Optical sensors for acoustic detection

Optical sensors for acoustic detection Lucas Borges da Silva A thesis submitted in partial satisfaction of the requirements for the degree of Master in Physics Engineering Supervisor: Prof. Orlando Frazão co-Supervisor: Prof. José Luís dos Santos Department of Physics and Astronomy Faculty of Sciences of the University of Porto October 2014 Acknowledgments First I would like to thank Prof. Orlando Frazão for his support and expertise throughout this journey. His immense experience and practical knowledge were a source of invaluable advice. To him I am very grateful. I would also like to thank Prof. José Luís dos Santos for his kindness and sincere comprehension. It allowed me to conciliate my work with the pursuit of my own dreams outside the academic life, something I am truly thankful for. To all the INESC staff and researchers I would like to leave a kind word of gratefulness for the warm welcome and the ever present will to help in spite of the long and annoying hours of buzzing piezoelectrics. A special acknowledgement to Hugo Martins whose aid was indispensable for the development of this thesis. To all the staff and researchers at the University of Manchester Aerospace Research Institute I would like to express my gratitude for the sympathy and kindness with which I was welcomed and treated throughout my staying in the UK. To Dr. Clara Frias a special note of gratitude for the opportunity to work in such high-standard facilities. Her professionalism and guidance were a great example for me. I would also like to thank Mohamed Saleh and Shankhachur Roy for the companionship and shared knowledge. To my friends, whose words of motivation never ceased to spring throughout all these years, a big thank you. In between light cafe conversations and lunch time philosophical shared thoughts I’ve learnt a great deal from you all. To my family, whose unconditional support has led me here, I am forever grateful. Without your support and comprehension I would never have made it this far. Lastly I am deeply thankful to Inês, for her unassailable support and tenderness, for loving and caring regardless of everything. Thank you for giving me strength when none seemed to be found. Thank you for being a part of my life. Abstract In the last few decades optical fibre sensors emerged as a great promise as sensing devices. High sensitivity coupled with small-size features and remarkable shielding properties in demanding environments motivate the gradual replacement of the conventional electric-based setups. In particular, their development in the branch of acoustic sensing came to help suppress the need for a more stable and reliable method to monitor large scale structures and measure challenging surroundings. Concerning the marked necessity for fibre optic sensors to establish themselves as trustworthy designs with improved performance over the traditional piezoelectrics, this dissertation provides a study of two major optical fibre sensor configurations. Fabry-Pérot interferometers and fibre Bragg gratings were developed for the detection of acoustic waves and had their performances compared. Several sensing heads were characterized with the aid of a piezoelectric acting as an acoustic source. The results emphasize the prospects of micro-applications for Fabry-Pérot sensors and enhance the qualities of FBGs as very sensitive and flexible devices capable of multiplexing. The ability to detect acoustic emissions that are typical from aging processes makes these sensors suitable for structural health monitoring. Keywords: Fibre optic sensors; acoustic sensing; Fabry-Pérot; fibre Bragg gratings. Resumo Nas últimas décadas os sensores em fibra ótica têm emergido como soluções muito promissoras na área dos dispositivos sensores. Alta sensibilidade associada às suas características próprias de pequenas escalas e propriedades notáveis de blindagem eléctrica em ambientes difíceis motivaram uma gradual substituição destes sensores em detrimento de muitos sistemas eléctricos convencionais. Em particular os recentes desenvolvimentos nas aplicações em deteção acústica vieram ajudar a suprimir uma necessidade no estabelecimento de um método estável e fidedigno para monitorizar estruturas em larga escala e efectuar medições em meios rigorosos. Dada a profunda necessidade de uma afirmação das fibras óticas como elementos de alta fiabilidade com um desempenho superior aos tradicionais equipamentos eléctricos, esta dissertação proporciona um estudo sobre duas das mais importantes configurações em fibra ótica. Interferómetros Fabry-Pérot e redes de Bragg em fibra ótica foram desenvolvidos para deteção de ondas acústicas e os seus desempenhos comparados. Vários sensores foram desenvolvidos e caracterizados com o auxílio de um piezoelétrico como fonte acústica. Os resultados reforçam as boas perspectivas da utilização dos interferómetros Fabry-Pérot em micro-aplicações e realçam as qualidades das redes de Bragg como elementos muito sensíveis, flexíveis e possibilitadores de multiplexagem. A capacidade de detectar emissões acústicas típicas de processos de envelhecimento de várias estruturas tornam estes sensores adequados para monitorização não invasiva de estruturas de larga escala. Palavra-chave: Sensores em fibra ótica; emissão acústica; Fabry-Pérot; Redes de Bragg em fibra ótica. Table of contents 1 Introduction ............................................................................................................ 1 1.1 Motivation ....................................................................................................... 1 1.2 Dissertation objectives .................................................................................... 2 1.3 Dissertation structure ...................................................................................... 2 2 State of the art ....................................................................................................... 4 2.1 Interferometric fibre optic sensors ................................................................... 4 2.1.1 Mach-Zehnder interferometer .................................................................. 5 2.1.2 Michelson interferometer ......................................................................... 6 2.1.3 Sagnac interferometer ............................................................................. 8 2.1.4 Fabry-Pérot interferometer....................................................................... 9 2.2 Fibre Bragg gratings ..................................................................................... 11 2.3 Final considerations ...................................................................................... 12 3 The Fabry-Pérot interferometer ............................................................................ 14 3.1 Fabry-Pérot cavities ...................................................................................... 14 3.1.1 Principles and setup .............................................................................. 14 3.1.2 Multi-wave interference phenomena ...................................................... 15 3.1.3 Acousto-optic effect ............................................................................... 19 3.2 In-line Fabry-Pérot sensors ........................................................................... 20 3.2.1 Experimental setup ................................................................................ 20 3.2.2 Results .................................................................................................. 23 3.2.3 Discussion ............................................................................................. 28 4 Fibre Bragg grating sensors ................................................................................. 31 4.1 Description of a FBG .................................................................................... 31 4.2 FBG sensing head ........................................................................................ 34 4.2.1 Experimental setup ................................................................................ 34 4.2.2 Results .................................................................................................. 37 4.2.3 Discussion ............................................................................................. 41 4.3 FBGs in 3D woven composites ..................................................................... 41 4.3.1 Introduction to composites ..................................................................... 41 4.3.2 Experimental method ............................................................................. 44 4.3.3 Results .................................................................................................. 46 4.3.4 Discussion ............................................................................................. 47 5 Conclusions and future work ................................................................................ 