972

Pultruded fibre reinforced polymer reinforcements with embedded fibre optic sensors Alexander L. Kalamkarov, Anastasis V. Georgiades, Douglas O. MacDonald, and Stephen B. Fitzgerald

Abstract: The use of the pultrusion process for the manufacture of fibre reinforced polymer (FRP) composites with embedded fibre optic sensors is discussed. The specific application is the use of smart composite reinforcements for strain monitoring in innovative concrete bridges and structures. The Bragg grating and Fabry–Perot fibre optic sensors are embedded during the pultrusion of FRP rods and the process-induced residual strains are evaluated using these sensors. The behaviour of optic sensors during pultrusion is assessed, and the effect of the embeddment of optical fibres and their surface coatings on the mechanical properties of the composite material is investigated. To verify the operation of the optic sensors embedded in the smart pultruded rods, mechanical tests were conducted and the output of the fibre optic sensors was compared to that of an extensometer. These mechanical tests were performed at room temperature as well as under conditions of low and high temperature extremes. The reliability assessment of the fibre optic sensors further entailed the study of their fatigue and creep behaviour as well as their performance when the rods in which they are embedded are placed in a severe environment (e.g., alkaline solutions) that may simulate conditions encountered in concrete structures wherein the composite rods will be used as prestressing tendons or rebars. Key words: smart composite reinforcements, fibre optic sensors, pultrusion, residual strain, fatigue and creep behaviour, reliability assessment. Résumé : L’utilisation du processus de pultrusion pour la fabrication de composites en polymère renforcée de fibres (PRF) avec des capteurs à fibres optiques encastrés est discutée. L’application spécifique est l’utilisation de renforts en composites futés pour la surveillance de la déformation dans des ponts et structures en béton innovatrices. Les capteurs à fibres optiques à grilles de Bragg et de Fabry–Perot sont encastrés pendant la pultrusion des barres de PRF et les déformations résiduelles induites par le processus sont évaluées à l’aide de ces capteurs. Le comportement des capteurs à fibres optiques pendant la pultrusion est évaluée, et les effets de l’encastrement des fibres optiques et de leurs enduits extérieurs sur les propriétés mécaniques du matériau composite sont étudiés. Afin de vérifier le fonctionnement des capteurs optiques encastrés dans les barres futées fabriquées par pultrusion, des tests mécaniques ont été effectués et la réponse des capteurs à fibres optiques a été comparée à celle d’un extensomètre. Ces tests mécaniques ont été réalisés à température ambiante ainsi que sous des conditions extrêmes de basses et de hautes températures. L’évaluation de la fiabilité des capteurs à fibres optiques a nécessité l’étude de leur fatigue et de leur glissement ainsi que de leur performance quand les barres dans lesquelles elles sont encastrées sont placées dans un environnement rigoureux (e.g., exemple solutions alcalines) qui peut simuler les conditions produites dans les structures en béton où les tiges composites seront utilisées en tant que câbles ou barres de précontrainte. Mots clés : renforts en composites futés, capteurs à fibres optiques, pultrusion, déformation résiduelle, comportement en fatigue et en glissement, évaluation de la fiabilité. Traduit par la Rédaction

Kalamkarov et al.

984

Introduction The potential for the use of composite materials in the fields of civil engineering, transportation, and marine engiReceived July 5, 1999. Revised manuscript accepted March 7, 2000. A.L. Kalamkarov,1 A.V. Georgiades, D.O. MacDonald, and S.B. Fitzgerald. Department of Mechanical Engineering, Dalhousie University, P.O. Box 1000, Halifax, NS B3J 2X4, Canada. Written discussion of this article is welcomed and will be received by the Editor until February 28, 2001. 1

Author to whom all correspondence should be addressed (e-mail: [email protected]).

Can. J. Civ. Eng. 27: 972–984 (2000)

neering is very promising. High strength-to-weight ratios and inherent corrosion resistance are among the properties that make composites attractive for these applications. The introduction of composite materials in these areas has been slowed by the high cost of most composites and a lack of reliable long-term data as compared to more traditional materials such as steel and concrete. To be successful, composites engineers have to utilize the materials as efficiently as possible and to use automated manufacturing techniques. One method of compensating for the lack of longterm data is to monitor the health of the structure using socalled smart composite materials. Smart composite materials are adaptive composite structures that incorporate sensors and actuators. Depending on their type, smart composites can be classified as passively or actively controlled. Passive © 2000 NRC Canada

Kalamkarov et al.

smart materials incorporate sensors and they can assess their own state and integrity, whereas the actively controlled smart materials incorporate both sensors and actuators and they can perform self-adjustment or self-repair as conditions change. The smart materials discussed here (passive) are pultruded FRP reinforcing rods with embedded fibre optic strain sensors. Pultrusion has received little attention in the area of smart composites, and there are currently only a few publications on this subject (Friebele et al. 1994; Friebele et al. 1995; Kalamkarov et al. 1997a, 1997b; 1998a, 1998b; 1999). However, pultrusion is one of the fastest and most cost effective composites manufacturing processes. It is well suited to produce prestressing tendons and reinforcing bars because it can provide the structures with a high degree of axial reinforcement. This makes pultrusion a prime candidate for manufacturing high quality and low cost components for civil engineering applications. The basic operation of pultrusion is simple in concept but quite complex in detail because of the number of mechanical, chemical, and physical factors that are simultaneously involved in the process. In the pultrusion of thermosetting resin composites, the reinforcing fibre rovings are pulled from a fibre rack or creel into a resin bath containing the liquid resin together with appropriate catalysts and promoters. The resin-impregnated fibres then pass through an array of guides that orient the rovings in the desired direction and subsequently enter the pultrusion die that has the profile of the part to be manufactured. Electric strip heaters attached to the die provide the thermal energy necessary to initiate the reaction. Many pultrusion setups have a cooling water supply at the die entrance to control the transition from room temperature to that of the die. The processes that go on inside the pultrusion die are the following (Sumerak 1985). First, the application of heat to the die initiates the curing process, and this chemical reaction progresses under the influence of complex pressure profiles within the die. The reaction is highly exothermic, and at some location inside the die the relationship of heat flux is inverted and the degree of cure progresses to the point where shrinkage allows the part to detach itself from the die walls. The consolidated product finally exits the die and is cut to the desired length. A schematic of the pultrusion setup is shown in Fig. 1. The pullers, which continuously move the product, are usually hydraulic and may be of either the reciprocating or the caterpillar type. The pultrusion process inherently has the potential to generate residual stresses within a composite component. The high speed of the process requires high thermal gradients and a fast cure rate, as the raw materials typically travel through the pultrusion die in about 1 min. The resin matrix is thus subjected to a dynamic cure profile created by strip heaters attached to the die surface. Accelerators and promoters are needed to cure the resin in addition to the normal catalysts. All considered, not much is known about the effect of these factors on the development of residual stresses. One must also consider that the infeeding of reinforcing fibres to the pultrusion die is also a dynamic process, and problems associated with the balance and symmetry of the fibre distribution may occur. Once again, this effect may generate residual stresses within the component. It is therefore useful to investigate the ability of embedded fibre optic sensors to

