Experimental analysis of postbuckling behaviour on composite preformed plates

Proyecto Fin de Carrera Ingeniería de Telecomunicación Formato de Publicación de la Escuela Técnica Superior de Ingeniería Autor: F. Javier Payán Somet Tutor: Juan José Murillo Fuentes Dep. Teoría de la Señal y Comunicaciones Escuela Técnica Superior de Ingeniería Universidad de Sevilla Sevilla, 2013 Trabajo Fin de Grado Grado en Ingeniería Aeroespacial Experimental analysis of postbuckling behaviour on composite preformed plates Autor: Jesús Barroso Báez Tutores: Jesús Justo Estebaranz Antonio Blázquez Gámez Dpto. Mecánica de Medios Continuos y Teoría de Estructuras Escuela Técnica Superior de Ingeniería Universidad de Sevilla Sevilla, 2022 Trabajo Fin de Grado Grado en Ingeniería Aeroespacial Experimental analysis of postbuckling behaviour on composite preformed plates Autor: Jesús Barroso Báez Tutores: Jesús Justo Estebaranz Profesor Titular Antonio Blázquez Gámez Profesor Titular Dpto. Mecánica de Medios Continuos y Teoría de Estructuras Escuela Técnica Superior de Ingeniería Universidad de Sevilla Sevilla, 2022 Trabajo Fin de Grado: Experimental analysis of postbuckling behaviour on composite preformed plates Autor: Jesús Barroso Báez Tutores: Jesús Justo Estebaranz Antonio Blázquez Gámez El tribunal nombrado para juzgar el trabajo arriba indicado, compuesto por los siguientes profesores: Presidente: Vocal/es: Secretario: acuerdan otorgarle la calificación de: El Secretario del Tribunal Fecha: A mis abuelos Agradecimientos Lo s que me han acompañado en éste arduo camino que ha sido la elaboración del proyecto Fin de Grado, saben que, pese a que sólo puede nombrarse a un único autor, el resultado final no es realmente si no la obra de todos ellos. De mis tutores, Jesús Justo Estebaranz, por su total implicación y ayuda, que ha hecho de faro en un mundo desconocido para mí como era el de la fabricación de composites; y Antonio Blázquez Gámez, por sus extensos conocimientos y paciencia ante cada duda que me surgía, a ambos por introducirme el gusanillo del cálculo estructural, que ahora parece ser el camino que ante mí se abre, y no puedo si no empezar con la mayor de las ilusiones, A todos mis compañeros del laboratorio: Migue, Encarna, Antonio, José Ramón, Serafín (y todos los demás). La ayuda y consejos que me habéis dado son incalculables, cada apuro de los que me habéis sacado, cada anécdota que me guardo, sabed que esto también es parte de vosotros, A mi familia, pues ellos son los únicos que saben realmente lo difícil que ha sido ésto, cada bache; y cada superación, todo lo que al final hemos conseguido, nada hubiera sido si no fuera por vosotros. A todos, de corazón, Gracias Jesús Barroso Báez Sevilla, 2022 III XContents 4.2 Stacking 28 4.2.1 Flat stacking 28 4.2.2 Over mold stacking 29 Mold 29 Mold Preparation 30 4.3 Vacuum bag 33 4.3.1 Vacuum bag elements 33 4.3.2 Vacuum bag manufacturing 34 4.4 Curing 37 4.5 Post-process 38 4.5.1 Edge trimming 39 4.5.2 Distortion Check 40 4.5.3 Traceability 41 5 Testing setup 43 5.1 Tooling 43 5.2 Monitorization 45 5.2.1 Strain gauges 46 Installation 48 5.3 Correlation system 53 5.3.1 Painting 54 5.3.2 Calibration 58 5.4 Test bench 60 6 Test results 61 6.1 Specimen testing 61 6.1.1 PP0001 61 6.1.2 PP0002 62 6.1.3 PP9001 63 6.1.4 PP9002 64 6.1.5 PP4501 65 6.1.6 PP4502 66 6.1.7 PPISO01 67 6.1.8 PPISO02 68 6.1.9 H1V3001 68 6.1.10 H3V1001 69 6.1.11 H1V39001 69 6.1.12 H3V19001 70 6.1.13 H1V34501 71 6.1.14 H3V14501 72 6.1.15 H1V3ISO01 73 6.1.16 H3V1ISO01 75 7 Results 77 7.1 Strain gauge results 77 7.1.1 PP0001 78 7.1.2 PP0002 78 7.1.3 PP9001 78 7.1.4 PP9002 79 7.1.5 PP4501 79 7.1.6 PP4502 80 7.1.7 PPISO01 80 7.1.8 H1V30001 81 7.1.9 H3V10001 81 7.1.10 H1V39001 82 Contents XI 7.1.11 H1V34501 82 7.1.12 H3V14501 83 7.1.13 H1V3ISO01 83 7.1.14 H1V3ISO01 84 7.2 Videocorrelation results 85 7.2.1 PP0001 86 Displacements 86 Strain εxx 87 Strain εyy 88 Strain εxy 89 7.2.2 PP0002 91 Displacements 91 Strain εxx 92 Strain εyy 92 Strain εxy 93 7.2.3 PP9001 95 Displacements 95 Strain εxx 95 Strain εyy 96 Strain εxy 96 7.2.4 PP9002 98 Displacements 98 7.2.5 PP4501 99 7.2.6 PP4502 100 Displacements 100 Strain εxx 100 Strain εyy 101 Strain εxy 102 7.2.7 PPISO01 103 Displacements 103 Strain εxx 103 Strain εyy 104 Strain εxy 104 7.2.8 H1V3001 106 Displacements 106 Strain εxx 106 Strain εyy 107 Strain εxy 107 7.2.9 H3V1001 109 Displacements 109 Strain εxx 109 Strain εyy 110 Strain εxy 111 7.2.10 H1V39001 112 Displacements 112 Strain εxx 113 Strain εyy 113 Strain εxy 114 7.2.11 H1V34501 115 Displacements 115 Strain εxx 116 Strain εyy 116 Strain εxy 117 7.2.12 H3V14501 119 XII Contents Displacements 119 Strain εxx 119 Strain εyy 120 Strain εxy 120 7.2.13 H1V3ISO01 122 Displacements 122 Strain εxx 123 Strain εyy 124 Strain εxy 125 7.2.14 H3V1ISO01 126 Displacements 126 Strain εxx 127 Strain εyy 128 Strain εxy 129 8 Conclusions 131 8.1 Conclusions 131 8.2 Issues and posible improvements 132 8.2.1 Manufacturing 132 8.2.2 Test setup and campaing 132 8.2.3 Results 132 8.3 Future researches 132 Appendix A Scraps 133 A.1 Estimaciones de gasto 133 Appendix B Distortions 135 B.1 Configuraciones iniciales 135 B.2 Optimización 136 Appendix C Measurements 139 C.1 Curvaturas 139 C.2 Grosores 142 Appendix D Pliego de condiciones 143 D.1 Lugar de trabajo 143 D.2 Método y útiles de fabricación 143 D.3 Recursos informáticos 143 List of Figures 145 List of Tables 151 Bibliography 153 Notation aPlate length in X direction bPlate length in Y direction hPLate thickness u,v X and Y offsets (respectively) of a point on the middle surface of the plate w Displacements of the mean surface in the direction normal to the undeformed plate EYoung‘s modulus νPoisson coefficient mSemiwaves number in the X direction nSemiwaves number in the Y direction DiBending stiffness in a given direction P,NCompression resultant load in X direction P0Critical load sin sine function cos cosine function ∂y ∂x,y,xPartial derivative of ywith respect to x. x◦Degree notation, xdegrees. XIII 1 Introduction The most important step in getting a job done is the recognition of the problem. Leó Szilárd, 1961 Pr esent chapter serves as an introduction where is talked background, motivation and objectives of the present project. Besides, a proper State of Art revision is done, and an analitical schem of studied case is presented. 1.1 Background This document is included within the Final Degree Project normative of Escuela Tecnica Superior de Ingenieria de Sevilla, inside Grado en Ingeniería Aeroespacial studies plan. It has been done with the goal of achieving corresponding credits of "Final Degree Project" in order to obtain bachelor´s title in Aerospace Engineering. The department where the project has been developed is "Departamento de Mecánica de los Medios Continuos y Teoría de Estructuras" of "Escuela Técnica Superior de Ingeniería de Sevilla", and, inside of it, in "Grupo de Elasticidad y Resistencia de Materiales". Tutors of this work have been Dr. Antonio Blázquez Gámez and Dr. Jesús Justo Estebaranz. Both testing and manufacturing have been done at "Laboratorio de Elasticidad y Resistencia de Materiales" (LERM) facilities. 1.2 Motivation Composite materials have have been used since Ancient times, being a game changer in terms of history several times, including a lot of examples, such as usage of bone-wood (as seen on figure 1.1) bows by asian tribes around 1700–1100 BCE , was one of the main topics related with the grow of central-asian empires[ 1 ], later adopted by europeans [ 2 ]. Although, by talking with the current definition used :"machroscopic scale combination of two or more materiales with separation interphases between them in order to make a new material with new properties", we can not talk about the usage of modern composites since the last quarter of XX century. This definition has been changing both narrowed down since that times. 1 2Chapter 1. Introduction Figure 1.1 Composite bow, made of bone and wood, played an important role in history because of it´s small size and high power, comparing with conventional bows [3]. When talking specifically about aeronautics, the relationship bewteen weight and resistance has been always one of the most important factors in the designing of the diferent parts of aircraft, being critical in many ways of performance, from civil to military purpouses. An example is the current usage of composites in primary structures of modern aircracft, such as Airbus A350, as seen in figure 1.2. Figure 1.2 A350 upper wing surface, made entirely on composites. Credit: Airbus Defense and Space. Futhermore, recent aplications have appeared in the eclosion of so called new space, hence, new vectors have necessites of diminishing weight even in a harder order than in aircraft, having high load requirements whereas improving structural efficiency at most. Its clear that, as currently , composite materials may play an important role of development in these ways, between others. 