Dissertations / Theses on the topic 'Liquid composite molding'

Thesis (Ph. D.)--Ohio State University, 2004.
Title from first page of PDF file. Document formatted into pages; contains xix, 245 p.; also includes graphics. Includes bibliographical references (p. 233-245).

As the aerospace industry continues to incorporate composites into its aircraft, there will be a need for alternative solutions to the current autoclaving process. Liquid composite molding (LCM) has proven to be a promising alternative, producing parts at faster rates and reduced costs while retaining aerospace grade quality. The most important factor of LCM is controlling the resin flow throughout the fiber reinforcement during infusion, as incomplete filling of fibers is a major quality issue as it results in dry spots or voids. Void formation occurs at the resin flow front due to competition between viscous forces and capillary pressure. The purpose of this work is to characterize capillary pressure in vacuum infusion, and develop a model that can be incorporated into flow simulation. In all tests performed capillary pressure was always higher for the carbon fiber versus fiberglass samples. This is due to the increased fiber packing associated with the carbon fabric. As the fabric samples were compressed to achieve specific fiber volumes an increase in capillary pressure was observed due to the decrease in porosity. Measured values for capillary pressure in the carbon fabric were ~2 kPa, thus the relative effects of Pcap may become significant in flow modeling under certain slow flow conditions in composite processing.

Global competition is pushing the composites industry to advance and become more cost effective. Liquid Composite Molding or LCM is a family of processes that has shown significant promise in its potential to reduce process times and cost while maintaining high levels of part quality. However, the majority of research and information on composite processes have been related to prepreg-autoclave processing which is significantly different than LCM. In order for LCM processes to gain large scale implementation, significant research is required in order to model and simulate the unique nature of the resin infusion process. The purpose of this research is to aid in the development of in situ void measurement and characterization during LCM processing, particularly for carbon fiber composites. This will allow for the gathering of important empirical data for the validation of models and simulations that aid in the understanding of void formation and movement during LCM. For such data to be useful, it needs to include details on the formation, mobility and evolution of the void over time during infusion. This was accomplished by creating a methodology that allowed for in situ images of voids to be captured during the infusion process. A clear mold was used to visually monitor infusions during RTM with UV dye and lighting to enhance contrast. Consecutive images were acquired through the use of macro lens photography. This method proved capable of yielding high quality images of a variety of in situ voids during infusions with carbon fiber composites. This is believed to be the first instance where this was accomplished. A second methodology was then developed for the analysis of the collected images. This was done by using ImageJ software to analyze and process the acquired images in order to identify and characterize the voids. Success was found in quantifying the size and circularity of a wide range of micro and macrovoids in both a satin weave and double bias NCF woven fabrics. To facilitate the burden of collecting large amounts of data, this process was made to be automated. A user generated macro script could be applied to large sets of images for rapid processing and analysis. This automated method was then evaluated against manually processed images to determine its overall effectiveness and accuracy as tool for validating void theory.

Two objectives have served as the basis for the present work. The first objective was to experimentally verify the output of a Monte Carlo-based model on particle deposition in porous structures (i.e. fiber preform) during the resin transfer molding (RTM) process. This model is expected to have a positive impact on process optimization and reduce costs by enabling theoretical prediction of how particles are deposited when nanoparticle filled resin is used in RTM. The second objective was to produce a multiphase composite where the epoxy (EP) matrix makes use of the mechanical enhancement that cellulose nanofibres (CNF) can impart when incorporated into the liquid phase of the used polymer matrix. The study was conducted by production of the CNF-filled EP matrix, and by using this to produce multiphase composites out of glass- as well as carbon fibers. The RTM was carried out in collaboration with CSI Composite Solutions and Innovations Oy (Vilppula, Finland). Characterization techniques, including Raman spectroscopy, optical and electron microscopy were used to investigate the microstructure and for assessment of the CNF distribution in the produced composites. These observations were qualitatively compared with the output from the proposed model to evaluate its applicability. EP/CNF nanocomposites (i.e. the consolidated resin) were evaluated by tensile test to investigate the influence of CNF on the mechanical properties of epoxy. Three-point bending tests (ISO 14125) was performed on the multiphase composite to evaluate the impact of CNF-inclusion in the matrix. Obtained results indicate that the model is consistent with the process by which the CNF are deposited in RTM, as both the model and experiment show that the CNF are accumulated in the upper layers (injection side) of the preform. However, work remains to be done for the model to fully comply with specific aspects of the used reinforcement in RTM (e.g. pore size and geometry of the used fiber reinforcement), and thus predict the correct deposition profile and penetration depth of the CNF. The mechanical tests showed that the incorporation of CNF in the epoxy provides a composite with enhanced performance relative a reference without the CNF. The percentage change relative the reference composite was up to 22 % in flexural strength and 25 % in flexural modulus. It is suggested that the increase in mechanical performance is a result of CNF accumulation due to filtering in the fiber preform.

