Doctor of Philosophy / Department of Mechanical and Nuclear Engineering / Youqi Wang / Ballistic performance of textile fabric is affected by numerous elements, such as fabric architecture, material property, and projectile characteristics. Near fiber-level microstructures of soft body armor composed of multi-layer Kevlar KM-2 fabrics are generated for numerical simulation. The modified digital element approach (DEA) is applied to determine the ballistic limit of textile fabrics against fragment simulating projectiles (FSP). Different from other numerical models, the DEA takes a considerable amount of fiber-level detail into consideration and models the fabric at filament-level. In this approach, fabric is an assembly of yarns weaved and relaxed into pre-arranged pattern; yarn is simulated as a bundle of digital fibers. When the number of digital fibers per yarn reaches the number of actual fibers per yarn, fiber-level simulation is achieved. The DEA model successfully simulates real scale multi-layer fabric impacted by spherical projectile and accurately predicted fabric displacement and failure mechanism. It was assumed that the digital fiber is fully flexible and its bending rigidity is negligible. Shear force was thus neglected. However, for projectiles with sharp edge(s), such as FSP, due to resultant shear force, fabric failure starts where it interacts with projectile edge. As a result, the numerical results derived from the previous DEA overestimated the impact strength of fabrics against projectiles with shape edges. Therefore, shear force and fiber bending rigidity must be considered. In the modified DEA approach, numerical tests are employed to determine the effective bending rigidity of digital fiber. A combined tension-shear failure model is then incorporated into the DEA in order to calculate the shear force applied to fibers. The 3-D microscope is applied to measure the radius of FSP along the edge. The surface of the FSP is meshed into triangle elements. A unique algorithm is developed and employed to search contacts between textile fabric and projectile of arbitrary shape. In this research, first, an overview of ballistic impact analysis is discussed; the previous DEA model used in simulating ballistic impact and penetration process is presented. Second, the modified DEA approach used in simulating arbitrary shape projectile perforation process is established and verified. The method of searching and calculating contacts between textile fabric and solid body projectile is explained. The convergence and accuracy of digital element mesh are investigated statistically using tension-shear failure model. Third, fabric shear force and fiber bending rigidity are investigated using tension-shear failure model. The effective digital fiber area moment of inertia is numerically determined. Fourth, standard ballistic tests of real scale multi-layer Kevlar KM2 fabrics are simulated using FSP. Numerical results are compared to high-resolution experimental test data. The modified DEA is validated.

This thesis presents a model of fabric bending hysteresis. The hysteresis model is designed to reproduce the fabric bending measurements taken by the Kawabata Evaluation System (KES) and the model parameters can be derived directly from these property measurements. The advantage to using this technique is that it provides the ability to simulate a continuum of property curves. Results of the model and its components are compared and constrasted with experimental results for fabrics composed of different weaves and yarn types. An attempt to incorporate the bending model as part of a fabric drape simulation is also made.

A new approach to local nonlinear image restoration is described, based on approximating functions using a regular grid of points in a many-dimensional space. Symmetry reductions and compression of the sparse grid make it feasible to work with twelve-dimensional grids as large as 22¹². Unlike polynomials and neural networks whose filtering complexity per pixel is linear in the number of filter co-efficients, grid filters have O(1) complexity per pixel. Grid filters require only a single presentation of the training samples, are numerically stable, leave unusual image features unchanged, and are a superset of order statistic filters. Results are presented for additive noise, blurring, and superresolution.

An automatic question answering system must find, within a corpus,short factual answers to questions posed in natural language. The process involves analyzing the question, retrieving information related to the question, and extracting answers from the retrieved information. This thesis presents a novel approach to answer extraction in an automated question answering (QA) system. The answer extraction approach is an extension of the MultiText QA system. This system employs a question analysis component to examine the question and to produce query terms for the retrieval component which extracts several document fragments from the corpus. The answer extraction component selects a few short answers from these fragments. This thesis describes the design and evaluation of the Redundant Inverse Term Frequency (RITF) answer extraction component. The RITF algorithm locates and evaluates words from the passages that are likely to be associated with the answer. Answers are selected by finding short fragments of text that contain the most likely words based on: the frequency of the words in the corpus, the number of fragments in which the word occurs, the rank of the passages as determined by the IR, the distance of the word from the centre of the fragment, and category information found through question analysis. RITF makes a substantial contribution in overall results, nearly doubling the Mean Reciprocal Rank (MRR), a standard measure for evaluating QA systems.