973

monitor the strains during the processing and to measure the residual strains created by the pultrusion process. It is also desirable to characterize the pultrusion process with respect to the interactions of the optical sensor and process-induced strains resulting from the compaction of reinforcements within the die, processing temperatures, chemical cure reactions, and differences in the thermal expansion of the fibres and resin. Such information has not been published at the present time. The literature contains many references to the study of process-induced residual stresses in composite materials. Residual stresses are generally modeled as a pre-load that may reduce the strength of a composite. In turn, this may lead to problems such as warping, delamination, or microcracking in the matrix. All of these problems have the potential to decrease the service life of the composite structure through premature failure or increase the need for costly maintenance and downtime. It was reported (White and Hahn 1992; Wang et al. 1992) that the reinforcing fibres are essentially inert to chemical change during cure. Conversely, typical thermosetting resins may shrink by as much as 5–12% during cure. The degree of shrinkage is dependent upon the type of matrix as well as the rate of the cure reaction. The large difference between the coefficient of thermal expansion (CTE) for the fibres and the matrix also induces stresses. The larger CTE for the resin matrix and the fact that the resin and fibres bond together result in residual stress development. The residual stresses induced by these effects may exceed the transverse strength of a single ply, resulting in cracking of the matrix. Not only do these cracks act as failure initiation sites, but they also expose the reinforcing fibres to environmental conditions of moisture and chemical degradation. A number of researchers have experimented with embedded fibre optic sensors as a unique means of measuring the process-induced residual strains in composite laminates. The features of such sensors that make them particularly advantageous over their foil gauge counterparts include immunity to electromagnetic interference, corrosion resistance, reduced cabling, excellent sensitivity, and small size. Dunphy et al. (1990) used Bragg grating sensors to monitor cure in a graphite–epoxy layup. The fibre with two inscribed Bragg gratings was embedded between plies of the six-layer prepreg layup in a hot press. The fibre was positioned in a way to enable one of the inscribed gratings to be located inside the uncured laminate while the other grating remained outside for reference. The response of the gratings was monitored throughout the curing program. One of the gratings was left outside the laminate because the the authors wanted to be able to determine the free thermally induced strains and thus factor them out. The sensor readings enabled them to extract valuable information as to the details of the curing process, such as, for example, the onset and completion of vitrification (consolidation) as well as the residual strains produced when the part contracted on cooling. An experimental setup consisting of one Bragg grating sensor, one thermocouple, and two extrinsic Fabry–Perot sensors was used to measure process-induced strains during the cure of a carbon fibre/epoxy [05/905]s laminate (Lawrence et al. 1996). The laminate was assembled by hand and cured in a hydraulic press at a peak temperature of 125°C. In © 2000 NRC Canada

974

Can. J. Civ. Eng. Vol. 27, 2000

Fig. 1. Experimental setup of pultrusion line for embeddment of fibre optic strain sensors.

comparing the time–temperature profile for the overall curing process recorded by the thermocouple with a time–strain profile obtained from the Bragg grating sensor, the authors made the following observations. During ramp-up to the peak cure temperature, i.e., from 16 to 125°C, little or no chemically induced strain existed in the laminate. The small compressive strain observed may have been caused by some initial curing of the epoxy matrix. During the curing phase at 125°C, compressive strains gradually build up, probably due to chemical shrinkage of the thermoset polymeric matrix. The compressive strain reached at the end of the cure cycle was approximately –100 microstrain. Finally, the cool-down period saw large thermal compressive strains reaching levels of a little over 400 microstrain. The results from the Fabry– Perot sensors showed basically the same trend, i.e., little or no chemically induced strains during the ramp-up to the cure temperature as well as during cure and large thermal compressive strains (around –300 microstrain) during the cooldown phase. Another important aspect of the present study is to assess the effects that the embedded optical fibres and their surface coatings have on the mechanical properties of composite materials. Leka and Bayo (1989) conducted a series of experiments involving carbon–epoxy laminates with 150- and 350-mm optical fibres embedded within them. For all experiments the optical fibres were embedded between the outermost laminae of the composite specimens. Their main conclusions were that the 350-mm optical fibre had a more detrimental effect on the composite specimens than the 150mm ones and that the effect of the optical fibre on the host material is more pronounced for the case of thin laminates than for the case of their thicker counterparts. As well, tensile and fatigue tests indicated that laminates that had 150mm optical fibres embedded in them between 0° plies experienced a negligible loss in strength. More importantly, however, it was observed that optical fibres should be embedded between 0° laminae because this tends to eliminate bending of adjacent plies and also lets the matrix material provide the required bonding to the optical fibre. As a result there is no delamination between adjacent plies or between the plies and the optical fibres. Also to be considered is the interface between the optical fibre and the host material. To effectively transfer strain from the resin matrix to an embedded fibre optic sensor, a good bond of the optical fibre to the host material is required. In general, optical fibres are protected with a buffer