2016 - 2017
The research activity was devoted to the study of the composite materials manufacturing processes. In particular, the liquid composite molding (LCM) processes were the object of the performed study. In recent years LCM processes have gained a widespread diffusion in different industrial fields, from civil to automotive and aerospace due to their several advantages compared to the conventional autoclave processes. However, some disadvantages related to a not uniform preform impregnation due to a local variation of the preform permeability, fibers bundles misalignment, that would results in dry zones or matrix richer areas, affect the LCMs limiting their usage in industrial full scale. Other limits are due to a limited pressure driving force as well as a reduced pressure compaction influencing the final volume fraction achievable with detrimental effects on the mechanical properties of the composite material product. A more deep knowledge of the phenomena involved in the manufacturing of the composite materials are required to implement proper control action on the parameters (e.g. pressure, resin flow rate, thermal cycle as well as inlet/vent locations) to optimize the process. In order to improve the impregnation of the preform and reduce the time required to fully fill the mold cavity an in-line microwave preheating system was developed. The aims was to couple a microwave generator upstream the LCM mold to heat up the resin prior the entry into the mold. Indeed, the temperature increasing reduces the liquid viscosity allowing the resin to flow more freely through the dry preform. To perform a thorough study on the effectiveness of the proposed approach a laboratory scale apparatus for liquid composite molding processes was designed. The system was instrumented with ad-hoc designed sensors to monitor the resin flow during the process. Cheaper dielectric sensors are designed, produced and installed on the mold. A numerical model was also developed to simulate the resin flow through the fibers preform. The numerical model proved to able to deal with the dual-scale nature of the textile preform commonly used in the LCMs, that are characterized by two different regions (inter- and intra-tow) with different values of permeability. The numerical outcomes were also used to validate the data obtained from the dielectric sensors. They demonstrated to be able to monitor the both the impregnation and the saturation of the fiber preform. The developed microwave heating system proved to be effective to both reduce the total infusion time as well as improve the wetting of the fibers, achieving a more uniform impregnation with a limited amount of residual voids.[edited by Author]
XVI n.s. (XXX ciclo)

One portion of this work was concerned with injection molding pregenerated microcomposites composed primarily of poly(ethylene terephthalate) (PET) as the matrix and HX1000 as the thermotropic liquid crystalline polymer (TLCP). Several factors were examined to maximize the mechanical properties of these composites, including injection molding temperature, matrix viscosity, and nozzle tip exit diameter. In addition, concentrated strands of HX1000/PET (50/50 wt%) were diluted using both an injection molding grade of PET and an injection molding grade of PBT. From this work, it was determined that the best mechanical properties were produced when the microcomposites were processed at the lowest injection molding temperatures, diluted with PBT, and injection molded using a large nozzle tip exit diameter. The pregenerated microcomposite properties were compared against theoretical predictions as well as glass-filled PET. It was found that the pregenerated microcomposites had tensile moduli of approximately 70% of theoretical expectations in the machine direction. Additionally, the comparisons against glass-filled PET revealed that at the same weight fraction of reinforcement, the pregenerated microcomposites had lower properties. Still, the composites were found to have smoother surfaces than glass-filled PET and at temperatures up to 150° C the storage and loss moduli of the pregenerated microcomposites were similar to those of glass filled PET. It was concluded that if the theoretically expected levels of reinforcement could be attained, the pregenerated microcomposites processing scheme would be a viable method of producing light weight, wholly thermoplastic composites with smoother surfaces than are obtained with glass reinforcement. An additional focus of this research was to evaluate the ability to modify the crystallization behavior of a high melting TLCP (HX6000, Tm = 332° C) with a lower melting TLCP (HX8000, Tm = 272°C). It was found that it was possible to tailor the crystallization behavior of these TLCP/TLCP blends by varying the weight fraction of each component, as determined by rheological cooling scans and differential scanning calorimetric cooling tests. Based on the analysis of these TLCPs at the maximum injection molding temperature of 360° C, it was speculated that they had reacted with one another.
Ph. D.