1.2 Motivation 3 Figure 1.3 Falcon 9 rocket fairing.Primary structure made in prepreg composites, with a metallic coating to protect it from reentry. Credit: Spacex. Spacex is one of the companies that has decided to incorporate composites in their rockets, paving the way to other manufacturers in the possible aplications of composites in fairings and interstages, originally made of aluminum alloys. A pair of examples of this can be seen in figures 1.3 and 1.4. Figure 1.4 Falcon 9 rocket interstage. Made in prepreg composites. Credit: Spacex. As in recent times diminishing weight and improving resistance is actually pursued as a main objective, composites have been demonstrated that can join an important role in related issues, figure 1.6 shows the current tendency of usage in aircraft industry. More specifically, prepreg laminates and infusion are the main manufacturing processes used in aerospace industry. In the present work one of the most used carbon fibers in the industry, AS4 8552, is discussed with several configurations and one load case in an extended talk. 4Chapter 1. Introduction Figure 1.5 Buckling appeared in aircraft stiffened fuselages . Credit: L. Rodríguez Tembleque. GERM. Coming to a more specific point, the present work will deal with the investigation of a common load case in the aforementioned context: plate buckling, which appears in the case of aircraft fuselages as well as other aerospace structural components1. Figure 1.5 shows one real case. Figure 1.6 Composite usage tendency in Airbus aircraft. Credit: Airbus. Main acomplishment about the work thats going to be presented comes from precedents works, serving as a continuation taking into account some research lines that were proposed, also proposing new procedures and methods within the context of the project. 1Such as rocket interstages and depots 1.3 State of Art 5 1.3 State of Art As it is already known, plates, unlike beams, still save a substantial structural strength when loaded in a compression regime (see figure 1.7). Thus, they present a noval behaviour; [ 4 ], called in literacy postbuckling wherein a stiffnest in a elasto-plastic regime is presented, not losing completely structural strenght when buckled, as beams. In this postbuckling mode, it has been observed that stiffnest decrease in a determined ratio (which depends on the main case), but doesn´t become zero, so that failure loads may achieve values several times higher than first critic load. Figure 1.7 Plate buckling. Theoretical load-displacement relation. Credit [4]. In metallic structures, this capacity has already been exploited [ 5 ] within aircraft design, achieving an important weight reduction and safety respect failure, so that non lineal analysis and suitable verifications must be done. Figure 1.8 Post buckling regime. Stress is distributed increasing at the edges and decresing at the centre. Credit:[4]. Altough, this is not composites case, where this issue takes a more complex shape, owing to different composites failure modes, besides adding present anisotropy. A first step in the current study requires a large displacements approach. Experience has demonstrated that Finite Element Metod (FEM) requires high computational resources and calculing time, besides parametric results doesn´t give good outcomes when deeming into postbuckling modes and its effects in distorsions neither mechanical properties. 12 Chapter 3. Testing Campaing Semiwaves configurations Having only available one mold (see chapter 5), and according to previous works, the following configurations have been proposed: Table 3.2 Configurations. Laminate Configuration 1 Flat 2m=1,n=3 3m=3,n=1 3.1.2 Pannels Accordingly to precedent works, L=200mm. To ease expected defectology inherent to manufacturing process, plates will be oversized plus then fixed to expected lenght Combining both laminates and pre-deformed configurations, 12 different pannel models result. In order to gain accuracy, plane pannels are duplicated (which will be appropiated as shown later). So 16 pannels will be made. Besides, a proper naming for pannels has been developed; in order to improve trazability, with the following structure: • Configuration: PP, H1V3 or H3V1 (where H1 corresponds to one semiwave in the X axis and V3 to 3 semiwaves in Y axis), whilst PP correspond to flat pannel •Laminates:00, 90, 45, ISO (ISO corresponds to case 4 in Table 3.1) •Number of pannel: 01 or 02 PP0001 and PP0002 In this case, being a plane pannel, m=n=0. 12 laminates that integrate composite pannel have 0ºorientation (X axis, load direction). Figure 3.1 represent schematic stacking of pannel: Figure 3.1 Pannel PP00 layer pattern. Displaying one of the symmetrical sides . PP9001 and PP9002 In this case, being a plane pannel, m=n=0. 12 laminates that integrate composite pannel have 90ºorientation (Y axis, transverse to load direction). Figure 3.2 represent schematic stacking of pannel: 3.1 Design of experiment 13 Figure 3.2 Pannel PP90 layer pattern. Displaying one of the symmetrical sides. PP4501 and PP4502 In this case, being a plane pannel, m=n=0. 6 laminates that integrate composite pannel have 45ºorientation, whilst the other 6 have -45º, being these disposed one after another (exception layers 6 and 7, both having -45ºin order to save symmetry). Figure 3.3 represent schematic stacking of pannel: Figure 3.3 Pannel PP45 layer pattern. Displaying one of the symmetrical sides. PPISO01 and PPISO02 In this case, being a plane pannel, m=n=0. 12 laminates that integrate the pannel follow the 45/-45º/0/45/-45/90 stacking. Figure 3.4 represent schematic stacking of pannel: 14 Chapter 3. Testing Campaing Figure 3.4 Pannel PPISO layer pattern. Displaying one of the symmetrical sides. H1V30001 In this case, X direction (load direction) has 1 semiwave ( m=1 ), while Y direction has 3 semiwaves ( n=3 ) 12 laminates that integrate composite pannel have 0ºorientation (X axis, load direction). Figure 3.5 represent schematic stacking of pannel: Figure 3.5 Pannel H1V300 layer pattern. Displaying one of the symmetrical sides. H1V39001 In this case, X direction (load direction) has 1 semiwave ( m=1 ), while Y direction has 3 semiwaves( n=3 ) 12 laminates that integrate composite pannel have 90ºorientation (Y axis, transverse to load direction). Figure 3.7 represent schematic stacking of pannel: 3.1 Design of experiment 15 Figure 3.6 Pannel H1V390 layer pattern. Displaying one of the symmetrical sides. H1V34501 In this case, X direction (load direction) has 1 semiwave ( m=1 ), while Y direction has 3 semiwaves( n=3 ) 6 laminates that integrate composite pannel have 45ºorientation, whilst the other 6 have -45º, being these disposed one after another (exception layers 6 and 7, both having -45ºin order to save symmetry). Figure 3.7 represent schematic stacking of pannel: Figure 3.7 Pannel H1V345 layer pattern. Displaying one of the symmetrical sides. H1V3ISO01 In this case, X direction (load direction) has 1 semiwave ( m=1 ), while Y direction has 3 semiwaves( n=3 ) 12 laminates that integrate the pannel follow the 45/-45º/0/45/-45/90 stacking. Figure 3.8 represent schematic stacking of pannel: 16 Chapter 3. Testing Campaing Figure 3.8 Pannel H1V3ISO layer pattern. Displaying one of the symmetrical sides. H3V10001 In this case, X direction (load direction) has 3( m=3 ) semiwaves, while X direction has just 1 semiwave ( n=1 ) 12 laminates that integrate composite pannel have 0ºorientation (X axis, load direction). Figure 3.9 represent schematic stacking of pannel: Figure 3.9 Pannel H3V100 layer pattern. Displaying one of the symmetrical sides. H3V19001 In this case, X direction (load direction) has 3( m=3 ) semiwaves, while X direction has just 1 semiwave ( n=1 ) 12 laminates that integrate composite pannel have 90ºorientation (Y axis, transverse to load direction). Figure 3.10 represent schematic stacking of pannel: 3.1 Design of experiment 17 Figure 3.10 Pannel H3V190 layer pattern. Displaying one of the symmetrical sides. H3V14501 In this case, X direction (load direction) has 3( m=3 ) semiwaves, while X direction has just 1 semiwave ( n=1 ) 6 laminates that integrate composite pannel have 45ºorientation, whilst the other 6 have -45º, being these disposed one after another (exception layers 6 and 7, both having -45ºin order to save symmetry). Figure 3.11 represent schematic stacking of pannel: Figure 3.11 Pannel H3V145 layer pattern. Displaying one of the symmetrical sides. H3V1ISO01 In this case, X direction (load direction) has 3( m=3 ) semiwaves, while X direction has just 1 semiwave ( n=1 ) 12 laminates that integrate the pannel follow the 45/-45º/0/45/-45/90 stacking. Figure 3.12 represent schematic stacking of pannel: 18 Chapter 3. Testing Campaing Figure 3.12 Pannel H3V1ISO layer pattern. Displaying one of the symmetrical sides. 