coating for increased durability. The bond strength of the optical fibre and host material is therefore dependent upon the characteristics of the buffer coating, i.e., its ability to survive in the harsh composite production environment and its chemical compatibility with the resin matrix used to produce the composite. Carman et al. (1993) investigated the effects of embedded optical fibres on the transverse tensile strength of graphite–epoxy laminates. Both coated and uncoated optical fibres were employed in the study. The polyimide-coated fibres had diameters ranging from 85 to 292 mm whereas the stripped uncoated fibres ranged in diameter from 40 to 237 mm. All optical fibres were appropriately embedded parallel to the surrounding reinforcing fibres. It was determined that the embedded uncoated optical fibres had a major effect on the transverse tensile strength of the laminates, reducing it by as much as 50%. This effect was more pronounced for the case of the thicker fibres. However, for optical fibres smaller than 100 mm in diameter the degradation in the transverse strength was negligible. It was also determined that the use of polyimide coating on the optical fibres had a smaller detrimental effect on the transverse strength of the laminates because of the chemical compatibility of the polyimide coating with the resin material, which promotes a good bond between them. Thus, even for the large-diameter 250-mm fibres, the reduction in the transverse tensile strength was less than 23%. It is understood that the method of embeddment of the optical fibre should be incorporated in the overall manufacturing process in a manner that the producibility of the composite is not affected in any way. Indeed, optical fibres may be readily integrated into composite structures, particularly ones that have a ply structure, and this issue has been successfully addressed by a number of different studies. However, the optical fibres must ultimately be connected to various fibre optic devices such as LEDs and detectors, and also different parts of the same structure must eventually be interconnected. This means that fibre optic egress is just as important as fibre optic ingress, but unfortunately, the issue of accessing the leads of optical fibres has not been dealt with in the sophistication and detail that its importance warrants. A number of researchers (Wood et al. 1989; Jardine et al. 1993) had to deal with the problem of fibre optic egress, but their techniques of lead recovery were specific to their manufacturing methodology, namely hand layup. The systematic lead recovery from pultruded products has not been examined prior to the present study. © 2000 NRC Canada

Kalamkarov et al.

In addition to the effects that the embeddment of the fibre optic sensors in the pultrusion process might have on the mechanical properties of the composite itself, it is necessary to assess the performance of the sensors themselves under conditions of static and dynamic loading. In particular, it would be important to obtain data that would reflect on the repeatability and accuracy of the fibre optic strain measurements and also to compare the output of the pultruded sensors to that from conventional strain gauges such as external extensometers. While fibre optic sensors and smart composite materials have shown promise in replacing or strategically complimenting traditional materials and strain gages, there is not a great deal of data available that reflect upon their long-term behaviour, which, as is the case with all materials, is significantly affected by the surrounding environment. For civil and marine applications, the composite materials and associated fibre optic sensors will encounter, in addition to mechanical stress, external conditions of high and low temperature, humidity, and chemical ion exposure. A typical application may have an expected life span of 50 to 150 years. Researchers must thus rely on accelerated aging tests to gain information with respect to the long-term reliability of the fibre optic sensors. To fully understand environmental effects on smart composite structures, one must consider both the individual components and the overall system. The manufacturers of Fabry–Perot strain sensors specify that the sensors are reliable over a temperature range of –100 to +125°C (RocTest Ltd. 1997). It has been shown that the decay of reflectivity in an optical fibre is fairly small at temperatures up to 300°C and that losses are negligible over a life-span of 50 years if the temperature during that period never exceeds 80°C (Erdogan et al. 1994). If the decay of reflectivity becomes substantial because of some adverse combination of moisture exposure and temperature, there may be some drift of the signal from sensitive devices such as Fabry–Perot and Bragg grating sensors. However, specialized fibre optic sensors can be manufactured to operate at elevated temperatures, although the high associated costs limit the use of such sensors to mainly aerospace applications. For example, Wang et al. (1994) demonstrated and tested a sapphire fibrebased polarimetric sensor that was operational at temperatures in excess of 1000°C. However, the overall effects of temperature extremes on an embedded fibre optic sensor have not been characterized to date. It has also been documented (Habel et al. 1994) that optical fibres and their protective coatings are susceptible to attack by exposure to certain chemicals. These authors studied the effect of highly alkaline solutions (typical in concrete) on the integrity of several coating materials normally applied to the bare optical fibre. It was discovered that at pH values of 11–14 polyimide coatings were seriously degraded, whereas acrylate coatings were only slightly affected. Fluorine thermoplastic coatings showed no evidence of degradation when exposed to alkaline solutions. In conclusion, neither acrylate nor polyimide-coated optical fibres were considered suitable for direct contact with cement and concrete mixtures. This conclusion can be extrapolated to the fibre optic sensors as well. However, if embedded in a protective composite material, these sensors could be incor-

975

porated in concrete structures to monitor the health of such structures. The reliability analysis of the embedded fibre optic sensors, however, should not be limited to the study of their performance in reactive solutions or under conditions of temperature extremes. Equally important is the fatigue behaviour of the fibre optic sensors. In particular, it will be of interest to determine if the sensors still perform adequately and provide a strain readout after enduring a large number of stress cycles and if they retain their accuracy and repeatability. As well, it is known that creep deformations may cause unacceptable dimensional changes or distortion and ultimately final failure if they are of significant magnitude. Both composites and more traditional materials such as steel can exhibit creep behaviour, although it is more often related to elevated service temperatures. Appropriately, another objective of the current research is to characterize the long-term creep behaviour of the composite tendons with the embedded fibre optic sensors and assess their suitability for monitoring in long-term service conditions. The objectives of the research reported herein are the following: to evaluate the residual strains induced during the pultrusion of FRP rods, to assess the behaviour of Fabry– Perot and Bragg grating sensors during pultrusion, to determine how the embeddment of optical fibres and their surface coatings affect the mechanical properties of the composite, to develop new methods of optical fibre lead recovery and sensor pre-reinforcement, to confirm the operation of the embedded sensors in the smart pultruded tendons, and to compare the output of the embedded sensors to that from an extensometer. This last objective involves testing of the smart tendons in ordinary laboratory conditions as well as at high and low temperature extremes and in chemically active solutions. In addition, the fatigue and creep behaviour of the smart pultruded tendons will be assessed.

Fibre optic sensors In this study, the first fibre optic sensors that were embedded during the pultrusion of fibre reinforced rods were of the Fabry–Perot type. These sensors and the demodulating equipment are currently “off the shelf” items (RocTest Ltd. 1997). The Fabry–Perot sensor has been developed to use a broadband light source as opposed to laser light. It is highly sensitive and can make precise, linear, and absolute measurements. Two multimode fibres are inserted and fused into a larger glass capillary tube with an overall diameter of 200– 250 mm. The ends of the fibres that are inserted into the capillary are polished and contain a semi-reflective coating. The distance between the fused locations defines the gauge length of the sensor. The sensor is designed such that a predefined air gap exists between the two polished optical fibre ends within the capillary tube (Fig. 2). Hence, some of the light introduced to the sensor reflects from the end of the lead-in fibre, while some travels through the air gap and reflects from the second polished fibre end. In the reflective mode of operation, both reflections are transmitted back through the lead-in fibre to a detector. As external forces are applied to the sensor, the length of the air gap changes and, hence, so does the phase difference between the two reflections. Several demodulation techniques are available to eval© 2000 NRC Canada

976

Can. J. Civ. Eng. Vol. 27, 2000

Fig. 2. The layout of a Fabry–Perot fibre optic sensor.

