient The other coefficients ε, ε s and are deterined as previously and are given on figure 4 in the case of a carbon sensor of,6% noinal ultiate elongation subjected to axial tension. θε ( ) = A 1 exp ε εs ε 1 θε ( ) ε = ε ln 1 + ε A s = 1 ε ln 1 LA ln 1 - R R + R 1 + εs (1) (11)

8 Electric resistance R-Ro (Ω) RESISTANCE ELECTRIQUE R-Ro Ω Weibull function Sigoïd function Experiental ean curve,2,4,6,8 1 1,2 1,4 1,6 ΕΦΟΡΜΑΤΙΟΝ % Strain (%) Paraeters Ro ean (Ω) 12,287 L (),16 ε s (%),65 3,51 ε o (%),185 A - Weibull Sigoid 12,287,16,65 3,94 6,88 34,5 Paraeters of Weibull and sigoid functions Fig.4. Coparison of experiental/calculated results for tensile test on test coupons (carbon tow of,6% noinal elongation at break) EXPERIMENTAL VALIDATION OF MODELS Application to test coupons subjected to uniaxial onotonic tension Figure 4 shows that there is a good correlation between the ean experiental data and the calculated Weibull curve for resistance increases going up to 17 Ω (corresponding to strains up to,87%). For higher resistance increases, the slope of the Weibull curve is higher than that of the experiental curve, because Weibull functions take into account only one fracture per carbon fibre, each fracture leading to an electric resistance increase. In a real life situation, the sae carbon fibre ay break any ties, and successive fractures on the sae fibres do not induce any significant electric resistance variation. The higher the nuber of broken fibres, the higher the probability of ultiple fractures: This could explain the observed divergences between experiental and calculated curves at high strains. In return the beginning of the experiental curve is well fitted by the Weibull odel, because the probability to obtain two fractures on the sae fibre is very low at the tensile test beginning. This probability is also reduced by the load transfer effect between neighbouring fibres, which favours distinct fibre fractures rather than ultiple fractures on the sae fibre [14] When the Weibull function starts to diverge fro the experiental curve, the sigoid function, which does not fit experiental data for lower strains, starts to fit the well. The odel change occurs at an electric resistance increase of 16 Ω, that is to say according to equation (5), for 56% of broken carbon fibres. At that point the fracture probability of a fibre already broken once becoes higher than the probability to break a new intact carbon fibre. Thus the Weibull function can be associated to an electric resistance / strain behaviour law, where each carbon fibre is only broken once, and the sigoid function to a behaviour law taking into account ultiple fractures for a sae carbon fibre (and this induces a slower increase of the electric resistance). Multiple carbon fibre fractures are confired by a visual inspection of the tow after tensile tests: an average of six ajor fractures are regularly spaced on the whole width of the carbon tow.