ALes travaux présentés sont une contribution à l’élaboration de composites à matrices thermoplastiques (TP) par un procédé de type Liquid Composite Molding non réactif pour l’industrie automobile. La thèse a été effectuée dans le cadre du projet ANR TAPAS (ThermoplAstic Process for Automotive composite Structure) et s’est focalisée sur la mise en œuvre de plaques composites en renfort continu injectées avec des matrices polyamides 6,6 (PA 6,6) de hautes fluidités par Resin Transfer Molding. Le premier objectif est porté sur l’optimisation des cadences d’injection à travers l’étude de la perméabilité de préformes unidirectionnelles (UD) en fibres de verre et à hauts modules mécaniques. L’architecture de ces UD a ainsi été modifiée de manière à faciliter les écoulements. La perméabilité des différents tissus a pu être évaluée par un couplage entre des mesures expérimentales et une modélisation analytique basée sur un raisonnement à deux échelles de pores : l’écoulement intra et inter-torons. Le deuxième objectif sur lequel les travaux de thèse se sont concentrés s’est reposé sur la maitrise de l’état d’imprégnation par le bais d’une étude complète sur les phénomènes qui se déroulent à l’interface entre la fibre et la matrice à haute température. Plusieurs viscosités et formulations du PA 6,6 ainsi qu’un traitement appliqué sur le verre ont pu être caractérisés et discutés en termes de mouillabilité et d’adhésion. Enfin, la dernière partie du manuscrit présente les résultats obtenus sur les plaques mises en œuvre par RTM-TP en injection in-plane. Les conditions optimales de fonctionnement ainsi que les aspects de saturation, de santé matière et des propriétés mécaniques sont ensuite présentés et discutés
The present work is a contribution to the thermoplastic composites manufacturing by a non-reactive Liquid Composite Molding process for the automotive industry. The thesis was carried out by the « ANR TAPAS » project (Thermoplastic Process for Automotive Composite Structure) and was focused on the elaboration of continious-fiber reinforced composites plates injected with a high-fluidity polyamide 6,6 (PA 6,6) by the Resin Transfer Molding process. The first goal was focused on increasing injection rates through the study of the in-plane permeability of unidirectional (UD) glass fiber fabrics with high mechanical modulus (HM). Experiments and modelling results showed that the permeability of these UD has been enhanced by modifying specific structural parameters of their architecture. The analytical model developped and used is based on a flow distribution according two differents scales of porosity : in and inter-yarns. The second part of the work was focused on the understanding of phenomenas that take place at the interface created between glass fiber and the matrix during the impregnation step. The wettability and adhesion of molten PA 6,6 dropped on a glass substrate is studied at different processing temperature. The last part introduce the thermoplastic composite plates elaborated by RTM-TP process. The optimum operating conditions as well as preforms saturation and mechanical properties are also studied and discussed

The purpose of this research is to empirically investigate flow front void formation rates and post-formation bubble mobility behavior for composites produced via resin transfer molding (RTM).For this study, in situ observation of bubble formation and migration was accomplished by photographing resin flow progression during infusion tests of carbon reinforcements. An analysis strategy for use in batch processing sequential image sets is presented. The use of MATLAB to process and analyze binary images of infusions for void content has garnered satisfactory results and has shown that analysis of progressive image sequences can greatly enrich the volume of in situ measurements for a given study without compromising the data quality.Semi-automated MATLAB software analysis employed the representative image area (RIA) method to evaluate v0. It was found that the shorter the RIA length, and the more it follows the true flow front shape, the more representative the measured v0 was of the void formation at the flow front.Experimental evidence of in situ bubble formation and mobility behavior is presented. Stitch architecture of NCF reinforcements is shown to influence bubble formation at the flow front. Bubble mobility mechanisms (such as escape and entrapment) are related to stitch orientation relative to the fluid flow direction. Different stitching orientations exhibited different effects on post-formation mobility.Void formation is presented as a function of flow front velocity. Despite differences in preform configurations (stitch orientation with respect to flow) and injection flowrates, bubbles seem to form in a similar fashion for the 3 infusions of carbon fiber NCF reinforcement analyzed in this study. It is observed that bubbles form at stitch lines, regardless of stitch orientation.Bubble migration is documented for infusion of NCF reinforcement with stitching at different orientations. Qualitative observations of bubble migration during infusions of a dense preform of STW, plain weave fabric are discussed. Recommendations are given for future studies involving image-based analysis of in situ bubble formation and migration.

Dans cette étude, une nouvelle approche numérique intitulée modèle de fuite multicouche a été développée permettant la simulation efficace par des éléments coques 2D multicouches de la nature tridimensionnelle de l’écoulement dans des préformes multicouches pourvues ou non d’un drainant et caractérisées par un fort gradient d’anisotropie le long de l’épaisseur. La convergence de l’approche développée a été démontrée via une comparaison avec un modèle 2D dans le plan xz. L’efficacité du modèle de fuite vis-à-vis du modèle 3D a été quantifiée. L’intérêt de la nouvelle approche numérique a été révélé via des problèmes à échelle industriel à savoir l’optimisation de la stratégie de dépôt de drainant dans un raidisseur en Hi-Tape, la simulation de remplissage d’une coque de bateau pendant le procédé VARTM et l’étude de l’écoulement dans un stratifié NCF
In this study, a new numerical approach called « Multilayer Leakage Model » has been developed for the efficient numerical simulation of the 3D flow by 2D multilayered shell elements in anisotropic multilayer preform with or without a distribution medium. The convergence of the developed approach has been demonstrated by a comparison with a 2D model in the xz plan. The efficiency of the multilayer leakage model versus the 3D model has been quantified. The advantage of the new numerical approach has been verified through industrial part simulations such as the optimization of distribution medium position in a Hi-Tape stiffener, the flow simulation of a boat in VARTM process and the flow analysis of a NCF laminate