3.1.3 Material In the wide range of composites actually used, carbon fiber with an epoxy matrix has been selected, following the next reasons: • Actual uses of carbon composites in aerospace industry, so material studied is as much near as possible as real cases • Great experience of personnel and utilites available at LERM, having worked with CFRP for more than 20 years •Uses in previous works •Finally, logistic and disponibility reasons More particularly, the elected composite is the Pre-impregnated (Pre-preg) graphite-epoxy AS4/8552, manufactured by Hexcel corporation [ 13 ]. AS4/8552 is an high performance material of epoxy matrix used in primary aeronautic structures. It displays both good impact strenght and fatigue behavior in a wide range of conditions. Table 3.3 shows matrix resine and fiber material properties. Reology table shows a maximun viscosity in a 150ºC temperature, giving a 1 hour gelification time as shown in figure 3.13 Table 3.3 Fisical properties. Property Units Value Fiber density g/cm31.79 Resine density g/cm31.30 Curated laminate nominal thickness mm 0.130 Nominal fiber volume %57.52 Nominal laminate volume g/cm31.58 3.1 Design of experiment 19 Figure 3.13 Matrix resin material properties. Mechanical properties are shown in table 3.4, supposed dry condition. Table 3.4 Mechanical properties. Property Units Value Ultimate Tensile Strenght at 0ºMPa 2207 Ultimate Tensile Strenght at 90ºMPa381 Young modulus ExGPa 141 Young modulus EyeGPa 10 Tensile strengt under compression at 0ºMPa31531 4 Manufacturing In this chapter will be discussed the actual methodology used in order to create the specimens which will be tested. As said before, imperfections play an important role in searched results, so every step in this chapter has had to be done with the most caring dedication as possible, taking into account both logistics and time constraints. Serve this chapter not only as a register of the plate manufacturing, but also as a compilation of all the complexities, acknowledgements and solutions that appeared during the manufacturing chain. Bearing in mind past and future repeatabilities fo this kind, the same procedures used in [ 14 ], [ 15 ] and [ 16 ] have been used, all of them based in the deep knowledge about composite manufacturing adquired at LERM that can be accessed at [17]. 4.1 Cutting Laminate manufacturing ought to firstly, consider pre-preg cutting. Cutting proccess was realized in the clean room at LERM. Figure 4.1 Original reel,as provided by manufacturer. AS4/8552 is delivered by Hexcel in cylinder reels, as seen in figure 4.1, which are about to be stored in specific conditions that avoid it’s unintentional curing and property degradation, having to be frozen in low temperatures and isolated in freeze chambers. Although, this conditions also avoid proper cutting, so material needs to be tempered before proceding to cut it. 21 28 Chapter 4. Manufacturing Figure 4.12 Laminates in sealed bags ready to be stored. 4.2 Stacking The interphase that spawns during laminate stacking plays a very important role in the character and behavior of composites, as stress will allocate through it. Good continuity is essential so that final plate has good ambiental corrosion properties and thougness both strenght are suitable for. Following previous works (and disponibility), vaccuum bag proccess and autoclave method has been used, which will be discussed in more detail lately in this chapter. Prior to this, prepreg laminate placement was taken as first step. Two stacking methodologies were used as there are considered flat and curved plates. Coming from the original reel 4.1, laminates present one paper shell in one side and other plastic layer (and easy to remove) on the other side. 4.2.1 Flat stacking Considering flat plates, laminates were arranged so that first layer is placed located paper shell facing down the table. Then, plastic protection is removed and next layer is placed so that paper protection is facing up (previously having removed this layer plastic protection before). An alignment tool is used to aid in positioning. Done this, a nylon tool is used just to apply pressure over the layers and removing ressidual air bubles located between layers. This has to be done if possible in fiber direction, starting at the center of the layer and ending at the edge of it. Nylon tool also helps to improve interface adhesion between layers. After 4 layers, a proper vaccuum is needed, so that laminate will be introduced in a vacuum bag and 5 minute vaccuum is done. This helps to eliminate residual air bubles and improve grip. The vacuum bag can be seen in figure 4.13 4.2 Stacking 29 Figure 4.13 Pre-stacking bag. Once this is done, the plates of 12 laminates each are ready to be cured. Last paper shell must not be removed until curing. 4.2.2 Over mold stacking Mold In order to make curve laminates, a 2024-T4 aluminum mold was available since it was used in previous works. Design and manufacturing of the mold is discussed in [ 14 ], being ready to use just needing proper setup, that include rust and resine removal and surface preparation. Figure 4.14 CATIA mold model[15]. The dimensions of the mold are 250x250 mm, although it has a margin of 50 mm on each of its sides where its surface is flat. Therefore, the waves are restricted to a square of 200x200 mm. As explained above, the waves are restricted to a 200x200 mm square. The deformed waves are in two directions, direction 1 with three longitudinal waves and direction 2 with only one longitudinal wave. 30 Chapter 4. Manufacturing Mold Preparation Cleanliness of the mold implied using of acetone and rubbing with a soaked cloth, grinding was considered but not applied so that it could have affected mold geometry distorting the waves, just applying a very soft sanding in order to remove the previous resin layer. Figure 4.15 Mold aspect before preparation. In order to ensure a proper cleaning of the residual resin, a heating of the mold was done, so that resin could be esily removed with sandind. Heating was performed in a way that it couldn´t create great distortions on the mold, just enough temperature (200º) to enhace viscosity of the resin, not scratching the metal. All the process can be seen in figures 4.16a and 4.16b. (a) Heating process . (b) Sanding machine. Figure 4.16 Mold preparation process, including heating and sanding in order to remove the resin. 4.2 Stacking 31 Figure 4.17 Mold once prepared. 32 Chapter 4. Manufacturing The preparation of the mold to stacking required use of Loktite’s Frekote(c) as a release agent, a sealer suitable for metal molds with micropores, scratches and surface defects that has high thermal stability. Application was done with a soaked cloth in the outdoors, because of high volatility. 10 coatings (each one perpendicular to the previous one) were firstly applied, then reduced to 7 in the following plates as it was considered enough. To check the correct application of the coatings, a Tesafilm(c) is glued and checked for non adherence with the mold. Frekote´s aplication can be seen in figure 4.18 Figure 4.18 Frekote applying on the mold surface. Stacking proccedure was almost the same as with flat plates, but in this case, nylon tool was also used to seek mold geometry during pressure application just to achieve that laminates adjust to the mold waves. A final laminate prior to curing can be seen in figure 4.19. 4.3 Vacuum bag 33 Figure 4.19 Complete laminate once stacked over the mold. 4.3 Vacuum bag Chamber curing cycle of composites requires previously the making of a vacuum bag[ 17 ]. This vacuum bag fulfills several goals: •Closes and isolates the setup from outside disturbances. •Improves the grip and compaction between layers, besides removing holes and delaminations. •Uniforms the curing process over the plate, achieving better finish and quality. •Allows piece extraction. 4.3.1 Vacuum bag elements Vacuum bag preparation is composed with several elements involved in its preparation: • Plastic : Acts as the first boundary, isolating the inside from external conditions.It is necessary to make sure that it doesn´t present damage such as holes that could allow outdoor air to get into the vacuum bag. It´s configuration must be done bearing in mind thermal distorsions that could break the bag during the cycle. • Chromate band : Acts as the gasket between plastic and surface, closing vacuum bag. Chromate is a non porous material that allows a proper isolation with high temperatures presented within the cycle • Teflon : Teflon is a polymer, known by anti-grip and inert properties. Also a good permeability and a low friction coefficient. In this case, it acts in the vaccuum bag as a contact between mold/surface and laminates underneath, also between laminate and airwave on the top. •Airwave : It is an that allows to a homogeneus distribution of vacuum along the plate. Also acts as a barrier avoiding damage that mold edges could do to the plastic •Vacuum intakes : Two metal valves wherein hoses will be connected, coming from compressor. • Breather : Thread tape that improves vacuum uniformity, placed around the mold, is recommended by manufacturer [ref]. 