48 List of figures Figure 2.1 – Schematic configuration of a Mach-Zehnder interferometer. Light undergoes phase-shifts whilst travelling in the sensing arm, which will then form a fringe pattern when recoupling with the signal from the reference arm. ......................... 5 Figure 2.2 – Schematic configuration of an in-line Mach-Zehnder interferometer. Modes are propagated both along the core and the cladding of the optical fibre; splitting is done with the aid of a LPG. ........................................................................................... 6 Figure 2.3 – Schematic configuration of a Michelson interferometer. The splitting and coupling of the beams occur in the same coupler after reflection in both mirrors. .......... 7 Figure 2.4 – Schematic configuration of a Sagnac interferometer. After counterpropagating within the ring the two beams interfere when reaching the coupler. The result is a fringe pattern obtained by the detector. .................................................. 8 Figure 2.5 – Illustration of a Fabry-Pérot interferometer. Splices or in-built mirrors can act as mirrors, creating the cavity. ................................................................................. 9 Figure 2.6 – Representation of an extrinsic Fabry-Pérot interferometer with a diaphragm as a reflective element. By encapsulating a single mode fibre (SMF) in this way, a Fabry-Pérot cavity is created in the gap between the tip of the fibre and the internal surface. .......................................................................................................... 10 Figure 3.1 – Representation of a longitudinal section of a Fabry-Pérot cavity with in-line design, where two fibres with index and are spliced together. The splice regions guarantee some reflectivity in order to create the cavity with length . ........................ 14 Figure 3.2 – Dependence of the transmittivity function on the phase difference for different values of the finesse. Peak values occur at intervals. Peaks sharpen with the increase of reflectivity, i.e. with the increase of the finesse. .................................. 18 Figure 3.3 – Simple scheme for visualization of the fringe pattern of a Fabry-Pérot interferometer. ............................................................................................................ 21 Figure 3.4 – Setup for acoustic characterization with Fabry-Pérot sensing heads. Laser emission was set to the quadrature point in order to guarantee maximum sensitivity. . 22 Figure 3.5 – Peak amplitude sensing as a function of the applied laser power for a constant PZT frequency of 500 Hz and nm, using a 8,5 mm long FabryPérot cavity. Inset: Peak height profile for 0,79 mW applied laser power. ................... 23 Figure 3.6 – Fringe patterns for Fabry-Pérot cavities with 0,5 mm, 1 mm and 3 mm. Different fringe periods and signal oscillations can be related to sensor length. Sampling OSA specifications: 0,5 mm – resolution: 0,01 nm; point average: 8; sweep average: 16 | 1 mm – resolution: 0,05 nm; point average: 1; sweep average: 8 | 3 mm – resolution: 0,05 nm; point average: 1; sweep average: 8 ............................................ 24 Figure 3.7 – Fringe patterns for Fabry-Pérot cavities with 7,5 mm, 13 mm and 21 mm. Fringe behaviour is non-sinusoidal, aperiodic and unstable. Sampling OSA specifications: 0,5 mm – resolution: 0,01 nm; point average: 8; sweep average: 16 | 1 mm – resolution: 0,05 nm; point average: 1; sweep average: 8 | 3 mm – resolution: 0,05 nm; point average: 1; sweep average: 8 ............................................................. 25 Figure 3.8 – Fringe pattern for a Fabry-Pérot cavity with 98 mm. Fringe amplitude oscillates heavily and it is difficult to establish a well-determined period for the pattern. Sampling OSA specifications: resolution: 0,2 nm; point average: 1; sweep average: 125 Figure 3.9 – Fringe period as a function of sensor length for several Fabry-Pérot sensing heads............................................................................................................. 26 Figure 3.10 – Spectrum of the detected peak amplitudes for a 13 mm Fabry-Pérot sensor with and without the application of averaging methods. ................................... 27 Figure 3.11 – Spectrum of the detected peak amplitudes for three different Fabry-Pérot sensing heads. Sample rate: 2M/s ; iterations: 100 ; points per iteration: 1M. Quadrature points: 0,5 mm: nm | 1 mm: nm | 3 mm: nm. Laser power: 0,5 mm: -20,90dBm | 1 mm: +1,45 dBm | 3 mm: -8,00 dBm ............................................................................................................................ 27 Figure 3.12 – Time response for a Fabry-Pérot cavity with 0,5 mm of length with an applied PZT frequency of 56 kHz ................................................................................ 28 Figure 4.1 – General scheme of a FBG; a) 2D representation of a fibre optic segment containing a FBG; b) FBG refractive index profile as a function of length , where denotes the pure refractive index of the fibre the modified refractive index and as the grating’s period. .................................................................................................... 32 Figure 4.2 – Spectral response of a FBG. When light with a broad spectrum is launched into the fibre, the sensor will act as a filter reflecting a specific λ B wavelength and transmitting the other wavelengths. ...................................................................... 32 Figure 4.3 – Experimental setup for preliminary acoustic sensing and testing with FBGs. The OSA allowed real-time monitoring of the reflected peak Bragg wavelength and study of the lasing threshold current. Noticeable is the fact that since the sensing head works in reflection mode it is not deeply immersed within the setup, making it an easily embeddable element in other measurand structures......................................... 35 Figure 4.4 – Peak laser emission collected at the OSA with an applied PZT frequency of 1 kHz ...................................................................................................................... 36 Figure 4.5 – Experimental design for acoustic sensing with FBGs. Substitution of the OSA for a more sensitive ESA system granted more resolution and data acquisition speed. ......................................................................................................................... 37 Figure 4.6 – Peak amplitudes at sensed by an OSA as a function of the EDFA input current. The red fitting line indicates the expected linear behaviour. ........ 38 Figure 4.7 – Characteristic spectrum from a WDM (black line) used as a filter to sense Bragg wavelength shifts through changes in the received amplitude. Data was acquired by illuminating the FBG with a broadband light source and collecting the light beam in an OSA. Also present is the laser emission at the Bragg wavelength. ........................ 38 Figure 4.8 – FBG sensor frequency response to several exterior perturbations by the PZT. ............................................................................................................................ 39 Figure 4.9 – Peak amplitude spectrums for two different frequency sweeps with the FBG acquired consecutively. The piezoelectric element was adjusted in slightly different positions on top of the grating to produce the two separate spectrums. Red spectrum: SNR = 19 dB; black spectrum: SNR = 17 dB. ............................................. 40 Figure 4.10 – Spectrums for the FBG sensor taken with 20 and 100 iterations. Noticeable is the effect this parameter has on averaging and on the final data display, influencing sensitivity and stability............................................................................... 40 Figure 4.11 - General structure of a 3D woven composite (adapted from [47]) ........... 43 Figure 4.12 - Twill angle interlock glass fibre woven fabric with the FBG inserted between a warp and a weft. ........................................................................................ 44 Figure 4.13 – Infusion process with resin approaching the FBG sensor. ..................... 45 Figure 4.14 - Profile of the reflected Bragg wavelength during the infusion process. ... 46 Figure 5.5.1 – Impedance response of the piezoelectric. ............................................ 50 Figure 5.5.2 – Angular response of the piezoelectric................................................... 