Experimentation exploring the movement of voids within carbon fiber reinforced plastics was performed using fluorescent dye infused into the laminates observed through a transparent mold under ultraviolet light. In situ photography was used as an inspection method for void content during Resin Transfer Molding for these laminates. This in situ inspection method for determining the void content of composite laminates was compared to more common ex-situ quality inspection methods i.e. ultrasonic inspection and cross-section microscopy. Results for localized and total void count in each of these methods were directly compared to test samples and linear correlations between the three test methods were sought. Test coupons were then cut from these laminates and were used to calculate the interlaminar shear strength at certain locations throughout the laminates. Although this research did not adequately observe correlations between results obtained from ultrasonic C-scans, cross-sectional microscopy and in situ photography of the surface, it was seen that the fluid dynamics of the thermosetting epoxy used in this experimentation correlated to results obtained from previous experimentation performed by students at Brigham Young University using vegetable oil as a substitute for resin.

Composite materials have traditionally been used in high-end aerospace parts and low-end consumer parts. The reason for this separation in markets is the wide gap in technology between pre-preg materials processed in an autoclave and chop strand fiberglass blown into an open mold. Liquid composite molding has emerged as a bridge between inexpensive tooling and large, technical parts. Processes such as vacuum infusion have made it possible to utilize complex layups of reinforcement materials in an open mold style set-up, creating optimal conditions for composites to penetrate many new markets with rapid innovation. Flow simulation for liquid composite molding is often performed to assist in process optimization, and requires the permeability of the reinforcement to be characterized. For infusion under a flexible membrane, such as vacuum infusion, or for simulation of a part with non-uniform thickness, one must test the permeability at various levels of compaction. This process is time consuming and often relies on interpolation or extrapolation around a few experimental permeability measurements. To accelerate the process of permeability characterization, a small number of methodologies have been previously presented in the literature, in which the permeability may be tested at multiple fiber volume contents in a single test. Some of the methods even measure the permeability over a continuous range of thicknesses, thus requiring no later interpolation of permeability values. A novel method is presented here for the rapid measurement of permeability over a continuous range of fiber volume content, in a single unidirectional vacuum infusion flow experiment. The thickness gradient across the vacuum bag, as well as the fluid pressure at several locations in the mold, were concurrently measured to calculate the fabric compressibility. An analytical flow model, which accounts for the compressibility, is then used by iterating the fitting constant in a permeability model until the predicted flow front progression matches empirical measurement. The method is demonstrated here for two reinforcement materials: 1) a fiberglass unbalanced weave and 2) a carbon bi-ax non-crimped fabric. The standard deviation of calculated permeabilities across the multiple infusion experiments for each material and flow orientation ranged from 12.8% to 29.7%. Validation of these results was performed by comparing the resulting permeability with multiple non-continuous permeability measurement methods.

L'objectif du projet COMPTINN (COMPosites Tièdes INNovants) est d'obtenir des matériaux composites pouvant être utilisés sur de longues durées, à des températures comprises entre 150°C et 400°C, pour des applications structurales de l'aéronautique civile. Les travaux de thèse s'inscrivent dans un objectif de développement de matériaux thermostructuraux mis au point par un procédé industrialisable pour la production de pièces en série, respectueux de l'environnement et économiquement viable. Les procédés d'élaboration choisis sont ceux utilisés pour la mise en œuvre des CMO (Composites à Matrice Organique) thermodurcissables en moule fermé et par voie liquide. Les procédés les plus conventionnels ont été sélectionnés : l'injection par transfert de résine (RTM : Resin Transfer Molding) et l'infusion de résine sous vide (LRI : Liquid Resin Infusion). Les constituants des composites sont d'une part une matrice vitrocéramique, issue d'une résine dérivée d'un système géopolymérique, et d'autre part des renforts 2D et 3D en fibres de carbone. La viscosité d'une résine est la propriété principale qui conditionne sa mise en œuvre par les procédés d'élaboration par la voie liquide. La résine utilisée étant une suspension dont la viscosité est relativement élevée, la faisabilité de l'élaboration de composites par RTM ou LRI est a priori délicate. Une attention particulière a donc été portée à la rhéologie de la résine. Les résultats de cette étude ont permis d'optimiser d'importants paramètres procédés. L'étude s'est ensuite dirigée vers l'élaboration et la caractérisation des composites. L'impact du procédé de mise en œuvre sur la microstructure et sur les propriétés thermomécaniques des composites a été évalué. L'influence d'autres paramètres procédés, tels que le taux de dilution de la résine, la direction d'imprégnation du renfort et le différentiel de pression, a été étudiée.