34 Chapter 4. Manufacturing 4.3.2 Vacuum bag manufacturing During the realization of the present work, 16 vacuum bags were realized in total. All of them made on a metallic plane surface, 1 laminate needed to be made again because of a failure during curing process that made a crack inside the plate, the other one couldn´t be realized due to logistic issues. This section explains the general procedure to make a vacuum bag, as followed from the literature and previous experiences[REF]. Firstly, mold with laminate inside (without retiring paper shell) is placed upon the metallic table, a teflon layer big enough to cover all mold contour (and a bit more, just enough to ensure completely cover afterwards) is cut and placed between mold and table. (a) Metal table. (b) Teflon. Figure 4.20 One of the metal tables used (left) and first layer of teflon(right). Then, breather is located all along the mold contour, leaving about 3 cm between mold and breather. Paper shell is removed and a second layer of teflon is placed above the mold/laminate, making sure that it completely covers all the plate, mold, and leaving a gap for safety. (a) Breather positioning. (b) Teflon above mold. Figure 4.21 Breather (left) and second layer of teflon (right)). Once done, airwave is placed above the second teflon layer, sized just a bit larger than the layer, once again bearing in mind that it should cover all the laminate and mold. Next, chromate band is placed all around the contour of the plate, without removing the upper paper. 4.3 Vacuum bag 35 Figure 4.22 Airwave and chromate bands positioning. Hereinafter, plastic bag is cut, making sure that has a proper size that makes it possible not to fall short when positioning. Vacuum intakes are positioned, preferably over the breather band 4 . Paper chromate is removed and plastic positioned above, bearing in mind that is should be as tight as possible 5 . Finally, upper part of vacuum intakes are placed (just by turning on the thread of the lower part below the plastic, and removing the circular plastic derived from the hole just made). Vacuum intake is portrayed in figures 4.23a and 4.23b. (a) Lower intake. (b) Upper intake. Figure 4.23 Parts of the vacuum intake. 4If not possible, just 2 cm beyond mold corners 5Obviously assuring it could not break 36 Chapter 4. Manufacturing Figure 4.24 Vacuum bag prior to vacuum. Finally, wrinkles over the chromate band must be removed. For that, previously nylon tool used for stacking is applied again, in this case, over the chromate band. Once done, vacuum is made. In this final step, it is very important to check that there are no leaks, because air entrance in the vacuum bag can spoil the entire manufacturing, having to do it all over again. Checking is both visual and by hearing. Corners are the most risky areas. It is recommended,once vacuum is finished, to remove vacuum hose and wait for a few minutes and check again if there are no pressure leaks all over the bag. Figures 4.25 and 4.24 show vacuum bags prior and after vaccum is done Figure 4.25 Vacuum bag after vacuum. 4.4 Curing 37 4.4 Curing Composite materials require specific conditions in order to achieve its final state, wherein they go from being in manipulable state to final state, with desired mechanical properties and geometry. This proccess is known as composite curing, and can be done in many ways [ 17 ]. Current work makes use of the autoclave available at LERM (as seen in figure 4.26), capable of maintain a desired atmosphere with pressure and temperature monitored at all times. Figure 4.26 Autoclave placed at LERM. Pressure and temperature that the laminates require in order to cure are specific values (time-dependent) provided by manufacturer. Besides, autoclave has a self-vacuum capacity that helps reducing possible manufacturing flaws during curing proccess. Figure 4.27 show these magnitude variations throughout the entire process, taking a time of 6 hours (same for each laminate). 44 Chapter 5. Testing setup Preparation consisted mainly in the cleaning of the rust formed on the surface, for that it was used a sanding procedure, that can be seen in figures 5.3 and 5.2. Figure 5.2 Rust removal procedure by sanding. Note upper faces clean in contrast to side faces . Figure 5.3 Final cleaning using acetone . Finally, in order to ensure a proper alignment of load application, several measures where taken and some marks added to tooling so that it could ensure visual verification. Futhermore, a small metal piece was added to the fixed part in the bottom side, so that there would be no undesirable turns during tests. Figure 5.4 shows alingment and figure 5.5 shows final tooling assembly 5.2 Monitorization 45 Figure 5.4 Metal plate added in order to avoid turning. Alingment marks are also visible . Figure 5.5 Final complete assembly of tooling to test bench, with a specimen.. 5.2 Monitorization Two methods were used to take measures during testing, so that data dispersion could be reduced correlating results obtained : Strain gauges and videcorrelation. 46 Chapter 5. Testing setup 5.2.1 Strain gauges In order to take strain measures on the plate during tests, strain gauges were installed on the plates. A strain gauge is a sensor based in piezo-resistive effect. This effect makes that during gauge deformation, it´s internal circuit lenght variation prompts to a section wire variation, which leads to a electrical resistance variation itself[18]. Figure 5.6 Strain gauge schem, sensitive pattern is the responsible of measuring strain as it deforms. . Typical gauges consists in the stamping of a metal circuit bonded to an insulating and flexible base. It glues to the specimen via an adhesive. During specimen deformation, galge deforms too in the same manner. During this project, biaxial gauges were used, so both X and Y axis strains were measured See figure 5.6 Originally, several configurations were proposed in order to get the most data as possible. Blazquez et al [ 12 ] sugests that bests results fit were octhogonal pattern (figures 5.7a and 5.7). Despite this, the high cost of each gauge lead to make a new study in order to reduce number of gauges but trying to get the most and best data measures as possible. (a) (b) Figure 5.7 Original and octhogonal gauge distributions firstly proposed). This included considering gauge possible positions in each plate, trying to avoid nodal lines (figure 5.8) when possible1. 1As in nodal lines expected strain values are zero, so that the gauge gives no practical information in these positions but noise 5.2 Monitorization 47 Figure 5.8 Expected buckling modes of each plate, used to know where to put the gauges avoiding nodal lines [12]. Hence, a proper studio was done in order to analyze the best positions to get strain gauges with the minimum cost as posible but the most accurate data it could be adquired. Number of strain gauges reduced from 128 to only 16 at the end, just by placing each in specific locations where it is predicted that the most deformation will happen, so that [ 12 ] was used to get to these locations. For more detail ,the study can be consulted in annex B. Figure 5.9 Previous configurations presented, then discarted. Following the study, the final setups can be seen in figures ?? and 5.11 48 Chapter 5. Testing setup (a) (b) Figure 5.10 Final configurations presented, indicating optimal gauge location. (a) (b) Figure 5.11 Final configurations presented, indicating optimal gauge location. Gauges elected are given my manufacturer HBM, being biaxial type . Correction factor (provided by manufacturer) is 0.45. Installation Correct installation of the gauges require following a particular procedure, detailed below. 1. Sanding of plate surface where gauge will be located. This allows to a better adherency between gauge and plate.Figure 5.12. 5.2 Monitorization 49 Figure 5.12 Surface once sanding is applied. 2. Gauge position marking. It is done using the calibration tool, that consists of a vertical guide with a blade on its movable part. Supporting the plate on the table, and adjusting the height two marks are done, making a cross, figure 5.13. Figure 5.13 Cross marking where the gauge will be located. 