uate this phase difference and relate it to strain. One such device, which uses a Fizeau interferometer to aid in the measurement of the Fabry– Perot cavity length, is described in more detail by Belleville and Duplain 1993. The other type of fibre optic sensor used in the present study was of the Bragg grating type. Bragg grating sensors are based on creating a pattern of refractive index differentials directly onto the material of the fibre core. This may be achieved by directing two laser beams operating in the ultraviolet into the fibre from the side. An interference pattern results with alternating bright and dark fringes. At the zones of constructive interference, permanent optical damage is induced at sites occupied by germanium atoms as a result of the intensity of the ultraviolet light. This changes the refractive index of the glass material and creates a periodic pattern in the fibre that resembles a diffraction grating. Fibre gratings selectively reflect certain wavelengths and transmit others, as shown in Fig. 3. Which wavelengths are transmitted and which ones are reflected depend on both the refractive index of the core material as well as the spacing of the pattern. Changes in temperature or pressure will change the refractive index of the core material and hence cause a change in the wavelengths of peak reflection (or transmission). The presence of mechanical strain along the length of the fibre will have a similar effect, since it will change the grating spacing. Measurements of these wavelength shifts provide the basis of operation of Bragg grating sensors (Electrophotonics Corporation 1996). One advantage of Bragg grating sensors is that the shift in the wavelength of peak reflection and (or) transmission is linear with temperature and axial strain. On the other hand, it is not possible to decouple the effects of temperature and strain with just one sensor. In addition, Bragg grating sensors, unlike Fabry–Perot sensors, are quite sensitive to transverse strain because of the photoelastic effect (Saleh and Teich 1991). There are several techniques available to determine the wavelength shift, including optical spectrum analyzers and tunable filters.

The Bragg grating sensors used in the present study were of a single mode type operating at a Bragg wavelength l0 of about 1300 nm.

Experimental materials and equipment Pultruded carbon and glass FRP rods were produced using a urethane modified bisphenol-A-based vinyl ester resin system known for its good mechanical properties and excellent processability. Two types of organic peroxide catalysts were used to cure the resin, diperoxydicarbonate and tert-butyl peroxybenzoate. Adequate release from the die was achieved using an internal lubricant. The mechanical properties of the resin system as well as the properties of carbon and glass fibre rovings that contained a sizing compatible with the vinyl ester resin are listed in Table 1. The 9.5-mm diameter carbon rods were pultruded with 22 ends of rovings giving a volume fraction of 62%, whereas their glass counterparts were pultruded with 26 ends giving a volume fraction of 64%. The Fabry–Perot and Bragg grating sensors were acquired as prepackaged assemblies. The sensing element was located at the front end of an optical fibre of approximately 2.0– 5.0 m total length. The fibre optic core and cladding were protected by a polyimide coating, which acts as the contact surface between the optic fibre and the surrounding composite. Polyimide coating was used to ensure survival of the optical fibre when exposed to the high temperatures in the pultrusion die (Kalamkarov et al. 1997a, 1997b). In these experiments the actual maximum die temperature was 150°C, whereas the polyimide provides protection to 350°C. The Fabry–Perot sensors used in the experiments were rated for ±5000 or 0–10 000 microstrain, and the Bragg grating sensors were rated for ±5000 microstrain. The sensors were not temperature compensated. Pultrusion was carried out on an experimental pultrusion line (Kalamkarov et al. 1997b). The machine is equipped with three temperature zones, each with its own PID temperature controller. The pulling force is supplied by two sets of © 2000 NRC Canada

Kalamkarov et al.

977

Fig. 3. Principle of operation of a Bragg grating sensor. Optical fibre

Bragg Grating sensor I T

R

0

0

Incident light - I

0

Reflected light - R

counter rotating wheels, which provide very consistent pulling speed. The glass and carbon FRP tendons were pultruded at a speed of 12 in./min (30.5 cm/min). The fibres were wet out in a dip-type wet bath, and then distributed evenly in the cross section of the rod by a series of specially machined high density polyethylene cards. One roving traveled a straight path through the cards to the centre of the rod. This centre roving was used to carry the optical fibre through the feed system and locate it in the centre of the rod (Fig. 1). Strains were read using fibre optic demodulating units. To determine how the embeddment of optical fibres and their surface coatings affect the mechanical properties of composite materials, a microstructural analysis was carried out on both the cross section of the pultruded profile and on fracture surfaces obtained from mechanically fractured samples. Kalamkarov et al. (1998b) provide the details on the materials and equipment used for this analysis. The new methods of optical fibre lead recovery and sensor prereinforcement have been developed by the authors, and two patents pertaining to these technologies have been filed. In order to examine the mechanical behaviour of the embedded fibre optic sensors and assess their reliability, smart tendons were produced with the pultrusion process described above. Two lengths of samples were fabricated: short lengths of approximately 0.6–1.0 m and long lengths of approximately 3.6–4.2 m,which would act as a test model for rebars or prestressing tendons. The mechanical tests described herein were performed on the short length samples using an Instron servohydraulic load frame and an appropriate controller. The capacity of the installed load cell is 10 000 lbf (44 kN). Strain was measured by the embedded fibre optic sensors as well as externally attached extensometers. During the actual tests, four analog signals were read into a data acquisition system: one from the extensometer, one from the load cell, one from an LVDT, and a final strain signal from the Bragg grating or Fabry–Perot sensors. A FIZ 10 demodulation unit (RocTest Ltd. 1997) was used to record strain from the Fabry–Perot sensors, and a BIS 1000 PC card (Electrophotonics Corporation 1997) was used to record strain from the Bragg grating sensors.

Transmitted light - T

Table 1. Properties of vinyl ester resin, carbon fibre, and glass fibre rovings. Property

Carbon fibre rovings

Glass fibre rovings

Vinyl ester resin

Axial modulus Radial modulus Strength Poisson’s ratio

227 GPa 70 GPa* 3500 MPa 0.3*

73 GPa* 73 GPa* 2760 MPa* 0.22*

3 GPa 3 GPa 73 MPa* 0.4

*Values taken from literature for similar materials.