The manufacture, laminate design, and modeling of a part with complex geometry are explored. The ultimate goal of the research is to produce a model that accurately predicts part stiffness. This is validated with experimental results of composite parts, which refine material properties for use in a final prototype part model. The secondary goal of this project is to explore manufacturing methods for improved manufacturability of the complex part. The manufacturing portion of the thesis and feedback into material model has incorporated a senior project team to perform research on manufacturing and create composite part to be used for experimental testing. The senior project was designed, led, and managed by the author with support from the committee chair. Finite element modeling was refined using data from coupon 3-point bend testing to improve estimates on material properties. These properties were fed into a prototype part model which predicted deflection of composite parts with different layups and materials. The results of the model were compared to experimental results from prototype part testing and 3rd party analysis. The results showed that an accurate mid-plane shell element model could be used to accurately predict deflection for 2 of 3 experimental parts. There are recommendations in the thesis to further validate the models and experimental testing.

En aéronautique, l’élaboration via des pré-imprégnés n’est pas toujours adaptées àla fabrication de nouvelles pièces de formes complexes ou de grandes dimensions. Desprocédés directs existent, dénommés Liquid Composites Molding (LCM), tels que leResin Transfer Moulding (RTM) ou les procédés d’infusion de résine, comme le LiquidResin Infusion (LRI) et le Resin Film Infusion (RFI). Actuellement, environ 5 à 10%des pièces composites sont fabriqués par ces procédés directs. Avec le procédé RTM,les tolérances dimensionnelles et la porosité peuvent être maîtrisées et on peut atteindredes pièces haute qualité, mais son industrialisation est complexe et les modèlesmécaniques doivent être améliorés pour réaliser des simulations représentatives. Parcontre, les procédés d’infusion peuvent être utilisés dans des conditions plus flexibles,par exemple, dans des moules ouverts à sac vide en nylon ou silicone, à faible coût. Parconséquent, les procédés de LRI et RFI sont particulièrement adaptés pour les petites etmoyennes entreprises car les investissements sont plus faibles par rapport à d’autresprocédés de fabrication.Les procédés par infusion de résine LRI ou RFI sont basés sur l’écoulement d’unerésine liquide (pour RFI, après le cycle de température, la résine solide obtenir son étatliquide) à travers l’épaisseur d’un renfort fibreux sec dénommé préforme.L’optimisation du procédé est difficile à réaliser car le volume de la préforme changefortement pendant le procédé car elle est soumise à une pression extérieure et qu’il n’ya pas de contre-moule. Pour optimiser les paramètres de fabrication des matériauxcomposites par infusion de résine, il est nécessaire de mettre en oeuvre un modèlenumérique. Récemment, une modélisation de l'écoulement d’un fluide isotherme dansun milieu poreux compressible a été développée par P. Celle [1]. Avec ce modèlenumérique, nous avons simulé des cas test en 2D pour des géométries industriellesclassiques. Pour valider ce modèle numérique, des essais d’infusion d’une plaque par leprocédé LRI dans des conditions industrielles ont été réalisés. D’une part, la simulationnumérique permet de calculer le temps de remplissage, l’épaisseur de la préforme et lamasse de la résine durant l’infusion. D’autre part, nous avons suivi de procédéexpérimentalement par des micro-thermocouples, la fibre optique et la projection defranges. Un des points clefs de l’approche expérimentale est que l’écoulement de larésine et le comportement de la préforme dépendent intrinsèquement de paramètres quiévoluent pendant l’infusion de la résine, tels que la variation de l’épaisseur, le temps deremplissage et le taux volumique de fibres, via la perméabilité. Enfin, une comparaisonentre les résultats expérimentaux et la simulation numérique permet de valider lemodèle numérique. Cette confrontation des résultats permettra de mettre en lumière lesdifficultés et les limites de ce modèle numérique, afin d’améliorer les futurs modèles.De plus, ces deux approches constituent un bon moyen d’étudier et d’approfondir nosconnaissances sur les procédés d’infusion de résine, tout en développant un outil desimulation indispensable à la conception de pièces composites avancées
Weight saving is still a key issue for aerospace industry. For instance 50% in weightof the B787 and A350 aircraft structures is made of CFRP, so it is necessary to makelighter thick and complex parts. Direct processes called Liquid Composite Molding(LCM), such as Resin Transfer Moulding (RTM) or Resin Infusion Process (LRI, RFI).At the present time, around 5 to 10% of the parts are manufactured by direct processesand the current trend is clearly to go ahead. In RTM process, the dimensional tolerancesand porosity fraction can be kept under control and high quality parts produced, but itsindustrialisation is complex and refined models are still needed to perform simulations.On the contrary, the resin infusion process can be utilized in flexible conditions, such asin low cost open moulds with vacuum bags in nylon or silicone. This type of processonly requires low resin pressure and the tooling is less expensive than RTM rigidmoulds. Therefore LRI and RFI processes are particularity suitable for small andmedium size companies because the investments are rather low compared to othermanufacturing process.Liquid Resin Infusion (LRI) processes are promising manufacturing routes toproduce large, thick or complex structural parts. They are based on the resin flowinduced across its thickness by pressure applied onto a preform / resin stacking.However, both thickness and fibre volume fraction of the final piece are not wellcontrolled since they result from complex mechanisms which drive the transientmechanical equilibria leading to the final geometrical configuration. In order tooptimize both design and manufacturing parameters, but also to monitor the LRIprocess, an isothermal numerical model has been developed by P. Celle [1], whichdescribes the mechanical interaction between the deformations of the porous mediumand the resin flow during infusion. With this numerical model, we have investigated theLRI process with classical industrial piece shapes. To validate the numerical model andto improve the knowledge of the LRI process, the researcher work details a comparisonbetween numerical simulations and an experimental study of a plate infusion testcarried out by LRI process under industrial conditions. From the numerical prediction,the filling time, the resin mass and the thickness of the preform can be determined. Onanother hand, the resin flow and the preform response can be monitored bymicro-thermocouples, optical fibre sensor and fringe projection during the filling stage.One key issue of this research work is to highlight the major process parameterschanges during the resin infusion stage, such as the preform and resin temperature, thevariations of both thickness and fiber volume fraction of the preform. Moreover, thesetwo approaches are both good ways to explore and improve our knowledge on the resininfusion processes, and finally, to develop simulation tools for the design of advancedcomposite parts