3. Surface cleaning, using acetone. It is important so that contact surface between gauge and plate must be as clean as possible, avoiding particles or oils that can interfere during measurements 4. Gauge placement, with the help of tweezers, a piece of adhesive band is cut, and gauge placed on a clean cristal surface. Then, adhesive is placed upon the gauge, and gauge is placed aligned with the cross done before with the help of adhesive band. Figure 5.14. 50 Chapter 5. Testing setup Figure 5.14 Gauge before colocation . 5. Catalyst is applied (heated before, as it is stored in a fridge) below the gauge and over the colocation surface (Figure 5.15.) . Catalyst main function is to improve adherence. Then, adhesive is distributed in the same places. Once done, uniform pressure must be applied so that adhesive is distributed uniformly. Time of pressure is around 20-30 seconds. Done this, adhesive band is retired and gauge adhesion is checked. Figure 5.15 Gauge colocation and catalyst aplication. 6. Wire cutting. 4 wires are cut with a lenght of aproximately 30 cm, both ends are stripped and weaved. An electric dice is placed on one end (the one that will be connected to test bench) 7. A layer of paint stripper is then applied (Figure 6.17) and tin is deposited on the surface of the stripped wires. For this purpose, a soldering iron is used at a temperature of 300ºC and a coil of tin filament 5.2 Monitorization 51 from which material is taken to deposit in. The wires are left with a shiny surface (Figure 5.16.). Figure 5.16 Wires extremes once tinned. 8. To join gauges and wires, first it is neccessary to scracht the terminals of the gauges with a fiber glass pen (Figure 5.17.). Then a drop of tain is deposited in each terminal (Figure 5.18.). Figure 5.17 Fiberglass pencil aplication. 9. A piece of adhesive band is placed on the contour where the terminal is placed, to protect carbon fiber specimen from wire contact. Then wires are welded to the terminals, completing gauge and wire 52 Chapter 5. Testing setup joining. Figure 5.18 Tinning of the terminals. 10. Once joined, as a final step, a comprobation is needed. For this purpose, a multimeter is used in which it is necessary to place in its Ohms reading and in continuity mode. If it had been done correctly, touching the two ends of the cable that are joined, a continuous beep should be heard, while when applying the terminals to any other combination (always in continuity) should show a reading of 120 Ohms 2 A gauge once properly installed can be seen in Figure 5.19. Figure 5.19 Gauge once installed. Adhesive bands are added for further protection during subsequents steps as painting. 2Approximate as we consider gauge factor 5.3 Correlation system 53 5.3 Correlation system The digital image correlation system (Digital Image Correlation) developed by Correlated Solutions provides an optical measurement of the deformation on the surface of the object in question by observing the variation of the gray pattern on the surface as it deforms. The validity of this method can be checked by comparing the results with those provided by a finite element analysis[ 19 ]. For data acquisition, two Teledyne Dalsa cameras connected to a computer and placed at a certain distance from each other are used as hardware (as seen in figure 5.20). These camera are connected via wire to a computer and arranged at a certain distance from the panel, about 1 m, and not oriented perpendicularly to it, as in the case of the perpendicular to the panel. Otherwise, it would not be possible to appreciate the deformation of the plate during the test. Figure 5.20 Camera and lighting setup. The system works with master-slave logic, with the master chamber being the leftmost one in the image. The software used is Vic-Snap, provided by Correlated Solutions, which is used to take pictures of the panel. The program triangulates the position of the cameras, as well as eliminating possible distortions of the lenses. Finally, any bias in the measurement is eliminated and a coordinate system is set up and a threedimensional coordinate system is established on the surface of the object to be tested [ 19 ] . This process is critical, since with an erroneous calibration the digital correlation system does not guarantee that the results will be correct. A sample of images taken for calibration is presented in figure 5.21. During the first few times of the set-up of the system, the lack of synchronization between the cameras and the image capturing software was a constant. Several solutions were tried, such as lighting modification, focus and aperture adjustments of the cameras (mainly the master), and finally, digital gain adjustment by software, which turned out to be the most effective. Once calibrated and checked, no further changes were needed. Note that fine tuning of the system took a few days until it was completely finished, taking a substantial amount of time of the work. Calibrated images can be seen in figure 5.21. 60 Chapter 5. Testing setup 5.4 Test bench In order to make the tests, the test bench used was INSTRON 4482; that can be seen in figure 5.30 with a load cell of 40000 N. This load cell gives displacements measures during the testing process, to that, although not very accurate, can be used to correlate with videocorrelation and strain gauge data. Figure 5.30 INSTRON 4482. For data acquisition from the gauges, there are two yellow connectors with four available docks each, which can be seen in the figure mentioned above on the left, connected to a series of electronic cards that send the data to the computer. Before performing the tests, channels are checked so that strain measures obtained don´t show great distorsions, reseting the data adquisition when necessary 6 Test results In this chapter it is explained and developed the whole testing process of the specimens, as well as the preliminary results (visual and force versus displacement) obtained. In addition, the problems that arose during the whole testing process and the final state of the specimens once they have been tested are shown. A total of 14 specimens of the original 16 were tested, not being able to made two of them because of logistical issues (lacks of material and time during manufacturing process). About test performing, every of them were performed in the same manner, with a displacements control with 0.6 mm/s test velocity. Except for the first one (that was performed at 0.5 mm/s), this values proved a enough quick testing with a significant batch of photos taken that allows to show pannel evolution during the process. Testing is stopped once pannel brokes, either seen in the load/displacement graph or by listening. 6.1 Specimen testing As a general comment, an initial period of plate resettlement is observed during all tests, corresponding to the initial part of the Load vs. Displacements curve. The test itself (and the data availability) begins at the moment when the curve takes on a linear character. 6.1.1 PP0001 This laminate, formed by 12 layers with the same direction (that is the direction of the compression load) and being plane, was the first to be tested. In figure 6.1, a clear change in the slope of the curve can be seen at 1000 N-0.38 mm, where even the load experiments a small decline. This is due to pannel settlement in the tooling, starting the proper testing at this moment. Approximately at 2500 N-1.30 mm, a change in the slope of the curve is observed, so the stiffness increases getting a new constant value. 61 62 Chapter 6. Test results Figure 6.1 Load vs displacements - PP0001 . Pannel broke at 28500 N-2.36 mm, not showing any major defects until the crack. The crack propagated quickly creating a high decrease of the bearing load. Cracks reported have the direction of the fiber (figure B.1). Figure 6.2 Specimen tested - PP0001 . 6.1.2 PP0002 This laminate, being the same kind of the former one, showed the same behaviour (as seen in figure 6.3), without pannel settlement and a higher failure point (30000 N-1.68 mm). Identical 0ºcracks appeared at failure. 6.1 Specimen testing 63 Figure 6.3 Load vs displacements - PP0002 . As seen in figure 6.4, both laminates show same stiffness and strength, being the main difference founded at the point where the test starts, which is most likely due to the settlement of the plate. Figure 6.4 Load vs displacements - PP0001 and PP0002 . 6.1.3 PP9001 This pannel has a perpendicular fiber orientation, being plane.In figure 6.5, Slope of the Load vs displacements curve seems to be almost constant, so the stiffness did not present any change during the test, except for the first part where laminate looks to be settling. 64 Chapter 6. Test results Figure 6.5 Load vs displacements - PP9001 . In this case, failure appears at 12800 N-1.79 mm. Crack propagated along the direction of the fiber, making the upper part of the tooling to fall down (without major consequences). The tested specimen with the crack can be seen in figure 6.6. Figure 6.6 Specimen tested - PP9001. . 