50 List of abbreviations EDFA – Erbium-doped fibre amplifier EFPI – Extrinsic Fabry-Pérot interferometer ESA – Electronic spectrum analyser FBG – Fibre Bragg grating FFT – Fast Fourier transform FPI – Fabry-Pérot interferometer FSR – Free spectral range FWHM – Full width at half maximum IFPI – Intrinsic Fabry-Pérot interferometer LPG – Long-period grating MI – Michelson interferometer MZI – Mach-Zehnder interferometer OFS – Optical fibre sensor OSA – Optical spectrum analyser PD – Partial discharge PZT – Piezoelectric SHM – Structural health monitoring SI – Sagnac interferometer SMF – Single mode fibre SNR – Signal to noise ratio SR – Sample rate VI – Virtual instrument WDM – Wavelength-division multiplexer 7 Figure 2.3 – Schematic configuration of a Michelson interferometer. The splitting and coupling of the beams occur in the same coupler after reflection in both mirrors. The fringe pattern is acquired after an interference phenomenon between both rays coming from the sensing and the reference arm. Generally speaking the MI possesses more sensitivity than the MZI since the propagating beam in the sensing arm passes twice through the sensing head. From an integration capability point of view, this sensing method has advantages over the MZI setup. In contrast to the latter, it allows the optical apparatus to be set in just one cluster, bundling together both the signal generation and signal detection equipment. Regarding distributed sensing using optical fibres, Hong et al. proposed a system in which two Michelson interferometers were used as phase detectors [13]. An outside vibration would cause a phase shift in the sensing arm, and from there a realtime application could be developed for a movement detector to be installed in a security system. Nonetheless, improvements must be made before a real application can be made since a modest spatial resolution of just ±51m in a 4012m fibre was achieved. Biomedical imaging technology engaging fibre optic sensors has been an active topic of research in the past few years. In this context, Michelson interferometers have been recently applied in photoacoustic tomography as pressure sensors for quadrature phase detection. This procedure eliminates the frequent need of tuning the laser wavelength for sensitive regions of the fringe pattern, an obstacle commonly found in Fabry-Pérot interferometers [14]. Although not directly an acoustic sensor, the MI in this setup allows for overall ultrasound detection with fast surface scanning properties that are vital for clinical applications. The device’s sensitivity has been estimated to be 0.43/ν Hz·Pa-1, where v is the frequency of the photoacoustic wave, an achievement which is 5 to 10 times greater than other previously results reported elsewhere in literature. 8 2.1.3 Sagnac interferometer Since its first developments in 1983 [15], Sagnac interferometry applied to acoustic sensing has been employed regularly in various applications given their ease of fabrication, simplicity of structure and sturdy response to demanding environmental conditions. They surpass the misalignment challenge encountered in conventional MIs and MZIs, leading to a more straightforward implementation and better noise shielding. Although the classic configuration contained several plain mirrors, the modern Sagnac interferometer (SI) consists of a fibre optic ring configuration where two beams propagate in opposite directions with different polarization states (Figure 2.4). Figure 2.4 – Schematic configuration of a Sagnac interferometer. After counterpropagating within the ring the two beams interfere when reaching the coupler. The result is a fringe pattern obtained by the detector. When continuously rotating the device in a specific direction, the optical path of one of the beams will become larger and the other one shorter, causing them to produce a fringe pattern when recoupled. Likewise other interferometric devices seen so far, coupling between the incident acoustic wave and the light beam will be responsible for changes in the refractive index of the fibre's core, introducing another phase-shift. By analysing the induced variations in the fringe pattern one can then obtain information about the acoustic wave. After being initially applied in the context of rotational sensors, SIs were soon spotted for their advantages as acoustic sensors. Although the first applications were directed to improved hydrophones over the already existent MZI setups, it was in structural health monitoring (SHM) that SIs managed to found their biggest breakthrough. Several studies have reported SI-based sensors for microcracks and structural defects in large-scale concrete structures [16-18]. One particular recent paper reported the development of a leak detection system for long distance natural gas pipelines [19]. Improvements in pipeline monitoring are always highly praised since cracks and subsequent leaks are synonymous of high losses and environmental safety hazard. 9 The most recent trends seem to explore SIs as sensing agents for partial discharges (PD) [20,21]. These electrical discharges occur in high-voltage environments such as high-power transformers and can turn an insulation material into a conductive one. This phenomenon can damage the structure significantly and therefore there is much interest in designing sensing elements that could predict eventual failures and prevent the collapse of the whole structure. High voltage environments present challenges that limit the use of conventional acoustic detectors and hence attentions have been focused in OFSs, given their chemical inertness and electromagnetic shielding coming from their dielectric nature. SIs can be very useful for PD detection since they offer superior stability in face of thermal influences and can sense ultrasonic bursts in a high frequency range. Wang et al managed to design a PD sensor incorporating a SI with a high-frequency response of up to 300 kHz and capable of detection both partial discharges and weak corona discharges [20]. 2.1.4 Fabry-Pérot interferometer Fabry-Pérot interferometry has played an important role over the years in acoustic detection, given its high potential for micro-applications and miniaturization. A Fabry-Pérot interferometer (FPI) is essentially composed of two parallel highly reflecting surfaces and the interference pattern is obtained from the multiple superpositions of the transmitted and reflected beams on both surfaces (Figure 2.5). Figure 2.5 – Illustration of a Fabry-Pérot interferometer. Splices or in-built mirrors can act as mirrors, creating the cavity. The spacing between the mirrors is called a Fabry-Pérot cavity, but in-line configurations can be achieved using splice regions as equivalent mirror surfaces. Whether or not the cavity is internal or external relatively to the optical fibre turns the FPI into an intrinsic or extrinsic interferometer (IFPI or EFPI, respectively). EFPIs are particularly relevant for acoustic sensing since they easily work as sensing heads [22,23]. 10 Like the other interferometric setups, this technique was first applied to the development of optical hydrophones and marine applications, and this trend has been maintained until recently [24,25]. The latest devices are already resistant enough to withstand intense acoustic waves, an obstacle commonly found when working with more fragile piezoelectrics (PZTs). Kim et al have recently proposed a refractive index sensor based on a band-pass filter deposited on the tip of the fibre [25]. The resulting system presented itself as robust and capable to detect frequencies up to 1.56MHz, with a relatively low cost associated. Detection in the megahertz order of magnitude in liquids is of high-importance for industrial applications and medical imaging. Figure 2.6 – Representation of an extrinsic Fabry-Pérot interferometer with a diaphragm as a reflective element. By encapsulating a single mode fibre (SMF) in this way, a Fabry-Pérot cavity is created in the gap between the tip of the fibre and the internal surface. Plenty of studies have been made where diaphragms of different materials were incorporated as reflective elements (Figure 2.6). This configuration allows for significant improvements in the resulting sensitivity of pressure and ultrasound sensors. Graphene is one of the most studied materials for this application, with its high Young modulus and flexibility being the reason behind its choice [26]. Yet, the best results have been found with thin diaphragms of silver with 125 µm of diameter and 300 mm of thickness, with a sensitivity of 70.5 nm·kPa-1. Better results are hoped to be achieved if one manages to reduce the diaphragms thickness without affecting the range of detection. Another major breakthrough was done by Chen et al with the design of a FPI with just 1.5 µm thick chitosan diaphragm, exhibiting excellent physicochemical properties [27]. A response range between 20 Hz and 20 kHz with an impressive sensitivity of 0.002 V·mPa-1 was achieved. In the field of acoustic detection for SHM, Fabry-Pérot interferometry has seen some important applications, especially in the detection of PDs. FPIs present themselves as excellent solutions to detect and locate precisely these discharges. Given the fact that they can resist these harsh environments, contrary to the ordinary PZTs, they can be integrated internally, largely reducing the signal noise and improving the device's sensitivity. Dong et al fabricated an advantageous multiple-point sensing system for weak PDs detection in power transformers from a diaphragm-based Fabry- 11 Pérot sensor, capable of a high-frequency response up to 200 kHz [28]. A more recent work demonstrated an EFPI with a quartz diaphragm and a short cavity length of 50 µm with the ability to sense signals in a frequency range up to a 2MHz, although it was applied to detect PDs at about 350 kHz [29]. 