Experimental and discussion As a first attempt at embedding a sensor in a pultruded rod, an unmodified Fabry–Perot fibre optic sensor was added to the fibre feed side of the pultrusion process. The forward end of the sensor lead was bonded with a 5-min epoxy to one of the carbon fibre rovings to ensure that it would feed into the die. From the location at which it was bonded, the sensor had to pass through two of the fibre feed cards before entering the die. The sensor was also located towards the outer surface of the carbon fibre rod. After the sensor had passed through the die and had been embedded in the composite rod, the pultrusion process was stopped to enable trimming away of several carbon fibre rovings in order to pass the pigtail and connector through the die. The result of the first trial was a length of carbon fibre rod with an embedded Fabry–Perot sensor. However, when the sensor was tested using the fibre optic readout unit, it was apparent that it had failed and the readings tended to jump from low (as expected) microstrains to very high readings of strain far in excess of the 5000 microstrain limit. Two explanations for the cause of fluctuations in the microstrain readings were proposed. One was that the harsh conditions of temperature, fibre compaction, or resin cure shrinkage in the pultrusion die damaged the sensor. The second explanation was simply that the sensor was handled too roughly before or after processing or that it may have been damaged by contact with the fibre feed cards or entrance into the die. It was decided to conduct a series of experiments that would expose a Fabry–Perot sensor separately to each of the © 2000 NRC Canada

978

Can. J. Civ. Eng. Vol. 27, 2000

Fig. 4. Comparison of outputs from reinforced and unreinforced Fabry–Perot sensors during normal pultrusion.

Fig. 5. Comparison of output from Bragg grating sensor during normal and dry pultrusion.

factors in the pultrusion process. These factors are fibre compaction pressure, elevated temperature, liquid resin, and resin curing shrinkage stresses. Changes were also made to the fibre feed system to allow the sensor to be located more accurately in the centre of the rod and to protect it from damage in the fibre feed system. In order to determine the effects of fibre bundle compaction pressure on the sensors, a sensor was pulled through the die at room temperature with no resin in the process. This trial was carried out using a sensor rated for ±5000 microstrain. The fibre optic readout unit was used to monitor the strain in the sensor as it traveled through the die. When the data were plotted it was observed that the strain levels were very low, of the order of 10 microstrain (Kalamkarov et. al. 1999). Thus it was concluded that fibre compaction pressure had no significant effect on the Fabry–Perot sensor and was probably not responsible for the failure of the first embedded sensor. A second trial was carried out using the same sensor to study the effects of elevated temperatures. The second trial was identical to the first except that the pultrusion die heaters were turned on and allowed to reach their set points, which in this case were 120, 150, and 120°C over three equal zones along the length of the pultrusion die. The data obtained from the demodulating unit as the sensor traveled through the heated die were plotted as microstrain vs. time. It was observed (Kalamkarov et. al. 1999) that the elevated temperatures had a more pronounced effect on the sensor than the fibre compaction pressure did, but strains were still an order of magnitude lower than the range capability of the Fabry–Perot sensor. The strain readings followed closely the temperature variation within the die and were therefore attributed to the thermal expansion of the sensor itself. This was to be expected, since the Fabry–Perot sensors used in the study were not temperature compensated. After the sensor had exited the die and cooled down, the strain readings gradually went back to zero. The sensor was operational after this trial, and so it was deduced that the elevated temperatures within the die were not solely responsible for the failure of the first embedded Fabry–Perot sensor. The third experiment using the same sensor was a normal pultrusion run. The reinforcing carbon fibre rovings were

pulled through the resin bath and into the heated die via the fibre feed guides. The experiment did not commence until the three thermal zones within the die had attained their steady state at 120, 150, and 120°C. As a further check, a thermocouple was also pulled along with the sensor and rovings to record the temperature profile. The sensor survived the process and managed to successfully travel through the fibre feed guides, the die, and both sets of the pulling wheels. However, it failed shortly after the end of the experiment and there was no obvious cause for this failure. The experiment was repeated several times with new Fabry– Perot sensors. A typical strain output is shown in Fig. 4. These sensors, however, failed shortly after they passed through the pulling wheels. These post-fabrication failures were attributed to the radial shrinkage of the composite tendon as it cooled down from the die temperature to room temperature. This shrinkage in turn exerts an external pressure on the sensing element, causing it to collapse. To overcome this problem, a novel method was developed to prereinforce the sensors prior to pultrusion. These prereinforced sensors were pultruded in the same way as the previous sensors and they did survive the process and were still operational after the end of the experiment. The constituent materials used for the pre-reinforcement of the sensors are the same (and in approximately the same proportions) as for the final pultruded product. Thus, the pre-reinforcing technique does not affect far-field strains. Figure 4 shows the strain output from the pre-reinforced sensor together with the output of the unreinforced sensor. Subsequently all Fabry–Perot sensors were pre-reinforced and this ensured their successful pultrusion in both carbon and glass FRP tendons. Unlike Fabry–Perot sensors, Bragg grating sensors showed enhanced survivability during pultrusion and hence it was not necessary to pre-reinforce them. Nevertheless, we did subject Bragg grating sensors to “dry” pultrusion runs (passing of the sensors and glass fibres through a heated die but with the fibre rovings not soaked in resin). Subsequently to the dry runs, a number of normal pultrusion experiments with different Bragg grating sensors were performed, just as was previously done with Fabry–Perot sensors. Figure 5 shows the strain plots from the dry and normal pultrusion runs superimposed. The difference between the two curves is due to the curing of the resin. For example, the peak strain during normal pultrusion is much higher than the equivalent strain during dry pultrusion. This is likely to have been © 2000 NRC Canada

Kalamkarov et al.

979

Fig 6. SEM micrograph shows: (a) excellent interface between polyimide-coated optical fibres and host material and (b) debonding between acrylate-coated optical fibres and host material.

Table 2. Average tensile properties of GFRP and CFRP rods.

Material Glass FRP No optical fibre One polyimide fibre Carbon FRP No optical fibre One optical fibre

Ultimate strength (MPa)

Standard deviation (MPa)

Tensile modulus (GPa)

Standard deviation (MPa)

902 949

126 55

48.2 47.3

1.1 0.7

1245 1246

118 56

140.8 144.2

3.4

caused by an increased thermal expansion of the sensor due to the exothermic reaction accompanying pultrusion and by the expansion of the surrounding resin due to increased temperature. Also to be noted is the fact that, as the product exits the die, the difference between the two curves represents the process-induced strains. As was mentioned earlier, another major component of the current research is to study the survivability of optical fibres in pultrusion and also to determine the effect of these fibres and their surface coatings on the mechanical properties of glass and carbon fibre reinforced tendons. SEM photomicrographs on pultruded rods with embedded optical fibres (Kalamkarov et al. 1998b) indicated an excellent interface between polyimide-coated optical fibres and the host material (Fig. 6a). On the other hand, substantial debonding was evident in the case of acrylate-coated optical fibres (Fig. 6b). Previous studies have reported that acrylate-coated optical fibres could not withstand processing temperatures greater than 85°C, whereas polyimide coated ones had a melting point of 385°C. Since the maximum temperature in the current study was 150°C, the micrographs conformed with the anticipated superior ability of polyimide coatings to survive enhanced temperatures. The results of mechanical tests performed on various glass and carbon FRP rods (10 specimens were used for each test) are summarized in Tables 2 and 3. ASTM D3916 and ASTM D4475 standard tests were used to determine the tensile and shear properties, respectively. As Table 2 indicates, the ten-

Table 3. Average shear properties of GFRP and CFRP rods.