6.1.4 PP9002 This pannel broke at only 4500 N-0.92 mm, being much more weaker as was expected, as seen in figure 6.7. No change in slope was reported. 6.1 Specimen testing 65 Figure 6.7 Load vs displacements - PP9002 . Despite being unknown, possible imperfections during manufacturing process could have made the pannel to not have the expected behaviour. Same type of crack (longitudinalfiber direction) was reported. As seen in figure 6.8, settlement affects as well, also second specimen starts almost the smae as first, although failing before reaching linear behaviour. Figure 6.8 Load vs displacements - PP9001 and PP9002 . 6.1.5 PP4501 This laminate, composed with 12 layers with a fiber direction of 45º, figure 6.9 showed constant stiffness from begining of the test to 7630 N-0.85 mm, where a small step appears, then showing a slight change in stiffness, decreasing. 66 Chapter 6. Test results Figure 6.9 Load vs displacements - PP4501 . Laminate broke at 8800 N-1.12 mm, but already during 7000 N fiber failure could be heard. Crack appeared at the upper edge, as shown in figure 6.10 Figure 6.10 Specimen tested - PP4501 . 6.1.6 PP4502 In this laminate, a not proper settling on the tooling was reported, due to a bad fit in the lower left corner of the plate in the groove. This caused, as seen in 6.11, an interval of time during testing where the plate didn´t supported structural load (the first small step where it starts to get load, but then it decreases drastically), until it finally settled, so that proper testing began. On the other hand, two different zones of stiffness are easily visible, being the change at 2800N-2.02 mm. 6.1 Specimen testing 67 Figure 6.11 Load vs displacements - PP4502 . Specimen broke at 7600 N-2.56 mm. Crack propagation spread to the left upper edge. As seen in figure 6.12, in this case plate settlement is an issue as well, showing both plates same stiffness in only some part of curve. Figure 6.12 Load vs displacements - PP4501 and PP4502 . 6.1.7 PPISO01 This cuasi-isothropic laminate showed in figure 6.13 a clear postbuckling regime change at 13400N-1.18mm, with a change in stiffness before failure. A very small step can be seen at 1700 N-0.28mm. 68 Chapter 6. Test results Figure 6.13 Load vs displacements - PPISO01 . Specimen broke at 14400 N-1.27 mm. No cracks were visible. 6.1.8 PPISO02 As commented before, this laminate couldn´t be done due to logistical reasons, so no results are available. 6.1.9 H1V3001 In this laminate, imperfections are introduced with m=1 and n=3, being 1 semiwave in the X direction, and 3 semiwaves in the Y direction. Up to 2 changes in stiffness are reported in figure 6.14, the first one at 8300 N-1 mm, then stiffness decreases, the second one at 16000 N-2.82 mm. Figure 6.14 Load vs displacements - H1V30001 . Plate broke at 18000 N- 3.22 mm, prior to failure individual fiber failures could be heard. Crack was not visible at first, then was reported on the left upper corner, being internal. 6.1 Specimen testing 69 6.1.10 H3V1001 As the former one, a preformed state is imposed in this laminate, in this case m=1, n=3.No change in stiffness neither buckling mode is seen during testing, as shown in figure 6.15 Figure 6.15 Load vs displacements - H3V10001 . Specimen fails at 25000 N-1.7 mm, previously individual fiber breaks are seen and heard, but no more load is supported between first fiber failures and final collapse. A longitudinal crack is seen in figure B.2, following fiber direction. Figure 6.16 Specimen tested - H1V30001 . 6.1.11 H1V39001 As the former one, a preformed state is imposed in this laminate, in this case m=1, n=3. Just a very slight change in stiffness before failure is reported in figure 6.17, with no turning points. 7 Results In this chapter, data adquired during testing is discussed. Both strain gauge and videocorrelation results are shown and analyzed. Specific conditions during the testing (where the results are shown) are displayed on the marks in figures of the previous chapter. 7.1 Strain gauge results The purpose of the placement of the gauges was to measure the deformations in different zones chosen a priori for the correlation. Once installed and checked, data acquisition was carried out during the test, taking into account previous procedures to try to avoid possible failures. It is also necessary to indicate the relationship between the sign of the band measurement and the deformation of the panel itself, in order to facilitate the determination of a specific buckling mode. On the one hand, a positive deformation implies an elongation of the region surrounding the gauge, which leads to the appearance in that area of a small half-wave in each direction. Following the sign criterion defined in this chapter, the points of the aforementioned half-wave must be displaced in each direction. The above-mentioned half-wave must be displaced in a negative direction so that the gauge region is elongated, bearing in mind that the face of the panel where the strips are connected is the one opposite to the one facing the chambers. Similarly, a negative measurement of the deformation corresponds to the manifestation in the area surrounding the gauge of a half-wave in each direction, whose points protrude in the positive direction of the displacement measured by the cameras of the videocorrelation system. An scheme of the explained can be seen in figures 7.1a and 7.1b (a) Semiwaves with negative displacement. (b) Semiwaves with positive displacement. Figure 7.1 77 78 Chapter 7. Results 7.1.1 PP0001 Two main points are clearly observed in figure 7.2, where Load vs microstrains curves are shown. At 90º direction, buckling evolution is clearly visible, where strains reach a maximun and then decreases. At 0º, it seems that distortions on the measurements appear from 15000 N. Figure 7.2 Load vs µε - PP0001 . 7.1.2 PP0002 In this case, a lot of distortions seems to have appeared (as seen in figure 7.3) on the measurements, both 0º and 90º. Even so, the same type change in strains as before can be noticed in 90ºand 0º Figure 7.3 Load vs µε - PP0002 . 7.1.3 PP9001 Results obtained can be seen in figure 7.4, whereit is clearly noticeable the stiffness change due to buckling in the direction of compression (90º), being constant in the 0ºdirection1 1Note that now directions are the one on the plate, not the 0ºdiscussed before 7.1 Strain gauge results 79 Figure 7.4 Load vs µε - PP9001 . 7.1.4 PP9002 Results obtained can be seen in figure 7.5, where it is clearly noticeable the stiffness change due to buckling in the direction of compression (90º), being constant in the 0ºdirection2 Figure 7.5 Load vs µε- PP9002 . 7.1.5 PP4501 In figure 7.6, results showed distortions during the first part of the test, with similarities in the slope of the strain curves for both directions (but not the same exactly). It is seen the change in the slope on the final part of the test. 2Note that now directions are the one on the plate, not the 0ºdiscussed before 80 Chapter 7. Results Figure 7.6 Load vs µε - PP4501 . 7.1.6 PP4502 Strain curves, despite distortions, looked very similar (figure 7.7) in the two axis. Besides, stiffness change is visible, with the final part of the curve more free of noise. Figure 7.7 Load vs µε- PP4502 . 7.1.7 PPISO01 Strain curves show in figure 7.8 a lot of noise after buckling mode change. Besides, the small step commented before can be seen as a great distorsion in the gauge measurements 7.1 Strain gauge results 81 Figure 7.8 Load vs µε - PPISO01 . 7.1.8 H1V30001 In this plate, 0ºstrain gauge data adquisition failed during testing (figure 7.9). 90ºshows a change in stiffness, then, at 8000N, plate seems not admiting more structural load, as strain holds constant, but finally increases after 1400 microstrains until failure Figure 7.9 Load vs µε - H1V30001 . 7.1.9 H3V10001 In this pannel most strain seems to be in compression direction (90º), as seen in figure 7.10. Relation between Loads-Strain seems to be proportional until 10000 N, where buckling stars to act. 82 Chapter 7. Results Figure 7.10 Load vs µε - H3V10001 . 7.1.10 H1V39001 Both 90ºand 0ºhave similar behaviors (figure 7.11), showing distortions in the measures taken by the gauges. Figure 7.11 Load vs µε - H1V39001 . 7.1.11 H1V34501 As seen in figure 7.12, longitudinal strain (90º) is bigger than perpendicular. Both 0ºand 90ºshows a transition between buckling modes as there is a region were no more structural load is hold by the plate, then it rises again. 7.1 Strain gauge results 83 Figure 7.12 Load vs µε - H1V34501 . 7.1.12 H3V14501 In this case, in figure 7.13 both axis seem to show a symmetrical behavior, being both stiffnesses almost constant. A distorsion ought to a translation of the strain gauge during testing seems to be the main cause of the flat zone of the green curve. Figure 7.13 Load vs µε - H3V14501 . 