2.2 Fibre Bragg gratings Demonstrated for the first time in 1978 by Hill et al [30], fibre Bragg gratings (FBG) are microstructures printed directly in the core of the optical fibre. These microstructures are periodic modulations of the core's refractive index, making them act as reflective elements in a very specific and well-defined wavelength determined by the grating spacing. By analysis of the reflected signal one obtains information regarding the grating's period and the changes it has undergone. Since outside perturbations such as temperature variations, externally applied pressures or strains can significantly change the sensed core's refractive index, one can successfully design sensors for these parameters. As one of the most promising developments in fibre optic sensing, FBGs have spurred much research during the past decades [31-33]. They have been applied with great success in ultrasonic detection ever since the first preliminary experiments in 1996/7 by Webb et al [32] and Fisher et al [33] for medical applications. Motivation for these developments has been centred on their notable properties and special suitability for less invasive medical procedures, given their compact structure, small diameter and chemical inertness inherent to optical fibres. Applications in biomechanics [34] and ultrasonic biopsies [35] have been reported, but perhaps the field in which FBG has had greater impact is in SHM, where many studies and applications have been undertaken in fields such as civil engineering, aeronautics, aviation and other engineering branches [36-38]. Chan et al showed the appliance of FBGs for SHM of Tsing Ma bridge, the world’s longest suspension bridge, employing a set of 21 FBGs multiplexed in three strands of fibres [36]. With an adjustable scanning rate of up to 20 kHz, a resolution of 1 pm and an accuracy of 10 pm, this remarkable system monitored strain in different parts of the bridge, revealing the dynamic response of the structure during train passages. Acoustic emissions from structural damage are typically within the range between 1 kHz and 1 MHz and strains are on the micrometric order [39]. Detectors must then be able to respond in high-rate frequencies and possess great sensitivity, 12 properties in which FBGs overperform PZTs largely. In order to sense ultrasounds with FBGs it is necessary to employ signal demodulation and for this there are typically two available methods [40]. In the first one, a tuneable laser is set to emit in a steep region of the spectrum, where one can assume a linear regime (near the points that define the full width at half maximum (FWHM) for example). The detected intensity change will then relate to the ultrasound amplitude that is being exerted on the FBG. It has been demonstrated that the length of the FBG determines the maximum frequency of the detected ultrasound [2], with shorter fibres leading to a higher range of detection. On the other hand, shorter fibres also imply a less steep spectral gradient, which diminishes the sensitivity. To solve this compromise between sensitivity and range of detection, Rosenthal et al used phase-shifted FBGs (PS-FBG), decreasing the effective length of the grating and improving sensitivity [41]. Though highly sensitive this technique requires a tuneable laser, an expensive instrument not always available. The second solution uses a broad-spectrum light source instead of a laser. This allows for a cost reduction while making multiplexing an option. Adding a second filter is needed to demodulate the Bragg wavelength shift into optical intensity. The conventional method for ultrasound sensing requires fixing the fibre along the structure that one wants to investigate. Although good sensitivity is achieved, this method has the inconvenient that the reflection peaks suffers fluctuations according to different deformations applied to the structure. Tsuda et al placed a FBG in a small moving platform set by the structure, creating an ultrasound sensor immune to deformation [42]. The study demonstrate that the reduction in signal to noise ratio (SNR) for the non-fixed sensors relative to the fixed ones was just 6 dB and that the ultrasound-induced strain detection was within the typical range of sub-micron strains. The necessity to optimize sensors for reliable and precise monitoring of large scale structures has fuelled research with FBGs as sensing heads. Further developments are to be expected since these sensors have proved to be one of the best solutions available nowadays. 2.3 Final considerations Optical fibre sensors can be moulded into many shapes and designs for different applications, but their transversal qualities as reliable, flexible and sensible devices remain a powerful trademark. Over the years they have been steadily pointed out as futurist sensors, but as the price of optical components decreases and more and more 13 applications are reported, general implementation of OFSs in our daily lives seems more likely. Some interferometric configurations like the MI or the SI are progressively being left behind and only finding usage in niche applications, but MZI configurations still seem to be applied regularly in different fields. On the other hand Fabry-Pérot interferometry seems to have the focus of its development in EFPIs, towards miniaturization of devices and microapplications. Their ability to be easily immersed in demanding environments should promise more future applications. Increasing commercialization of FBG systems along with standardization of sensor characteristics make these devices the main propellers of OFSs study and research. Further applications of FBGs, especially in SHM, are to be expected. 14 3 The Fabry-Pérot interferometer In the domain of interferometry applied to acoustic detection, Fabry-Pérot interferometers (FPIs) stand out as a preeminent solution. Thanks to high integrability, FPIs are feasible for withstanding continuous progress towards miniaturization, a current trend transversal to many technological fields. Acknowledging their relevance, an understanding of the physical principles behind FPIs together with an experimental approach involving acoustic characterization has been done. With this effort one aimed to get a clearer view of the importance and room for improvement of this interferometric device. 3.1 Fabry-Pérot cavities The following section provides a theoretical treatment pertaining Fabry-Pérot interferometry. The main mathematical relationships that describe the behaviour and operation of these sensors are covered, as well as a fitting topic on acousto-optic coupling. 3.1.1 Principles and setup The standard plane parallel Fabry-Pérot interferometer is an optical device constituted by two highly reflective planar mirrors. They stand perpendicular to the optical axis, with their reflective surfaces facing each other, distanced by some specific length – the Fabry-Pérot cavity. Figure 3.1 – Representation of a longitudinal section of a Fabry-Pérot cavity with in-line design, where two fibres with index and are spliced together. The splice regions guarantee some reflectivity in order to create the cavity with length . 15 This design allows the setup to act as an optical resonator where multi-wave interference occurs. Using a coherent laser source, light is launched along the optical axis to the inside of the cavity through the first mirror. The beams are then repeatedly reflected within the cavity in a process resembling a feedback system. The waves eventually interfere at the end of the cavity forming a well determined fringe pattern. Apart from this conventional configuration, FPIs cavities can also be designed by splicing the ends of two segments of optical fibre with index to a fibre with index (Figure 3.1). This assures some reflectivity in the splice region which leads to multiple reflections in the cavity (i.e. the fibre with refractive index ). Hence, the theoretical treatment for the standard mirror setup can also be applied to the in-line design. This configuration is nowadays particularly important and will be treated with more detail further. As already seen, other Fabry-Pérot configurations can be achieved, for instance using external diaphragms where the cavity is made of a small air gap between the tip of the fibre and the diaphragm itself. 