Material

Shear strength (MPa)

Standard deviation (MPa)

Glass FRP No optical fibre One polyimide fibre Two polyimide fibres One acrylate fibre

29.3 27.2 25.4 29.2

1.3 1.1 1.0 2.2

Carbon FRP No optical fibre One polyimide fibre

29.7 29.5

1.6 1.5

sile strength and modulus values of carbon and glass fibre reinforced rods (CFRP and GFRP, respectively) were virtually unaffected by the embeddment of optical fibres. The reason for this is that the reinforcing fibres are essentially solely responsible for the tensile properties of unidirectional composites (Kalamkarov 1992). Table 3, however, shows that as the number of embedded polyimide-coated optical fibres increases from 0 to 1 and 2, the mean shear strength of GFRP rods decreases from 29.3 to 27.2 and 25.4 MPa. The carbon rod (CFRP), on the other hand, did not seem to be affected by the presence of a single optical fibre. To assess and characterize the overall behaviour of the embedded fibre optic sensors, mechanical testing of the pultruded tendons was carried out in ordinary laboratory conditions by applying various proof loads to the tendons while continuously monitoring strain via the embedded optical sensors and a standard extensometer clipped to the pultruded rod. The smart FRP tendons were subjected to two basic wave forms to evaluate their mechanical performance. The first waveform was trapezoidal whereby the load was ramped from a low value (typically 100 N) to a peak value up to 11 000 N, at a slow rate of 90 N/s. The load was held at this level for 20 s and then ramped back down to the initial value at the same rate. The second waveform applied to the smart tendons was a sinusoidal one. The frequency was © 2000 NRC Canada

980 Fig. 7. Strain vs. time plot from extensometer and Bragg grating sensor in a glass FRP tendon subjected to a sinusoidal waveform (peak load of 5000 N) at room temperature.

Can. J. Civ. Eng. Vol. 27, 2000 Fig 9. Strain from extensometer and embedded Fabry–Perot sensor in a glass FRP tendon subjected to a sinusoidal load at 60°C. 3500

Extensometer

3000

Microstrain

2500 2000 1500

Fabry Perot Sensor

1000 500 0 -2000

0

2000

4000

6000

8000

10000 12000

14000

Load (N)

Fig. 8. Strain from extensometer and embedded Fabry–Perot sensor in a glass FRP tendon subjected to a trapezoidal load at room temperature.

one cycle per minute (0.0167 Hz), and a typical range through which the load was cycled was from 400 to 5000 N. Figure 7 shows the results from a sinusoidal test performed on a GFRP tendon with an embedded Bragg grating sensor. The data are plotted as microstrain vs. time. The plot shows that the profile of the strain output from the extensometer follows very closely the profile of the corresponding sensor readings. In addition, there is a very high degree of conformance between the strain readings from the two devices over the entire load range. At the peak loads, the discrepancy is less than 100 microstrain (about 5%), which is quite reasonable given the resolution of the extensometer and the Bragg grating sensor. Many more tests (trapezoidal and sinusoidal) were performed on both GFRP and CFRP tendons with embedded Bragg grating sensors. The results of the experiments indicate that the strain output from the sensors was accurate and consistent and agreed very well with that from extensometers. Figure 8 shows the results from a trapezoidal test performed on a GFRP tendon, this time with an embedded Fabry–Perot sensor. This microstrain vs. load graph illus-

trates that there is a good agreement between the Fabry– Perot sensor and the extensometer. The same conclusion can be reached from many other experiments performed on CFRP and GFRP tendons with embedded Fabry–Perot sensors. For the purposes of testing in high and low temperatures, the pertinent experiments entailed subjecting the GFRP and CFRP tendons to sinusoidal and trapezoidal load waveforms of about 11 kN magnitude inside a temperature chamber. The temperature in the chamber was varied from –40 to +60°C in increments of 20°C. The strain output from the embedded sensors was compared with that from externally mounted extensometers as well as to theoretical strain values. Typical graphs are shown in Figs. 9 and 10. Figure 9 is a microstrain vs. load graph pertaining to a GFRP tendon subjected to a sinusoidal load waveform at 60°C. At the beginning of the test, prior to the application of any load (other than a very small preload), the Fabry–Perot strain reading was about 210 microstrain. This is a purely thermal strain caused by the expansion of the glass tendon. This thermal strain was factored out of both the sensor and the extensometer strain readings by calibrating (nulling) the two strain monitoring devices after the temperature in the chamber had reached steady state and just before any load application. This was repeated for all temperature tests. Thus, the strain outputs shown in Fig. 9 pertain to purely mechanical strain. The conformance between the embedded Fabry–Perot sensor and the externally attached extensometer is remarkable throughout the range of the applied loads. There is, however, more scatter associated with the sensor strain readings than with the extensometer output. Figure 10 shows the results from a trapezoidal test (peak load of around 11 kN) performed on a carbon FRP tendon with an embedded Fabry–Perot sensor. This time, however, the sensor data were compared with theoretical strain readings calculated on the basis of the applied load and the experimentally determined longitudinal modulus of pultruded CFRP tendons with an embedded polyimide-coated optical fibre (Table 2). The figure shows that there is a good agreement between the Fabry–Perot output and the theoretical data, the maximum discrepancy being about 8% at the peak applied load. Many more temperature tests were performed on both GFRP and CFRP tendons, and they all indicated that © 2000 NRC Canada

Kalamkarov et al. Fig. 10. Strain from extensometer and embedded Fabry–Perot sensor in a carbon FRP tendon subjected to a trapezoidal load at –20°C.