7.1.13 H1V3ISO01 In this case, figure 7.14 show that both axis had an almost constant stiffness, being the longitudinal axis (0º) the least rigid. 84 Chapter 7. Results Figure 7.14 Load vs µε - H1V3ISO01 . 7.1.14 H1V3ISO01 In this case (figure 7.15), it is the perpendicular axis (0º) the one who shows more strain. It is noticed a change in stiffness in both axis, where the slope of the Load-strain curve changes (approximately at 24000 N) Figure 7.15 Load vs µε - H1V3ISO01 . 7.2 Videocorrelation results 85 7.2 Videocorrelation results Data that are about to be explained in this section correspond to several points during the evolution of each pannel testing. As data obtained is a continuous and because of his proper nature the behavior of the postbuckling process is not stationary, in order to simplify the analysis, just what were considered the most interesting points are going to be discussed. Results are present in two ways, 3D plot and a heatmap over the real surface, wherein red colors indicate the most steepest gradient. It is important to explain that several correlation algorithms can be used with Vic-3D software, as during the postprocess of the results some data looked not to ve very reliable, by changing the options in the software the data obtained ended up looking much closer to what was intended at the cost of losing some resolution. Although some other changes were applied, the main one consisted on changing correlation method to the "least squares" option available, which proved to improve results largely, as seen in figure 7.16. (a) Before applying algorihtm. (b) After algorihtm. Figure 7.16 Results improved due to changes in postproccessing method. 92 Chapter 7. Results Strain εxx In this case, it is clear the formation of two isolated regions with larger strains on the sides of the plate, and a more-homogeneus zone on the center. (a) (b) Figure 7.32 εxx at F=2500 N. (a) (b) Figure 7.33 εxx at F=15000 N. (a) (b) Figure 7.34 εxx at F=30000 N. Strain εyy It can be seen a center region with the most strains, reaching negative values at edges 7.2 Videocorrelation results 93 (a) (b) Figure 7.35 εyy at F=2500 N. (a) (b) Figure 7.36 εyy at F=15000 N. (a) (b) Figure 7.37 εyy at F=30000 N. Strain εxy It looks that strains tend to go to the corners of the plate, being two of the corners the higher values and the ones opposing them, the lowest. 94 Chapter 7. Results (a) (b) Figure 7.38 εxy at F=2500 N. (a) (b) Figure 7.39 εxy at F=15000N. (a) (b) Figure 7.40 εxy at F=30000 N. 7.2 Videocorrelation results 95 7.2.3 PP9001 Displacements In this pannel, mode 1 is clearly the dominant, with two separate areas corresponding to the maximum and the minimum in displacements. No change in buckling mode is seen. (a) (b) Figure 7.41 Displacements at F=5000 N. (a) (b) Figure 7.42 Displacements at F=12000 N. Strain εxx εxx shows a particular distribution with two circular areas in the upper side and one area in the lower side, with the maximum and the minimum, respectively. (a) (b) Figure 7.43 εxx at F=5000 N. 96 Chapter 7. Results (a) (b) Figure 7.44 εxx at F=12000 N. Strain εyy It can be seen two different areas, with a maximum and a minimum. (a) (b) Figure 7.45 εyy at F=5000 N. (a) (b) Figure 7.46 εyy at F=12000 N. Strain εxy It looks that strains tend to go to the corners of the plate and to the centre of the edges, alternating highest and lowest values. 7.2 Videocorrelation results 97 (a) (b) Figure 7.47 εxy at F=5000 N. (a) (b) Figure 7.48 εxy at F=12000 N. 98 Chapter 7. Results 7.2.4 PP9002 Due to early failure of the plate, most of the data obtained don´t allow for sufficiently useful conclusions to be drawn. Displacements As pannel didn´t stood for enough bearing load, the only conclusion is that mode 1 (as in the former PP9001) seems to be the dominant. This can be seen in figure 7.49. (a) (b) Figure 7.49 Displacements at F=2000 N. Since the data obtained are not considered sufficiently valid, the plate deformations are not shown. 7.2 Videocorrelation results 99 7.2.5 PP4501 During the testing of this plate, image adquisition software had a failure, making imposible to get enough images to perform a proper analysis. Results obtained were in the form of figure 7.50 Figure 7.50 Displacements obtained by software, crearly not correct.. 100 Chapter 7. Results 7.2.6 PP4502 Displacements In this pannel, it seems that buckling regime is dominated by a combination of modes, being modes 1 and 3 influence visible due to the inclined behavior of the waves formed. (a) (b) Figure 7.51 Displacements at F=1000 N. (a) (b) Figure 7.52 Displacements at F=3000 N. Strain εxx εxx shows a distribution very homogeneous, with a little redistribution of strains in and inclined shape during buckling mode 7.2 Videocorrelation results 101 (a) (b) Figure 7.53 εxx at F=1000 N. (a) (b) Figure 7.54 εxx at F=3000 N. Strain εyy The distribution of strains holds almost the same in all the plate during the testing, just with a little inclined shape in the direction perpendicular to the fiber (a) (b) Figure 7.55 εyy at F=1000 N. 108 Chapter 7. Results (a) (b) Figure 7.72 εxy at F=5000 N. (a) (b) Figure 7.73 εyy at F=16000 N. 7.2 Videocorrelation results 109 7.2.9 H3V1001 During postprocess of the data obtained during testing, it was discovered that the painting coat given to the pannel was not optimal (mostly, because brightness), so that data obtained have less quality than in other plates. Displacements In this case, it seems that the pannel maintains it´s preformed state, not showing any postbuckling state nor stiffness change. (a) (b) Figure 7.74 Displacements at F=10000 N. (a) (b) Figure 7.75 Displacements at F=24000 N. Strain εxx εxx tends to decrease and uniforms over the pannel 110 Chapter 7. Results (a) (b) Figure 7.76 εxx at F=10000 N. (a) (b) Figure 7.77 εxx at F=24000 N.. Strain εyy εyy behaves as εxx, becoming uniform over the pannel and decreasing (a) (b) Figure 7.78 εyy at F=10000 N. 7.2 Videocorrelation results 111 (a) (b) Figure 7.79 εyy at F=24000 N. Strain εxy εxy holds almost uniformly during all the testing. (a) (b) Figure 7.80 εxy at F=10000 N. (a) (b) Figure 7.81 εyy at F=24000 N. 112 Chapter 7. Results 7.2.10 H1V39001 It was noticed that brightness during testing decreased the quality of the results, especially just before final moments before breaking. Displacements In this case, a very quick change in buckling appears, turning from the former preformed shape to a form where mode 1 seems to dominate (a longitudinal pattern with 3 semiwaves) (a) (b) Figure 7.82 Displacements at F=1000 N. (a) (b) Figure 7.83 Displacements at F=3000 N. (a) (b) Figure 7.84 Displacements at F=12000 N. 7.2 Videocorrelation results 113 Strain εxx εxx tends to get an almost uniform behavior, with some points in the upper centre with minimum values (a) (b) Figure 7.85 εxx at F=3000 N. (a) (b) Figure 7.86 εxx at F=12000 N.. Strain εyy εyy changes becoming a 3 semiwave longitudinal pattern (a) (b) Figure 7.87 εyy at F=3000 N. 114 Chapter 7. Results (a) (b) Figure 7.88 εyy at F=12000 N. Strain εxy εxy seems to gather in two hills in the centre, with maximum and minimum (a) (b) Figure 7.89 εxy at F=3000 N. (a) (b) Figure 7.90 εyy at F=12000 N. 7.2 Videocorrelation results 115 7.2.11 H1V34501 Displacements A clearly change on buckling regime is observed, so that the pannel behaves changing from the initial preformed to a center isolated hill, dissapearing the previous waves (a) (b) Figure 7.91 Displacements at F=5000 N. (a) (b) Figure 7.92 Displacements at F=20000 N. (a) (b) Figure 7.93 Displacements at F=25000 N. 116 Chapter 7. Results Strain εxx εxx shows a change during buckling regime change, from preimposed state appearing a hill in the centre that then dissappears and then tends to grow in the sides. (a) (b) Figure 7.94 εxx at F=5000 N. (a) (b) Figure 7.95 εxx at F=20000 N.. (a) (b) Figure 7.96 Displacements at F=25000 N. Strain εyy εyy behaves firstly increasing on a hill in the centre, then going to the sides of the pannel 7.2 Videocorrelation results 117 (a) (b) Figure 7.97 εyy at F=5000 N. (a) (b) Figure 7.98 εyy at F=20000 N. (a) (b) Figure 7.99 Displacements at F=25000 N. Strain εxy εxy behaves increasing mostly in the corners 124 Chapter 7. Results Strain εyy εyy shows a central zone with slightly higher values (a) (b) Figure 7.115 εyy at F=15000 N. (a) (b) Figure 7.116 εyy at F=34000 N. 7.2 Videocorrelation results 125 Strain εxy εxy behaves gathering mainly at the corners (a) (b) Figure 7.117 εxy at F=15000 N. (a) (b) Figure 7.118 εyy at F=34000 N. 126 Chapter 7. Results 7.2.14 H3V1ISO01 Displacements No behavior change is observed, maintaning the plate the preformed shape. (a) (b) Figure 7.119 Displacements at F=10000 N. (a) (b) Figure 7.120 Displacements at F=24000 N. 7.2 