3.1.2 Multi-wave interference phenomena Let us assume a light beam travelling parallel to the optical axis along a segment of optical fibre pertaining to an in-line FPI, like the one depicted in Figure 3.1. For the sake of simplicity, the cavity is treated as being lossless. After entering the cavity through the first splice, a light beam propagates along its length until it reaches the second splice, where it is partially reflected back. Then, as the reflected light beam travels backwards, it is reflected again in a similar fashion in the first splice, this time regaining its original direction of propagation. Thus an interference pattern is obtained after light transits through the second splice, as a result of the interference between the beams that passed through the splices without suffering reflection and the beams which suffered multiple reflections within the cavity. In the process of undergoing two reflections in one round trip inside the cavity, a light beam acquires a certain phase shift, determined by (3.1) where is the effective refractive index of the fibre that constitutes the cavity, is the wave vector defined as and is the wavelength of the laser. Neglecting any absorbance effect that takes place in the splice regions, the intensity distribution 16 obtained of the transmitted pattern after multiple interferences (also called transmittivity) assumes the shape of an Airy function: ( ) (3.2) (3.3) where is the reflectivity of the splice, and are the amplitudes of the transmitted and incident waves respectively, and is an auxiliary function introduced for simplification, generally referred as the parameter or as the coefficient of finesse. Some special cases of the Airy function are of particular interest, namely when it presents a minimum or a maximum, corresponding to the event of destructive or constructive interference, respectively. In order to attain constructive interference the various reflected waves must be in phase upon reaching the end of the cavity. Formally speaking, the reflected waves have to fulfil the condition that the different phases acquired during multiple reflections must add to be a multiple of : (3.4) where is an integer. Likewise, destructive interference is represented as valleys in the graphic and occurs when the phase difference of two incoming waves is a multiple of : (3.5) The transmittivity function contains two useful quantities that serve as quality factors for the FPI: the free spectral range (FSR) and the full width at half maximum (FWHM). The FSR is a measure of the sensitivity of the device and provides the range of frequencies or wavelengths that can be observed. It can be defined as the spacing in optical frequency or wavelength between consecutive peaks of interference: (3.6) (3.7) 23 distinguish the outside perturbation as an independent phenomenon among the background noise. The used VI (virtual instrument) lacked an automation routine that would connect the generator as a trigger element in the LabView program. This meant that the user had to manually sweep the frequencies one by one, making it a slow process. One single sweep could take typically somewhere between ten seconds to one minute or more, according to the desired specifications. Sweeps intended to cover large frequency ranges with short steps meant that sometimes one had to spend two to three hours characterizing a single sensor. 3.2.2 Results A verification of the right state of operation of the laser was done in order to assure a correct characterization of our sensors and to confirm the feasibility of the upcoming results. To this end a PZT was set to a specific frequency and placed directly on top of the sensor head, using the experimental setup displayed on Figure 3.4. Sampling of the signal’s amplitude relative to each specific frequency followed, while varying the laser output power (Figure 3.5). 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4 0,0 0,2 0,4 0,6 0,8 1,0 1,2 480 490 500 510 520 0,0 0,2 0,4 0,6 0,8 1,0 Signal amplitude (a.u.) Frequency (Hz) 0,79 mW Peak amplitude (a.u.) Power (mW) Figure 3.5 – Peak amplitude sensing as a function of the applied laser power for a constant PZT frequency of 500 Hz and nm, using a 8,5 mm long Fabry-Pérot cavity. Inset: Peak height profile for 0,79 mW applied laser power. Early experiments involved fabricating sensors 30 to 150 millimetres long. Though easy to produce and integrate due to good handling, these rather large sensors yielded unclear fringe patterns, not showing distinct fringes or periodic behaviour. As the relationship between sensor length and sensitivity started to be 24 understood, effort was placed in developing smaller sensors, with the smallest sensor fabricated having 0,5 mm of length. For sensors with less than 10 mm it became increasingly hard to fabricate ever smaller sensors as one tried to manoeuvre very tiny segments of fibre with mechanical cleavers and fragile splices. The most interesting results came indeed from the smallest sensors, namely the ones with 0,5 mm, 1 mm and 3 mm of length, all of them showing well defined fringe patterns (Figure 3.6). Stability in the fringe amplitude can be compared, when it is easy to note that sensor length relates to this parameter. Thus the 5 mm sensor shows less signal amplitude oscillation while in the other end the 3 mm sensor presents a more erratic behaviour than the other two. Also, between these three it is possible to see that there is a tendency of having a smaller fringe period with increasing sensor length. 1525 1530 1535 1540 1545 1550 1555 -30 -28 -26 -24 -22 -20 1525 1530 1535 1540 1545 1550 1555 -35 -30 -25 -20 -15 -10 -5 1525 1530 1535 1540 1545 1550 1555 -25 -20 -15 -10 -5 0,5 mm Signal amplitude (dB) 1 mm Wavelength (nm) 3 mm Figure 3.6 – Fringe patterns for Fabry-Pérot cavities with 0,5 mm, 1 mm and 3 mm. Different fringe periods and signal oscillations can be related to sensor length. Sampling OSA specifications: 0,5 mm – resolution: 0,01 nm; point average: 8; sweep average: 16 | 1 mm – resolution: 0,05 nm; point average: 1; sweep average: 8 | 3 mm – resolution: 0,05 nm; point average: 1; sweep average: 8 The studied sensors with lengths between 75 mm and 210 mm showed a new spectrum when compared with the smaller cavities (Figure 3.7). Large main fringes appeared which were not dependent on cavity length. In this case the cavity comprises both the core and the cladding since lights travels in both mediums. These cavities present the same behaviour as sensors of the Figure 3.6. 25 1530 1540 1550 1560 1570 1580 -30 -25 -20 -15 -10 -5 0 5 1530 1540 1550 1560 1570 1580 -35 -30 -25 -20 -15 -10 -5 0 1530 1540 1550 1560 1570 1580 -40 -35 -30 -25 -20 -15 -10 -5 7,5 mm Signal amplitude (dB) 13 mm Wavelength (nm) 21 mm Figure 3.7 – Fringe patterns for Fabry-Pérot cavities with 7,5 mm, 13 mm and 21 mm. Fringe behaviour is nonsinusoidal, aperiodic and unstable. Sampling OSA specifications: 0,5 mm – resolution: 0,01 nm; point average: 8; sweep average: 16 | 1 mm – resolution: 0,05 nm; point average: 1; sweep average: 8 | 3 mm – resolution: 0,05 nm; point average: 1; sweep average: 8 1564 1566 1568 1570 1572 1574 -20 -18 -16 -14 -12 -10 Signal amplitude (dB) Wavelength (nm) Figure 3.8 – Fringe pattern for a Fabry-Pérot cavity with 98 mm. Fringe amplitude oscillates heavily and it is difficult to establish a well-determined period for the pattern. Sampling OSA specifications: resolution: 0,2 nm; point average: 1; sweep average: 1 Sensors with even greater length presented aperiodic spectrums, resulting from multi-interference of light-waves on the fibre’s cladding, forming a multimodal interferometer (Figure 3.8). 26 The relationship between the fringe period and sensor length was studied for all the developed sensors (Figure 3.9). Since it was impossible to determine fringe periods for the larger sensors produced, one could only plot for lengths up to 8 mm. 0 2 4 6 8 10 0 2 4 6 8 10 (nm) L (mm) Figure 3.9 – Fringe period as a function of sensor length for several Fabry-Pérot sensing heads. It is possible to see an inverse relationship between sensor length and the fringe period when taking into account the sensors with 8 mm or less. For sensors larger than 8 mm this relationship is unclear, as is also unclear the fringe pattern itself. The acoustic characterization of the sensing heads was achieved by exposing the sensor heads to a large range of varied and discrete frequencies. Limitations on the range of characterization were imposed mainly by the LabView VI, cutting of frequencies above 500 kHz due to the available computer RAM, and by the PZT, whose impedance response indicated several resonance frequencies starting from 60 kHz. The frequency step applied was typically of 2 Hz, a value chosen after noticing that meaningful frequency oscillations in the spectrum usually spanned a few hundred hertz and also because this setting allowed one to see amplitude decays around the peak frequency. The gathered results prior to the implementation of a proper averaging method in the LabView VI led to only qualitative evaluations of the correct functioning of the tested sensors. After its implementation, efforts were centralised in the characterization of the smaller and more promising sensors, purposely neglecting the others. A quick visual comparison between results with and without averaging can confirm this point (Figure 3.10). 27 010000 20000 30000 40000 50000 60000 -1 0 1 2 3 4 5 6 ln(As) (a.u.) Frequency (Hz) Without averaging With averaging Figure 3.10 – Spectrum of the detected peak amplitudes for a 13 mm Fabry-Pérot sensor with and without the application of averaging methods. The frequency response of the three most promising sensors showed some similarities between them (Figure 3.11). It is worth noting that at some particular frequency intervals (e. g. 15-18 kHz, 43-46 kHz, 51-53 kHz) all the sensors responded with greater signal amplitude. 