En aéronautique, l'élaboration via des pré-imprégnés n'est pas toujours adaptées àla fabrication de nouvelles pièces de formes complexes ou de grandes dimensions. Desprocédés directs existent, dénommés Liquid Composites Molding (LCM), tels que leResin Transfer Moulding (RTM) ou les procédés d'infusion de résine, comme le LiquidResin Infusion (LRI) et le Resin Film Infusion (RFI). Actuellement, environ 5 à 10%des pièces composites sont fabriqués par ces procédés directs. Avec le procédé RTM,les tolérances dimensionnelles et la porosité peuvent être maîtrisées et on peut atteindredes pièces haute qualité, mais son industrialisation est complexe et les modèlesmécaniques doivent être améliorés pour réaliser des simulations représentatives. Parcontre, les procédés d'infusion peuvent être utilisés dans des conditions plus flexibles,par exemple, dans des moules ouverts à sac vide en nylon ou silicone, à faible coût. Parconséquent, les procédés de LRI et RFI sont particulièrement adaptés pour les petites etmoyennes entreprises car les investissements sont plus faibles par rapport à d'autresprocédés de fabrication.Les procédés par infusion de résine LRI ou RFI sont basés sur l'écoulement d'unerésine liquide (pour RFI, après le cycle de température, la résine solide obtenir son étatliquide) à travers l'épaisseur d'un renfort fibreux sec dénommé préforme.L'optimisation du procédé est difficile à réaliser car le volume de la préforme changefortement pendant le procédé car elle est soumise à une pression extérieure et qu'il n'ya pas de contre-moule. Pour optimiser les paramètres de fabrication des matériauxcomposites par infusion de résine, il est nécessaire de mettre en oeuvre un modèlenumérique. Récemment, une modélisation de l'écoulement d'un fluide isotherme dansun milieu poreux compressible a été développée par P. Celle [1]. Avec ce modèlenumérique, nous avons simulé des cas test en 2D pour des géométries industriellesclassiques. Pour valider ce modèle numérique, des essais d'infusion d'une plaque par leprocédé LRI dans des conditions industrielles ont été réalisés. D'une part, la simulationnumérique permet de calculer le temps de remplissage, l'épaisseur de la préforme et lamasse de la résine durant l'infusion. D'autre part, nous avons suivi de procédéexpérimentalement par des micro-thermocouples, la fibre optique et la projection defranges. Un des points clefs de l'approche expérimentale est que l'écoulement de larésine et le comportement de la préforme dépendent intrinsèquement de paramètres quiévoluent pendant l'infusion de la résine, tels que la variation de l'épaisseur, le temps deremplissage et le taux volumique de fibres, via la perméabilité. Enfin, une comparaisonentre les résultats expérimentaux et la simulation numérique permet de valider lemodèle numérique. Cette confrontation des résultats permettra de mettre en lumière lesdifficultés et les limites de ce modèle numérique, afin d'améliorer les futurs modèles.De plus, ces deux approches constituent un bon moyen d'étudier et d'approfondir nosconnaissances sur les procédés d'infusion de résine, tout en développant un outil desimulation indispensable à la conception de pièces composites avancées.