Videocorrelation results 127 Strain εxx εxx seems to form a "cross" pattern with maximum on the corners and minimums on the center (a) (b) Figure 7.121 εxx at F=10000 N. (a) (b) Figure 7.122 εxx at F=24000 N.. 128 Chapter 7. Results Strain εyy εyy shows a clear 3 semiwave pattern in the X axis (a) (b) Figure 7.123 εyy at F=10000 N. (a) (b) Figure 7.124 εyy at F=24000 N. 7.2 Videocorrelation results 129 Strain εxy εxy behaves gathering mainly at the corners (a) (b) Figure 7.125 εxy at F=10000 N. (a) (b) Figure 7.126 εyy at F=24000 N. 8 Conclusions On ce testing campaing is concluded, and all data are properly obtained, sorted out, and disposed, an overall analysis is done in this chapter. To begin with, it is worth remembering the main goal proposed at the beginning of this work: to study the postbuckling regime of composite preformed plates with several imperfections intentionally introduced. In view of the results obtained, it can be said that this goal has been properly achieved to the greatest extent, so that many conclusions have been drawn that allow to contribute to the knowledge of postbuckling regime. The conclusions drawn are detailed in the first point of this chapter. In the second point, several issues that were discovered during the process are described, with proposed improvements. Future researches are detailed in point 3. 8.1 Conclusions The conclusions obtained during the testing process of each curved laminate are detailed below • H1V30001: Comparing to flat case, changes in stiffness are reported, also a change in buckling mode from the initially imposed mode. • H3V10001: No change in stiffness is reported, showing the pannel a buckling behaviour similar to the preformed shape. • H1V39001: Changes in stiffness comparing to the flat case, being higher, then decreasing. Also it didn´t hold the imposed shape, changing the buckling to a similar one of the flat cases. • H3V19001: No change in stiffness. No information about buckling modes was available because of videocorrelation system failure. • H1V34501: Changes in stiffness comparing to the flat case, being lower in this case. Neither holding the preformed shape in buckling mode, developing a new buckling mode. Because of poor quality of videcorrelation in planar specimens buckling modes couldn´t be related • H3V14501: It held preformed shape, not changing buckling mode. Lower stiffness than flat and H1V345 cases • H1V3ISO01: Clear change in buckling mode, not holding pre-imposed state. Also a considerable increase in stiffness comparing to flat case. This panel proved to be the most structurally efficient of all • H3V1ISO01: Held pre-imposed buckling mode, and showed a lower stiffness than planar specimens In summary, it can be drawn as a conclusion the invariance of the buckling mode of the laminates with the configuration H3V1 (m=3, n=1), always holding the pre-imposed shape. In the case of laminates with configuration H1V3 (m=1, n=3), all of them show a change in buckling mode, not holding pre-imposed shape, and developing a mode 1 dominating shape. Worth mentioning the case H3V19001 where a change in buckling mode has been clearly observed. This case is specially remarkable because of the different buckling change mode (comparing with others) that showed, with 2 semiwaves in the direction of load and 0 in the perpendicular where the original were 3 in the perpendicular and 1 in the load direction. When talking about stiffness, generally preformed shape with H1V3 pannels have showed a better behaviour even than the flat 131 132 Chapter 8. Conclusions ones, and it is worth mentioning in particular the H1V3ISO01 laminate, that showed the best stiffness and strenght behaviour, which is demonstrated the practical utilities in this particular case of the use that can be made of postbuckling behaviour. 8.2 Issues and posible improvements Based on the results, it is legitimate to think about continuing the line of investigation followed, taking into account the flaws and mistakes made throughout the chain. Several issues were reported, so that some improvements are proposed as well. 8.2.1 Manufacturing Due to the inherent nature of the defects caused by the manufacturing process, although these can be minimized, there will always be errors that appear during this step, so they should not be taken into account. The possible improvements that can be proposed in these cases are due to logistics (although not being exactly related with manufacturing process), trying to improve the availability of materials and quantify the manufacturing times of each laminate, something critical in the case of manufacturing a large number of test pieces, as is the case. On the other hand, the use of a "countermold" is proposed in order to ensure a proper thickness 8.2.2 Test setup and campaing In this case, it should be emphasized that the boundary conditions set out in section 3 are not exactly those reproduced during the tests, but rather an approximation due to the use of available tooling. Furher adaptations in numerical calculations that take them into account would be appropriate to make later comparisons. On the other hand, during some test it was reported the difficulty for the introduction and extraction of the speciments in the lower left corner of the plates.This could have affected the test in some extent, modifying the values obtained as free lateral edge boundary condition is no longer applicable. A readjusment of the lateral sides of the test tooling would be recommended, either redo or adapt it, for the most exact replication of boundary conditions as possible, so that knowing where the real testing begins (as it was explained in chapter 6, settling of the plates in the initial step of testing has been an issue). 8.2.3 Results The harder constraint found during the performance of the current project resulted in the videocorrelation system performance . As was reported in chapters 5 and 7, data adquisition during testing required several changes in methodologies in order to get the correct results, although during the process of testing some data was lost due to failing of the software. Besides of the methodology implemented (mate coating, use of reel, changes in videocorrelation algorithm and brightness control), a proper single assembly of the setup with the same configuration that has demonstrated to be the optimal, also doing all the campaign test at once, all with the same characteristics to try to avoid possible noise, would be ideal in order to consider future developments. 8.3 Future researches The line of research that is followed during this project can and should be exploited if an exhaustive analysis of the post-buckling regime of composite plates is to be carried out. The study in particular of the situations in which there has been a change of regime to the pre-imposed shape is the most interesting, in particular the H1V3ISO case, where stiffness changes should be studied in more detail. Also other configurations, not only in fiber directions, but also in preformed shapes (and also materials and structural configurations, such as infusion and stiffened laminates) should be proposed in order to study postbuckling and its effects. Results and conclusions obtained throughout the entire project seems to be promising and invite to continue delving into the line of research, especially to the extent of obtaining results that allow the development of practical applications in the use of composites. Appendix A Scraps In this append, a proper study for scraps is done, in order to optimize material expenses. Work here is written in spanish ought to some measurements were done with the collaboration of colleagues at LERM, so that comunication was easier. A.1 Estimaciones de gasto Estimacion gasto de material: •Laminados a 0 y a 90 Estimación de gasto: – 0º: (2 planas+2 ondas) configuraciones distintas * 2 (simetria) *6 laminados por placa + 4*2*1 (Laminado ISO) = 56 apilados – 90º: (2 planas+2 ondas) configuraciones distintas * 2 (simetria) *6 laminados por placa +4*2*1 (Laminado ISO) = 56 apilados 540 mm para 9 apilados: 2800 mm = 3,36 m (para 0) 540 mm para 9 apilados: 2800 mm = 3,36 m (para 90) Valido para ambas •Laminados a 45 Estimación de gasto: 45º: 4 configuraciones distintas * 6 laminados por placa * 2(simetria) +4*2*4 laminados por placa ISO = 80 apilados 1200 (1145,42) mm para 15 apilados : 6,2 (6109,44) para 80 apilados Total = 2*3,4 + 6,2 = 13 m aprox (contando errores de uniformidad) REDONDEADO Tiras: Para 0-90: 1085 (+5mm de error) Da para 18 apilados En total: 6 tiras Para 45: 1150 (+5mm de error) Da para 15 apilados En total 5: tiras Desperdicio de material: Por 0-90: 0,41 % Por 45: 17,87 : podría mejorarse hasta un orden de magnitud si se cortara todo el rollo de una (desperdicio debido a garantizar la continuidad de la fibra) 133