010000 20000 30000 40000 50000 60000 0 20 40 60 80 100 120 140 010000 20000 30000 40000 50000 60000 0 50 100 150 200 250 300 350 010000 20000 30000 40000 50000 60000 0 5 10 15 20 25 30 35 3 mm Peak amplitude (a.u.) 1 mm Frequency (Hz) 0,5 mm Figure 3.11 – Spectrum of the detected peak amplitudes for three different Fabry-Pérot sensing heads. Sample rate: 2M/s ; iterations: 100 ; points per iteration: 1M. Quadrature points: 0,5 mm: nm | 1 mm: nm | 3 mm: nm. Laser power: 0,5 mm: -20,90dBm | 1 mm: +1,45 dBm | 3 mm: -8,00 dBm Acoustic characterization of the smallest sensors revealed disparate values for the SNR (Table 2), with splice quality certainly having a decisive effect on them. 28 Nevertheless, all of them indicate clearance of signal relative to the present background noise. Table 2 – Signal-to-noise ratios for sensors with 0,5 mm, 1 mm and 3 mm and respective applied laser power. The SNR was calculated as the squared ratio between the signal and noise root mean squared amplitudes. Sensor (mm) SNR (dB) Laser power (dBm) 0,5 13 -20,90 1 33 +1,45 3 29 -8,00 Regarding the temporal response of the sensors, the collected data indicates a stable response in the range 110-130 mV. It is from this response that it is possible to perform a FFT and obtain the acoustic spectrum for each sensor. 020 40 60 80 100 110 115 120 125 130 Signal amplitude (mV) Time (ms) Figure 3.12 – Time response for a Fabry-Pérot cavity with 0,5 mm of length with an applied PZT frequency of 56 kHz 3.2.3 Discussion All the experiments that were carried point to the fact that the smallest sensing heads display the best interference fringes. One likely cause for this effect is the fact that since the optical path sensed by the propagating light is shorter in smaller sensing heads, then also smaller is the portion of light that is dispersed and lost. This in turn translates into a higher fraction of light propagating back and forth along the cavity, thus magnifying the resonance effect inside the sensing head. Fringes from the 0,5 mm sensor oscillate less than their counterparts and fittingly the 1 mm sensor appears to 29 show more stability than the 3 mm sensor. Sensors with even greater length start to have an even more erratic behaviour. Apart from stability, fringe periodicity relates also, and more importantly, to sensor length. The results in Figure 3.9 are in accordance with what would be expected: smaller sensors have higher periods. This is particularly true and easy to confirm with the smaller sensors since the bigger ones (larger than 8 mm) fail to provide clear fringe patterns in the first place. Sensors with an intermediate size produced interesting but yet complex fringe patterns. Main patterns showing large fringes were composed by smaller oscillations, sometimes difficult to distinguish from noise oscillations. This behaviour is due to light travelling both in fibre’s core as in the fibre’s cladding, experiencing different indexes of refraction. It seems reasonable to assume the smaller oscillations as being the real fringe pattern since they follow the fitting line in the graph. When studying the acoustic response the best results were obtained with the smallest sensors developed, creating a logical link between fringe pattern quality and acoustic sensing performance. The production and testing of these three sensors in similar conditions allowed concluding more distinctly about their performance. The ideal case had the sensors displaying a flat and predictable response to the whole array of scanned frequencies together with a high SNR, but the real case deviates largely from this desirable pattern. The sensors show a wiggling spectrum, with amplitudes varying significantly within short frequency intervals. For the purposes of sensing it is not critical that the sensors behave in this way, whereas the important feature is a high SNR in the desirable frequency range so that the signal can be unravelled from the background noise. A further calibration according to the normalized spectrum can enable the sensor to equally perceive signals with the same amplitude but different frequencies. The fact that the wiggling of the spectrum is still due to too much instability not fully compensated by averaging can be in turn a big setback for real-purpose applications. Such an effect could compromise the speed and reaction time of a real system in which one may assemble the acoustic sensor, given that to sense more accurately the signal’s amplitude the sensor needs more time of data acquisition. It can also affect the locating of acoustic bursts characteristic of structural cracks if these are too brief, though it does still allow for the detection of the acoustic emission itself. The SNR exhibited by the smallest sensors is very reasonable, opening ways for their real application. Table 2 shows that the 1 mm sensor had a greater SNR than the 0,5 mm or the 3 mm sensor, but this was only due to the applied laser power being 30 very large compared to the other sensors. This was required because the 1 mm had high loss and so it was necessary to set the laser power to higher values so that the 100-120 mV signal could reach the photodetector. This in turn led to increasing power travelling inside the cavity and thus enhanced the sensed frequencies. When comparing these three sensors it is apparent that in some peculiar frequency intervals all the sensors exhibit above-average amplitudes for the sensed signal, such as the 15-18 kHz interval. The fact that three sensors with different lengths reveal these resonance-type incidents seems to give a clue that the origin of the resonance comes from sources outside the cavity, though it is hard precise something beyond that. It can be anything nearby the setup that senses the PZT’s frequency and reacts positively, feed backing the sensing head with the same frequency. It is also worth noting that among the three sensors the 0,5 mm and 1 mm seem to have more similarities in their frequency response between them than with the 3 mm sensor, a behaviour most likely motivated by their shorter length difference. This might indicate that spectrum changes induced by splice losses and other sources of perturbations are not as dictating when it comes to the acoustic response as initially thought. According to the temporal response displayed by all the sensors, the devices exhibit high stability. Real-time analysis of the temporal response allowed for the observation of instability caused by outside perturbations as these data would oscillate heavily. Steady temporal responses give more credibility to the FFTs that are obtained thereafter and hence are an important aspect when acoustically characterizing FabryPérot sensors. 31 4 Fibre Bragg grating sensors Widely regarded as the most promising branch of optical fibre sensors (OFS), fibre Bragg gratings (FBGs) have been employed with great success in recent applications spanning distinct technological demanding fields, from novel delicate medical procedures to structural health monitoring (SHM) in composites for aerospace parts. Such topicality has surged an interest to evaluate FBGs in the context of acoustic detection and to compare it to other sensing methods. Reported in this section is the characterization of a FBG laser sensor and its feasibility as acoustic sensors. Applications in structural health monitoring (SHM) are also discussed, including strain measurements with embedded FBGs in 3D woven fabrics. 4.1 Description of a FBG By subjecting a section of the core of an optical fibre to an external periodic modulation one can locally modify its refractive index. These fluctuations in the core's refractive index can be achieved with a variety of laser-based techniques, usually involving interferometric methods or photomasking. Using an intense laser source in the UV region, a small segment of the fibre senses the beam intensity distribution and undergoes changes in its refractive index according to that same profile. The printed pattern usually takes the form of a quasi-sinusoidal modulation of the core's original refractive index . A common refractive index profile is shown schematically in Figure 4.1, although as already seen here there are PS-FBG that have aperiodic gratings. Bragg gratings behave as selective mirrors, reflecting highly in a well determined frequency band in the spectrum and transmitting all the remaining wavelengths. The spectral behaviour of a FBG and its principle of operation are schematically presented in Figure 4.2. In its free state, each FBG has a very narrow reflection bandwidth around the peak which is directly dependent to the grating's spacing. This relationship is shown in Bragg's formula , (4.1) where is the reflected Bragg wavelength, is the effective refractive index of the core after modulation and is the grating's spacing. According to the last expression, 32 Figure 4.1 – General scheme of a FBG; a) 2D representation of a fibre optic segment containing a FBG; b) FBG refractive index profile as a function of length , where denotes the pure refractive index of the fibre the modified refractive index and as the grating’s period. one expects a spectral shift in the detected Bragg wavelength when tensional or compressive forces are applied longitudinally to the FBG, since they would cause a change