the sensor readings conformed well with the corresponding extensometer readings and theoretical results. Thus, the performance of the embedded Fabry–Perot sensors is not affected by ambient temperatures falling within the range of – 40 to +60°C. As far as fatigue testing is concerned, glass and carbon FRP tendons with embedded Fabry–Perot and Bragg grating sensors were subjected to a sinusoidal load waveform of 1 Hz frequency and ranging in magnitude from 6.5 to 11 kN (stress ratio of 0.6). Testing was carried out for a duration of 140 000 to 350 000 cycles. The strain values from the embedded sensor were compared with those from an externally mounted extensometer (Figs. 11 and 12). It can be seen that the sensors embedded in the carbon FRP tendons are unaffected by the fatigue load, and their strain load is repeatable and accurate after a large number of load cycles. The same conclusions were reached for the case of the glass FRP tendons. On comparing the plots from Figs. 11 and 12 one may observe that the extensometer readings were different even though the applied loads were the same. The reason for this discrepancy is the misalignment between the carbon tendons and the spelter sockets (grips) affixed to the ends of the tendons. These sockets are used for the attachment of the tendons in the load frame and the subsequent transmission of the tensile loads. Because of the misalignment, the frame subjected the tendons to a small bending load (in addition to the desired axial load). In the experiment to which Fig. 11 pertains, the extensometer was attached on the “tension side” of the bending moment, whereas in the experiment of Fig. 12, it was attached on the “compression side.” Hence, the readings in Fig. 12 were lower than the corresponding ones in Fig. 11. The next major objective in the research was to characterize the creep behaviour of the pultruded FRP tendons with the embedded fibre optic sensors and assess their suitability for monitoring long-term service conditions. The goal is to investigate the interaction of the composite host material and the embedded sensor under conditions of sustained load levels and to characterize the composite tendons themselves. Both carbon and glass FRP tendons were chosen for short-term creep testing. Both tendons contained an embed-

981 Fig. 11. Strain from extensometer and embedded Fabry–Perot sensor in a CFRP tendon subjected to tension-tension fatigue for 140 000 cycles.

Fig. 12. Strain from extensometer and embedded Bragg grating sensor in a CFRP tendon subjected to tension-tension fatigue for 350 000 cycles.

ded Fabry–Perot sensor. The applied load was 9 kN (for a period of 140 h) for the case of the glass tendon and 13.5 kN (for a period of 350 h) for the case of the carbon tendon. The results of these tests are shown in Figs. 13 and 14. It can be seen that the output from the Fabry–Perot sensors is practically uniform over the entire duration of the tests, and thus it may be concluded that they do not exhibit a shortterm creep behaviour at the given experimental conditions. Long-term testing involved the combined effects of load and alkaline exposure. The smart tendons were placed in a long-term creep test fixture and loaded to 11 000 N. Two long-term creep test fixtures were specifically designed and fabricated for this test. A simplified schematic of this setup can be seen in Fig. 15. The fixtures were constructed of square steel box tube, with a base, two uprights, and a single cross beam. Each fixture was approximately 2.13 m high and 0.91 m wide between the uprights. A smart tendon was then suspended from the cross beam of each fixture by using a chain and a turnbuckle. A shackle was used to attach the weights (solid steel bars) to the tendon. Each test sample was also exposed to an alkaline solution having a pH of 12.8 and composed of 0.32 mol/L KOH, 0.17 mol/L NaOH, and 0.07 mol/L Ca(OH)2. This was accomplished by adding a glass environmental chamber over the length of the tendon. The chamber was fitted with rubber stoppers and sealed with silicone to ensure that the test solution was contained within © 2000 NRC Canada

982 Fig. 13. Carbon FRP tendon at 11.5 kN for 350 h.

Can. J. Civ. Eng. Vol. 27, 2000 Fig. 16. Glass FRP tendon with embedded Fabry–Perot sensor (in an alkaline solution) with a sustained load of 11 kN.

Fig 14. Glass FRP tendon at 9 kN for 140 h.

Fig. 17. Carbon FRP tendon with embedded Fabry–Perot sensor (in an alkaline solution) with a sustained load of 11 kN.

Fig. 15. Experimental setup for long-term creep testing in alkaline environment.

the chamber. The alkaline test solution was held in a reservoir next to the test fixtures. A peristalic pump was used to circulate the solution between the reservoir and the environmental chambers. Testing was carried out for a period of 2.5 months. The first tendons that were tested were GFRP and CFRP tendons

with embedded Fabry–Perot sensors. The strain output from the embedded sensors was compared with that from surface bonded resistive strain gages. The strain gages had to be attached on a very small area of the tendons that was not inside the environmental chamber. In Figs. 16 and 17 the strain form the embedded sensors and the externally affixed gages was plotted vs. time. It can be seen that there is a good agreement between the sensors and the gages. As well, these tests demonstrated that the combination of the highly alkaline solution and the sustained load did not affect the functionality of the embedded sensors.

Conclusions This study has demonstrated that the pultrusion manufacturing process can be used to successfully embed Bragg © 2000 NRC Canada

Kalamkarov et al.

grating and Fabry–Perot fibre optic sensors into the smart FRP reinforcements. It was shown that it was necessary to pre-reinforce Fabry–Perot sensors prior to pultrusion. The Bragg grating sensors show a greater survivability in the pultrusion process and there was no need to pre-reinforce them. Dry pultrusion runs were performed with Bragg grating and Fabry–Perot sensors and the thermal strain output obtained conformed quite well with the temperature profile within the die. Fabry–Perot and Bragg grating sensors were subsequently incorporated in normal pultrusion runs. The different sensors showed strain outputs that had the same basic profile even though the absolute values of the strains varied from sensor to sensor. The information provided by these experiments yields valuable insight as to the specifics of the pultrusion process. Pertinent microscopic analysis indicated that polyimide coating on optical fibres results in a good interface between the optical fibre and the host material. On the other hand, acrylate coating cannot withstand the high production temperature characterizing the pultrusion process and causes severe debonding between the optical fibre and the surrounding resin. Therefore, polyimide-coated optical fibres should be a preferential selection for use in the pultrusion of smart FRP reinforcements. It was also determined that embedded optical fibres have no significant effect on the tensile properties of the pultruded FRP rods, but they slightly deteriorate the shear strength of the composites. This slight decrease in shear strength was seen for the glass rods but was not evident in the case of the carbon FRP rods. Mechanical testing was carried out to assess the overall behaviour of the smart FRP tendons and compare the performance of the embedded sensors to that of traditional strain monitoring devices such as extensometers. Glass and carbon FRP tendons with embedded Fabry–Perot or Bragg grating sensors were subjected to trapezoidal and sinusoidal load inputs at room temperature. In all cases there was a very good agreement between the sensor and the extensometer. GFRP and CFRP tendons were subsequently subjected to sinusoidal and trapezoidal load waveforms under conditions of high (up to 60°C) and low (down to –40°C) temperature, and the strain output from the embedded Fabry–Perot fibre optic sensors was compared with that from externally mounted extensometers as well as to theoretical strain values. It was determined that the strain output from the Fabry– Perot sensor showed an excellent agreement with the corresponding output from the extensometer. Thus, it can be concluded that the performance of embedded Fabry–Perot sensors was not affected by ambient temperatures falling within the range of –40 to +60°C. Fatigue testing of the smart FRP tendons entailed subjecting them to 140 000–350 000 cycles of load varying in magnitude from 6.5 to 11 kN. The output from the embedded Fabry–Perot and Bragg grating sensors showed an excellent conformance with that from the extensometer. Thus, many cycles of applied load had no effect on the performance of the sensors. Subsequent tests indicated that Fabry–Perot sensors embedded in composite materials did not exhibit long-term creep behaviour even if the composites were tested in an alkaline environment. Short-term creep tests performed under