Dans ce manuscrit, un modèle complet pour la simulation de l'écoulement d'un fluide thermor éactif à travers un milieu poreux fortement compressible est présenté. Ce modèle est utilisé pour l'étude des procédés d'élaboration des matériaux composites par infusion à travers leur épaisseur (Liquid Resin Infusion-LRI et Resin Film Infusion-RFI ). Dans ces procédés, le mélange entre les renforts et la résine liquide est réalisé dans la direction transverse aux plans des préformes pendant la phase de mise en forme. Les coˆuts sont ainsi réduits et les problèmes de remplissage éliminés. Ces procédés sont néanmoins peu maîtrisés et les caractéristiques de la pièce finale difficilement prévisibles (principalement les épaisseurs et les porosités). La mise au point d'un modèle numérique constituerait un bon outil pour développer et finaliser de nouvelles solutions composites. D'un point de vue physique, l'infusion de la résine à travers l'épaisseur des préformes est une conséquence de la pression appliquée sur l'empilement résine/préforme. Dans cette analyse multi-physique deux types de problèmes sont rencontrés. Tout d'abord, on connait mal les conditions de couplage entre les zones liquides, gouvernées par les équations de Stokes, et les préformes imprégnées assimilées à des milieux poreux, gouvernées par une loi de Darcy et une loi de comportement mécanique non-linéaire. Par ailleurs, les interactions entre l'écoulement de la résine et la compression des préformes ne sont pas bien maîtrisées. Le modèle développé inclut donc une condition de Beaver-Joseph- Schaffman modifiée pour le couplage entre les zones de Darcy et de Stokes. Une formulation ALE pour l'écoulement de la résine dans un milieu poreux déformable subissant de fortes déformations est utilisée et couplée à une formulation Lagrangienne Réactualisée pour la partie solide. Ces deux mécanismes physiques sont couplés à des modèles thermo-chimiques pour traiter la réticulation de la résine sous l'action du cycle de température. Dans ce travail, un certain nombre d'outils numériques et de nouvelles formulations ont été développés en vue de simuler les procédés LRI et RFI. Chaque outil est étudié et validé analytiquement ou numériquement avant d'être intégré dans les modèles LRI /RFI. Des simulations numériques d'infusion sont ensuite présentées et commentées, puis une première comparaison avec des essais expérimentaux est proposée.

碩士
國立中央大學
機械工程研究所
90
ABSTRACT In recent years, Resin Transfer Molding (RTM) has been widely used in aerospace, automobile and precision industries, because its processing cycle is short; besides the stiffness and precision of its products are high and the styles of the products can be easily changed. Due to the size of a finish is too large, or the permeability of the fiber is too low, the filling time would be longer. Therefore, we try to use the filling method of Injection-Compression liquid composite molding to shorten the filling time. The experiment use transparent acrylic mold, combined with flow field visualization to analyze and simulate the flow field condition of Injection-Compression liquid composite molding in filling period. We use Buckingham Theorem to couple the processing parameters, such as injection pressure, compression pressure, gap size, porosity, permeability, surface tension, velocity, and plot them as a function diagram; besides, we also compare our findings with RTM to show the advantages of Injection-Compression liquid composite molding. . Key word：reinforced plastic, Liquid Composite Molding, Injection / Compression Liquid Composite Molding, Resin Transfer Molding, porous media

碩士
大葉大學
機械工程研究所
88
There are various manufacturing processes to produce reinforced plastics. Resin Transfer Molding (RTM) is one of the most important processes. In producing large-surface-area parts with low fiber permeability, long mold filling time is needed, i.e. the cycle time is large. Moreover, the resin might gel before the filling period ends. To prevent the short shot, increasing the injection pressure is a possible choice. However, the equipment cost is increased. Excessive injection pressure would also produce the fiber deformation or the fiber wash-out, and it affects the quality of the reinforced plastics. The main goal of the proposed research is to provide a novel approach , Injection-Compression Liquid Composite Molding, which can reduce the injection pressure and improve the part quality. The research will be conducted through modeling , numerical simulation and experimental analysis. Control Volume-Finite Element Method has been widely used in RTM simulation. This research is going to apply this numerical approach to simulate the Injection-Compression LCM processes. The process parameters, injection pressure, part thickness before and after the compression, the permeability and compressibility of the fiber preform, on quality of Injection Compression LCM parts will be investigated. The experimental results will be compared with the theoretical predictions. The quality test of the composite samples will be conducted through a three point bending test and by using microscopes .