in the grating's length. Indeed, by simply physically tapping gently a FBG during real time monitoring (i.e. inducing slight changes in the grating's length) one can shift the Bragg wavelength significantly up to a few nanometres. Figure 4.2 – Spectral response of a FBG. When light with a broad spectrum is launched into the fibre, the sensor will act as a filter reflecting a specific λ B wavelength and transmitting the other wavelengths. Although straightforward, equation 4.1 does not explicitly show all the dependences that has with other parameters. FBGs are very versatile sensors in the sense that virtually any parameter that can cause the grating to elongate will be sensed. However, the core's effective index is also prone to fluctuate as a function of other perturbations. Indeed, FBGs have a rather complex dependence with strain and temperature, and the decoupling of these two when measuring poses some not so 39 only revealed residual changes in the sensed peak wavelength in the OSA display. During this test the maximum detected was 0,02 nm around the central peak emission wavelength although several parameters were tested and modified. This noise-like order of magnitude led to the conclusion that the was small compared to the available OSA resolution and that an indirect measuring method should be preferred over the direct use of the OSA. Hence a WDM filter was introduced in order to establish a relationship between the detected intensity and the wavelength of the incoming light wave through the WDM’s characteristic spectrum (Figure 4.7). The fact that the system’s laser emits at a steep region of the WDM spectrum near the quadrature point is very important, since this feature adds sensitivity to the system. In this way any small change in the Bragg wavelength should be more easily spotted in the spectrum. In order to study the acoustic response of the FBG, small step frequency sweeps were undertaken and the peak amplitudes for each set of data were retrieved. In this way, a profile of the device’s sensitivity could be drawn, similarly to what was done previously with the Fabry-Pérot interferometers. Single frequency spectrums show that the sensed signal peaks are highly distinguishable from background noise (Figure 4.8). The acoustic characterization of the FBG laser sensor shows spectrums with high-frequency stable regions of operation having SNRs of 17 dB and 19 dB (Figure 4.9). 010000 20000 30000 40000 50000 60000 0 1 2 3 4 5 6 Signal amplitude (a.u.) Frequency (Hz) Figure 4.8 – FBG sensor frequency response to several exterior perturbations by the PZT. Two sets of data were collected with the same initial conditions but with the exception that the position of the PZT on top of the grating was slightly changed. This shifted the amplitude and decreased sensitivity, but the overall behaviour was 40 maintained. Sharp peaks around 18 kHz and 20 kHz are due to external resonances as it can be seen in Figure 4.8. 010000 20000 30000 40000 50000 60000 1 2 3 4 5 6 Peak amplitude (a.u.) Frequency (Hz) Figure 4.9 – Peak amplitude spectrums for two different frequency sweeps with the FBG acquired consecutively. The piezoelectric element was adjusted in slightly different positions on top of the grating to produce the two separate spectrums. Red spectrum: SNR = 19 dB; black spectrum: SNR = 17 dB. 10000 20000 30000 40000 50000 60000 1,0 1,5 2,0 2,5 3,0 3,5 4,0 4,5 5,0 Peak amplitude (a.u.) Frequency (Hz) 100 it 20 it Figure 4.10 – Spectrums for the FBG sensor taken with 20 and 100 iterations. Noticeable is the effect this parameter has on averaging and on the final data display, influencing sensitivity and stability. An important parameter when sampling was the correct adjustment of the number of wanted iterations for each set of data, as it had a direct impact both on the quality of the averaging process and acquisition time. Quick tests revealed no relevant differences between spectrums acquired when applying 20, 30 or 100 iterations. Still, a 41 full sweep was done for 20 and 100 iterations in order to study the effect of this parameter (Figure 4.10). 4.2.3 Discussion The acoustic spectrum reveals that the FBG laser sensor is capable of highfrequency detection in the range between 0-60 kHz with a SNR over 17 dB. At low frequencies the detected signals suffer from high instability and the presence of noise, but at higher frequencies the time response remains constant. This means that the intensity changes in the range between 10-60 kHz are caused by modulation of the WDM through shifts in the Bragg wavelength. The system can hence be applied for the detection of high-frequency bursts, typically acoustic crack emissions generated by aging processes in concrete structures. In this frame of thought, an interrogation system encompassing three FBG laser sensors could be used to spatially localize cracks in a material through a triangulation process. 4.3 FBGs in 3D woven composites As part of a one-month summer internship supported by the University of Manchester, this subchapter reports the work developed at the Aerospace Research Institute in Manchester, United Kingdom. The goal was to embed OFSs in 3D woven composites and attempt a verification of strain distributions along the composite. After a short introduction to the field of composites, the experimental procedure for the sample preparation is described, followed by results for the infusion process monitoring. 4.3.1 Introduction to composites When two or more materials are combined to form another material with properties not seen in any of the original constituents, one obtains a composite. It consists of a bulk material that acts as the main structure (the matrix) and a reinforcing material that adds stiffness, usually in fibre form. The right combination of matrixes and fibres can produce exceptionally strong composites, leading to applications in aerospace parts where often it is required for materials to overcome environments with 42 extreme conditions [45]. Carbon, glass and aramid are among the most common fibre materials, while polyester and epoxies are the most used as matrixes. With the advent of material fabrication in fibre form, it has been discovered that some apparently fragile materials in bulk (such as glass) can be transformed into a very strong fibre form. In bulk form, applied stresses produce cracks that propagate and deteriorate the whole material rapidly due to random surface defects; however, in fibre form, the cracks will only affect a discrete number of fibres, leaving the rest of the material intact. In this way, the theoretical strength of the materials is best seen when in fibre form. Although strong, fibres have preferential directions in which they show their best properties, namely along its length. The ideal case of having a material that would present great isotropic mechanical properties started to be solved when resins were used to glue the fibres together. While resins do not have extraordinary mechanical properties on their own, they do have the ability to be easily shaped into different forms. Allying the tensile strength of the fibres with the homogeneity of applied stress distribution provided by the resin matrix creates light materials with superb mechanical properties, outperforming the industry-based metals in many aspects and possible applications. What is more interesting and promising about these composites though, is the fact that the set of final properties depends not only on the previous properties of the matrix and the reinforcing material, but depends also on the fabrication process. This is something not found when working with metals, for example, where the material’s specifications are given by the supplier. When building a composite, fibres are weaved together forming fabrics in order to attain the best mechanical properties. Traditionally, through processes of interlocking these well determined patterns produce 2D layer sheets of fibres that can eventually be stacked together to form thicker fabrics. By adding a matrix, 2D composites are then formed, revealing notable in-plane properties, but suffering however from delamination problems because of their sheet structure. This setback is translated into poor performance when strains and stresses from impact loads are applied in specific directions, causing the material to fail. To circumvent these problems 3D composites were developed integrating complex structures of orthogonally interlocked fibres, conferring isotropic mechanical properties [46]. The several available techniques (such as 3D weaving, 3D braiding or 3D stitching) all require specialised and expensive instrumentation which thus confines 3D composites usage to highly engineered applications. 