9 Application to pipes under increasing internal pressure The ai is to deterine the daage state of a pipe subitted to a given internal pressure and instruented with a hoop carbon tow only (ultiate elongation,36 %). For this purpose one needs to calculate the pipe strain fro the electric resistance increase due to the carbon fibre fractures, using the appropriate odel. Then, the value of this strain is copared to the strain critical threshold corresponding to the first daage of the structure, previously deterined by the user (finite eleent calculations carried out in design departents or experiental deterination on prototypes). The value of the strain threshold is here given by the knee-point of the stress-strain curves (,54 % ±,5) recorded during the initial characterisation of the coposite structure. Moreover it is assued that the electric resistance / strain curves of the carbon tows under biaxial loading can be odelled by the sae kind of equations as those used under uniaxial tension. The critical strain of,54% being just above the fracture strain of the carbon tows used (,51 % ±,7), it is necessary to use a Weibull function to odel the electric resistance / strain curve. The coefficients of the Weibull for pipes, fitted on the basis of a ean square regression, are given on figure 5. It appears that the calculated curve fits well the experiental data up to,59% strain. Electric resistance R-Ro (Ω) RESISTANCE ELECTRIQUE R-Ro Ω ,1 Weibull function,54 Experiental ean curve,4,45,5,55,6,65,7 DEFORMATION % Strain (%) Paraeters Weibull Ro ean (Ω) 4,889 L (),314 ε s (%),49,95 ε o (%),178 A - Paraeters of Weibull function Fig.5: Coparison of experiental/calculated results for a pipe under internal pressure (carbon tow of,36% noinal elongation at break) In short, fro a siple easureent of the electric resistance, the user is able to deterine whether the coposite structure is daaged or not. For an electric resistance increase lower than 7,1 Ω, the calculated strain reains under the pipe first daage threshold: the structure is undaaged. On the other hand, if the electric resistance increase easured is higher than 7,1 Ω, that eans that the structural integrity of the pipe has been lost. CONCLUSIONS This research progra ade it possible, in the first stage, to ipleent the resistive daage sensor developed on siple industrial parts: Coposite pipes subitted to an internal pressure loading. For this purpose, it was necessary to choose a type of carbon fibre having characteristics adapted to the first daage critical strains of the structure to be onitored. In the second stage, a odelling of the resistive sensor electric response has allowed to design a calculation code aiing to deterine in real tie, if the daage threshold of the coposite structure has been overgone or not. Thus a siple instruented coposite part being able to self-onitor its daage state has been successfully developed. This onitoring concept still

10 needs to be validated on ore coplex geoetries and ore coplex loading odes. ACKNOWLEDGEMENT The authors would like to thank the Nord-Pas-de-Calais Regional Council for its support. REFERENCES 1. Measures R. M., "Fiber optic sart structures progra at UTIAS", SPIE Fiber Optic Sart Structures and Skin II, 1989, 117 p Valis T. et al., "Passive-quadrature deolated localised-michelson fiber-optic strain sensor ebedded in coposite aterials", Journal of Lightwave Technology, Vol. 9, n 4, 1991, p Shaikh N. et al.,"sart structural coposites with the ability to onitor vibration and daage", 1st European Conference on Sart Structures and Materials, Glasgow, Muto N. et al., "Intelligent CFGFRP Coposites with self-diagnostic function for preventing fatal fracture", Sensors and aterials, Vol. 6, n 1, 1994, p Schulte K., Baron C., "Load and failure analyses of CFRP lainates by eans of electrical resistivity easureents ", Coposites Science and Technology, Vol.36, 1989, p FischeR C., Arendts F.J., "Electrical crack length easureent and the teperature dependence of the ode I fracture toughness of carbon fibre reinforced plastics", Coposites Science and Technology, Vol 46, 1993, p Ceysson O. et al., "Carbon fibres : Sensor coponents for sart aterials", Proceedings 3rd Int. Conf. on Intelligent Materials, Lyon (F), 3-5 Juin 1996, p Taiatto C. et al., "Integrated sensors for in-service health onitoring of glass-resin coposites", Journal of Advanced Materials, Vol.3, N 3, 1998, p Taiatto C., "Conception et analyse du coporteent d'un coposite à capteur intégré en fibres de carbone pour la détection in-situ des endoageents d'une structure verre/résine", PhD report, Université de Lille I/Ecole des Mines de Douai (F), Prabhakaran R., "Daage assessent through electrical resistance easureent in graphite fiber-reinforced coposites", Experiental Techniques, 14, 199, p Bunsell A. R., Soer A., "The tensile and fatigue behaviour of carbon fibres", Plastics, Rubber and Coposites Processing and Applications, Vol. 18, n 4, 1992, p Koornicki J., "Propriétés écaniques des fibres de carbone: fonction de distribution des allongeents à la rupture des onofilaents", Coposites, n 3, 1988, p Baxevanakis C. et al, "Fracture statistics of single-fibre coposite speciens", Coposite Science and Technology, Vol.48, 1993, p Fuwa M. et al, "Tensile failure echaniss in carbon fibre reinforced plastics", Journal of Materials Science, Vol.1, 1975, p