碩士
國立成功大學
航空太空工程學系
89
In the resin transfer molding (RTM) process, there are several important processing parameters, such as injection pressure, resin injection temperature, mold pre-heated temperature, mold heating rate, and cure temperature, which have major effect on the quality of a RTM product. In general, these parameters are determined based on engineer’s experience. In order to establish an efficient way for selecting the process parameter, optimization methods based on the computer aided process simulation are used. The optimization of manufacturing parameters on RTM was developed by Yu[5]. He used the numerical method to simulate the RTM manufacturing process and the genetic algorithm to search the optimal manufacturing parameters based on the simulation results. Yet there exists a problem which impede a wide application of this approch. The RTM manufacturing process simulation is takes too much computation time and the entire optimization is slow. In this study, the RTM process simulation program is replaced by the artificial neural network method. A neural network is used to learn the correlation between input and output data of the RTM program. With the simulation results, the neural network is trained to create a rapid RTM process model. Genetic algorithm is still applied to this rapid RTM manufacturing process model to search for the optimum solution for a RTM process design. Finally, the major objective of this research is to study that if the RTM simulation program can replaced with a neural network successfully. 授權書 摘要 英文摘要 誌謝 目錄 i 表目錄 iii 圖目錄 iv 第一章、 緒論 1.1簡介 1 1.2樹脂轉注成型法簡介 2 1.3文獻回顧 3 1.4研究動機與目標 4 1.5研究方法 6 第二章、 類神經網路簡介 2.1類神經網路概論 10 2.2倒傳遞類神經網路 12 2.2.1導論 12 2.2.2網路架構 12 2.2.3網路演算法 14 2.2.4網路參數選取 17 2.2.5網路的修正與效率改善 18 2.2.6網路的優缺點 20 2.3本文所使用類神經網路架構介紹 22 第三章、 基因演算法則簡介 3.1簡介 23 3.2編碼的形式 24 3.3二進位式（Binary）基因演算法操作流程 25 3.4基因演算法的特性 29 3.5本文所使用基因演算法架構介紹 30 第四章、 RTM製程類神經網路模擬之分析與討論 4.1類神經網路模擬操作流程 32 4.2類神經網路參數選取 34 4.3模擬結果討論 36 第五章、 製程參數之最佳化 5.1製程參數的分析 39 5.2參數範圍的選取及價值函數定義 41 5.3基因演算法之參數選取 46 5.4利用基因演算法搜尋結果討論 48 第六章、 結論與建議 52 參考文獻 55 自述 著作權聲明 表目錄 表4-1 疊代次數為100萬次時，處理單元個數2～70的誤差函數值 57 表4-2 疊代次數為100萬次時，處理單元個數10～70的運算次數 比較 57 表4-3 不同sse誤差收斂下，測試誤差的比較 58 表5-1 不同交配機率之搜尋成功率比較表 59 表5-2 不同突變機率之搜尋成功率比較表 59 表5-3類神經網路回想所得之488187種參數組合中，適應函數值 最大的前50種組合 60 表5-4 類神經網路回想所得之488187種參數組合中，適應函數值 最大的前50種組合。（限定樹脂充模溫度=25℃時） 61 表5-5 類神經網路回想所得之488187種參數組合中，適應函數值 最大的前50種組合。（限定樹脂充模溫度=25℃以及加熱速 率=3℃/min時） 62 表6-1 不同成化溫度範圍所對應的適應函數值範圍之列表 63 圖目錄 圖1-1 樹脂轉注成形法製造程序圖 64 圖1-2 製程參數最佳化流程圖（RTM模擬程式之架構） 65 圖1-3 製程參數最佳化流程圖（應用類神經網路之架構） 66 圖1-4 本研究所模擬之範例使用模具，及其進口與量測位置 67 圖2-1 神經元架構圖 68 圖2-2 倒傳遞類神經網路架構簡圖 68 圖2-3 倒傳遞類神經網路詳細架構圖（摘自MATLAB Toolbox） 69 圖2-4 轉換函數列表（摘自MATLAB Toolbox） 70 圖3-1 基因演算法的演化流程圖 71 圖3-2 基因演算法則單點交配示意圖 72 圖4-1 倒傳遞類神經網路演算過程（訓練及回想） 73 圖4-2 倒傳遞網路訓練誤差收斂圖（誤差震盪示意圖） 74 圖4-3 隱藏層處理單元個數=40、學習速率為0.0001 ，收斂到 sse誤差為2.5的誤差收斂圖 75 圖4-4 收斂到sse誤差=2.5的測試誤差分佈圖（適應函數值 之學習） 76 圖4-5 訓練誤差值sse為4.0、2.5及2.0的誤差值分佈比較圖 77 圖4-6 收斂到sse誤差=6.0的測試誤差分佈圖 （最高溫度之學習） 78 圖5-1 環氧樹脂中模具預熱溫度對充模時間的影響 79 圖5-2 不同成化溫度對位置1溫度之影響（預熱溫度：60℃ 、樹脂充模溫度：28℃、加熱速率：4℃/min） 80 圖5-3 不同預熱溫度對位置1溫度之影響（樹脂充模溫度：28℃ 、加熱速率：4℃/min、成化溫度：140℃） 81 圖5-4 不同加熱速率對位置1溫度之影響（預熱溫度：60℃ 、樹脂充模溫度：28℃、成化溫度：140℃） 82 圖5-5 基因演算法則收斂趨勢圖（橫軸為函數計算次數） 83 圖5-6 RTM製程中時間-溫度變化圖 84 圖5-7 基因演算法則收斂趨勢圖（樹脂充模溫度為25℃） 85 圖5-8 RTM製程中時間-溫度變化圖（樹脂充模溫度為25℃） 86 圖5-9 基因演算法則收斂趨勢圖（樹脂充模溫度為25℃、加熱 速率為3.0℃/min） 87 圖5-10 RTM製程中時間-溫度變化圖（樹脂充模溫度為25℃、加熱 速率為3.0℃/min） 88 圖5-11 基因演算法則收斂趨勢圖（模具欲熱溫度大於25℃） 89 圖5-12 RTM製程中時間-溫度變化圖（模具欲熱溫度大於25℃） 90