43 Currently there are several fabric types that can be achieved by interlacing warps (longitudinal threads) and wefts (transverse threads), with different configurations displaying different levels of fibre crimp, difficulty in draping and stability (Figure 4.11). Plain, twill and leno (with variants) are the most commonly seen woven fabrics configurations. Figure 4.11 - General structure of a 3D woven composite (adapted from [47]) The process of joining the fabric with the polymer based resin is generally a complex process, with the main objective being a homogeneous spread of the resin along the fabric. Several different infusion methods have been developed; from the simple spray or hand lay-up processes to the more sophisticated resin transfer moulding, each technique is more suited to a specific set of fibres and resins. Adding to this are constraints regarding both safety and environmental hazards imposed by the toxic nature of the resins. After spreading the resin through the fabric, the mixture needs to undergo a curing process where the resin reacts chemically to form a thermoset structure, achieving its best mechanical properties at the gel point. It is therefore common for the resin to be supplied together with a suitable hardener in the correct proportion, since this parameter weighs heavily in the final properties. Most resins require heat for the curing process to be held with the reaction rate doubling with each 10 ºC increase, approximately and curing temperatures range typically from 50-150 °C. Heating is made in a step-by-step basis, with slow increases in order to keep the highly exothermic reaction under control and to not allow the heat to build up locally in the sample. 44 4.3.2 Experimental method Key to the assessment of the final structure’s performance is the composite’s response to applied stresses and strains. For this effect there are mechanical tests available, but to understand at a more fundamental level how the composite behaves it is most useful to study locally the strain response, in particular at the interstitial junctions between warps and wefts. FBGs sensors quickly turned out to be the main choice as sensors since a fully automatized interrogation system for FBGs was available. In the other hand, the also available Fabry-Pérot sensors required other optical devices that would make the measurements much more difficult. The fabric used was a 2x2 twill angle interlocked (11 picks/cm) woven fabric made of glass fibre with dimensions 14 x 7,5 cm. The FBG was placed in the centre of the fabric to avoid boundary effects; its embedment in an interstitial zone between one warp and one weft was achieved with the aid of a needle and required great care due to the tightness of the fabric and the brittle nature of the FBG sensor. The chosen matrix was a hot curing epoxy system with Araldite® LY564, a low-viscosity epoxy resin, and Aradur® 2954, a cycloaliphatic polyamine acting as an epoxy hardener. This system combines long pot life and low viscosity with excellent post-cure mechanical properties, being particularly suitable for aerospace applications. Figure 4.12 - Twill angle interlock glass fibre woven fabric with the FBG inserted between a warp and a weft. 45 A wet lay-up method with vacuum bagging was carried for the infusion process. The fabric was laid on a steel plate and covered with peel ply to enhance the homogeneity of the finished surface after curing. A tacky-tape was placed alongside all the borders to prevent vacuum failure with extra attention to the zone containing the output fibre from the FBG sensor and vacuum pipes. After successfully sealing the system with a vacuum bag, the setup was ready to proceed to the infusion of the resin (Figure 4.13). A vacuum pump system was connected to the setup so as to control the rate flow of the resin inside the bag. Slow rates were preferred over fast rates since it improved the chances of the resin fulfilling every cavity and gap along and between the glass fibres. Ideally the resin would be in close contact with the fibres not allowing air bubbles to settle and being a source of material failures. Figure 4.13 – Infusion process with resin approaching the FBG sensor. A study of the sensor response to strains during the infusion was achieved through monitoring of the reflected Bragg wavelength. For this effect, a MicronOptics sm125 FBG interrogator was connected to the FBG, displaying real time shifts of the Bragg wavelength in a ready-to-use interface. It worked by sending broadband light into the fibre and analysing the reflected wavelengths in the same channel. After the infusion process the sample plate was taken to an oven where the curing took place. In accordance with the supplier’s guidelines the curing cycle took 4 hours at 60ºC followed by 8 hours at 120ºC and then 4 hours until decreasing to ambient temperature. 46 4.3.3 Results The monitoring of the infusion process shows the grating being mainly compressed as the resin spread around the fabric and induced strains (Figure 5.1). This indicates that distributing sensing in the fabric could be a good approach to achieve validation of a computational model for strain distribution. Pre-infusion tests already revealed high sensitivity from the FBG sensing head, with direct physical tapping resulting in shifts of Bragg wavelength in the order of 10-1 nm. Prior to the vacuum pumping the sensor was stable with a Bragg wavelength of 1550,4 nm, but after the vacuum had been established the grating underwent elongation. Then, as the resin started to be infused, the pressure inside the vacuum bag increased and caused the grating to enter a state of relaxation, shrinking towards its original length. This is quite possibly the reason because there is a linear decrease of the reflected Bragg wavelength, sensibly during the first 12,5 minutes after the beginning of the infusion process. As for the steep decrease between 12,5 and 18 minutes, it is caused by strains prompted by the resin thermosetting around the fabric. It is interesting to note that since the FBG was placed in a central zone of the fabric the resin did not immediately came into contact with the FBG; however, this did not hindered the strains 010 20 30 40 50 1551,0 1551,2 1551,4 1551,6 1551,8 1552,0 1552,2 1552,4 1552,6 Wavelength (nm) Time (minutes) Figure 4.14 - Profile of the reflected Bragg wavelength during the infusion process. caused by the resin away from the grating to also be sensed. Indeed, the resin only reached physically the FBG after 13 minutes. At 24 minutes the Bragg wavelength 47 showed signs of stabilization; surely around this point the resin had already covered the entire fabric, although it still continued to flow inside of the bag. After 34 and 43 minutes adjustments with the equipment positioning needed to be done and contact with the supporting table was unavoidable. These gestures were readily sensed as peaks and oscillations. 4.3.4 Discussion The FBG sensing scheme proved to be an effective device for monitoring infusion processes, and can be used for mapping strain distribution. Monitoring during the curing cycles and vacuum pumping can also provide a clearer view on the spectrum response of the sensor. Proper encapsulation and protection of both the FBG and the linking optical fibre are critical aspects to be developed if this monitoring method is to become economically feasible. 48 5 Conclusions and future work Chapter two saw an introduction to optical sensors and their features as acoustic sensors. A literature review was done, highlighting mainly the important developments in Fabry-Pérot interferometry and fibre Bragg gratings. In chapter three the focus was on in-line designs of Fabry-Pérot sensors. The development of several sensing heads and acoustic characterization led to the conclusion that they are mostly suitable for micro-applications, where small size and embedding capabilities are a must. High sensitivity coupled with flexibility and resistance to moisture and other chemicals make these sensors a viable alternative to the majority of electrical gauge technologies. Fibre Bragg gratings were examined in chapter four. Although there was no success in retrieving the reflected Bragg wavelengths, results showed very good sensitivity and ability to detect single frequencies amidst the background noise. According to the experiments FBGs are prone to have success in fields that require sensing of faint signals and strains. Smart structures involving optical fibre sensors and composite structures (such as polymers or carbon fibre structures) are likely to become a busy topic of research since the demanding industry that produces the applications of these structures also is in need of a reliable method of monitoring their performance. More specifically, future work connected to Fabry-Pérot interferometry and optical fibre sensing should see developments regarding:  A fast and concise way of producing FP sensing heads in the submicron range, abandoning the purely mechanical methods;  Higher splice quality, which would lead to a greater reflectivity and subsequently to an increase in sensitivity;  The use of pencil lead breaks as a cheap way of characterizing the acoustic response of the sensor;  Applications in harsh chemical environments where FPs outperform the existent electric gauges; In its turn, FBGs should experience developments in:  Improved embedding processes, with adequate coatings that do not interfere with the sensing scheme as well as introduced unwanted strains on the grating.  Higher reliability in temperature and strain simultaneous measurements.