983

ordinary laboratory conditions had previously shown the same results. Thus the smart composite materials with embedded fibre optic sensors have potential for significant benefit in the long-term monitoring of strain levels in field applications.

Acknowledgment This work is supported by the Canadian Network of Centres of Excellence on the Intelligent Sensing for Innovative Structures (ISIS-CANADA), through the Project T3.4 on Smart Reinforcements and Connectors.

References Belleville, C., and Duplain, G. 1993. White-light interferometric multimode fiber-optic strain sensor. Optics Letters, 18(1): 78–80. Carman, G., Paul, C., and Sendeckyj, D. 1993. Transverse strength of composites containing optical fibers. Proceedings of The International Society for Optical Engineering, 1917(1): 307–316. Dunphy, J.R., Lamm, F.P., and Morey, W.W. 1990. Multi-function, distributed optical fiber sensor for composite cure and response monitoring. Proceedings of The International Society for Optical Engineering, 1370: 116–118. Electrophotonics Corporation. 1996. User manual for FLS 3000. Concord, Ont. Electrophotonics Corporation. 1997. Technical data. Concord, Ont. Erdogan, T., Mizrahi, V., Lemaire, P.J., and Monroe, D. 1994. Decay of ultraviolet-induced fiber Bragg Gratings. Journal of Applied Physics, 76(1): 73–80. Friebele, E., Askins, C., Putnam, M., Florio, J., Fosha, A., Donti, R., and Mosley, C. 1994. Distributed strain with fiber Bragg Grating arrays embedded in CRTMTM composites. Proceedings of The International Society for Optical Engineering, 2361: 338–341. Friebele, E.J., Askins, C.G., Putnam, M.A., William, G.M., Kersey, A.D., Fosha, A.A. Jr., Florio J. Jr., Donti, R.P., and Blosser, R.G. 1995. Fabrication and application of low cost optical fiber sensor arrays for industrial and commercial applications. Proceedings of The International Society for Optical Engineering, 2447: 305–311. Habel, W.R., Höpcke, M., Basedau, F., and Polster, H. 1994. The influence of concrete and alkaline solutions on different surfaces of optical fibers for sensors. Proceedings of The International Society for Optical Engineering, 2361: 168–179. Jardine, A.P., Madsen, J.S., Donnelan, T.M., Meilunas, R., and Tobin, A. 1993. Manufacturing studies of fiber-optic embeddment in polymer-matrix composites. Proceedings of The International Society for Optical Engineering, 1918: 60–69. Kalamkarov, A.L. 1992. Composite and reinforced elements of construction. John Wiley & Sons, New York. Kalamkarov, A.L., Fitzgerald, S., and MacDonald, D.O. 1997a. On the processing and evaluation of smart composite reinforcement. Proceedings of The International Society for Optical Engineering, 3241: 338–346. Kalamkarov, A.L., MacDonald, D.O., and Westhaver, P. 1997b. On pultrusion of smart FRP composites. Proceedings of The International Society for Optical Engineering, 3042: 400–409. Kalamkarov, A.L., Fitzgerald, S.B., MacDonald, D.O., and Georgiades, A.V. 1998a. Smart pultruded composite reinforcements incorporating fiber optic sensors. Proceedings of The International Society for Optical Engineering, 3400: 94–105. Kalamkarov, A.L., Liu, H.Q., and MacDonald, D.O. 1998b. Experimental and analytical studies of smart composite reinforcements. Composites Part B: Engineering, 29B(1): 21–30. © 2000 NRC Canada

984 Kalamkarov, A.L., Fitzgerald, S.B., and MacDonald, D.O. 1999. The use of Fabry Perot fiber optic sensors to monitor residual strains during pultrusion of FRP composites. Composites Part B: Engineering, B30(2): 167–175. Lawrence, C.M., Nelson, D.V., Bennett, T.E., and Spingarn, J.R. 1996. Determination of process-induced residual stress in composite materials using embedded fiber optic sensors. Smart sensing processing, and instrumentation. Proceedings of The International Society for Optical Engineering, 3042: 154–158. Leka, L.G., and Bayo, E. 1989. A close look at the embedment of optical fibers into composite structures. Journal of Composites Technology & Research, 11(3): 106–112. RocTest Ltd. 1997. Technical information, equipment, and manuals. St-Lambert, Que. Saleh, B.E.A., and Teich, M.C. 1991. Fundamentals of photonics. John Wiley & Sons, Inc., New York.

Can. J. Civ. Eng. Vol. 27, 2000 Sumerak, J.E. 1985. Understanding pultrusion process variables. Modern Plastics, 62(3): 58–64. Wang, A., Zhang, P., May, R.G., Murphy, K.A., and Claus, R.O. 1994. Sapphire fiber-based polarimetric optical sensor for high temperature applications. Proceedings of The International Society for Optical Engineering, 2191: 13–22. Wang, T., Daniel, I., and Gotro, J. 1992. Thermoviscoelastic analysis of residual stresses and warpage in composite laminates. Journal of Composite Materials, 26(6): 883–899. White, S., and Hahn, H. 1992. Process modelling of composite materials: residual stress development during cure. Part I. Model formulation. Part II. Experimental validation. Journal of Composite Materials, 26(16): 2402–2451. Wood, R.L., Tay, A., and Wilson, D.A. 1989. Design and fabrication considerations for composite structures with embedded fiber optic sensors. Proceedings of The International Society for Optical Engineering 1170: 160–170.

© 2000 NRC Canada