A new approach to local nonlinear image restoration is described, based on approximating functions using a regular grid of points in a many-dimensional space. Symmetry reductions and compression of the sparse grid make it feasible to work with twelve-dimensional grids as large as 22¹². Unlike polynomials and neural networks whose filtering complexity per pixel is linear in the number of filter co-efficients, grid filters have O(1) complexity per pixel. Grid filters require only a single presentation of the training samples, are numerically stable, leave unusual image features unchanged, and are a superset of order statistic filters. Results are presented for additive noise, blurring, and superresolution.

This thesis presents a model of fabric bending hysteresis. The hysteresis model is designed to reproduce the fabric bending measurements taken by the Kawabata Evaluation System (KES) and the model parameters can be derived directly from these property measurements. The advantage to using this technique is that it provides the ability to simulate a continuum of property curves. Results of the model and its components are compared and constrasted with experimental results for fabrics composed of different weaves and yarn types. An attempt to incorporate the bending model as part of a fabric drape simulation is also made.

An automatic question answering system must find, within a corpus,short factual answers to questions posed in natural language. The process involves analyzing the question, retrieving information related to the question, and extracting answers from the retrieved information. This thesis presents a novel approach to answer extraction in an automated question answering (QA) system. The answer extraction approach is an extension of the MultiText QA system. This system employs a question analysis component to examine the question and to produce query terms for the retrieval component which extracts several document fragments from the corpus. The answer extraction component selects a few short answers from these fragments. This thesis describes the design and evaluation of the Redundant Inverse Term Frequency (RITF) answer extraction component. The RITF algorithm locates and evaluates words from the passages that are likely to be associated with the answer. Answers are selected by finding short fragments of text that contain the most likely words based on: the frequency of the words in the corpus, the number of fragments in which the word occurs, the rank of the passages as determined by the IR, the distance of the word from the centre of the fragment, and category information found through question analysis. RITF makes a substantial contribution in overall results, nearly doubling the Mean Reciprocal Rank (MRR), a standard measure for evaluating QA systems.

Technical textiles are increasingly being engineered and used in challenging applications, in areas such as safety, biomedical devices, architecture and others, where they must meet stringent demands including excellent and predictable load bearing capabilities. They also form the bases for one of the most widespread group of composite materials, fibre reinforced polymer-matrix composites (PMCs), which comprise materials made of stiff and strong fibres generally available in textile form and selected for their structural potential, combined with a polymer matrix that gives parts their shape. Manufacturing processes for PMCs and technical textiles, as well as parts and advanced textile structures must be engineered, ideally through simulation, and therefore diverse properties of the textiles, textile reinforcements and PMC materials must be available for predictive simulation. Knowing the detailed geometry of technical textiles is essential to predicting accurately the processing and performance properties of textiles and PMC parts. In turn, the geometry taken by a textile or a reinforcement textile is linked in an intricate manner to its constitutive behaviour. This thesis proposes, investigates and validates a general numerical tool for the integrated and comprehensive analysis of textile geometry and constitutive behaviour as required toward engineering applications featuring technical textiles and textile reinforcements. The tool shall be general with regards to the textiles modelled and the loading cases applied. Specifically, the work aims at fulfilling the following objectives: 1) developing and implementing dedicated simulation software for modelling textiles subjected to various load cases; 2) providing, through simulation, geometric descriptions for different textiles subjected to different load cases namely compaction, relaxation and shear; 3) predicting the constitutive behaviour of the textiles undergoing said load cases; 4) identifying parameters affecting the textile geometry and constitutive behaviour under evolving loading; 5) validating simulation results with experimental trials; and 6) demonstrating the applicability of the simulation procedure to textile reinforcements featuring large numbers of small fibres as used in PMCs. As a starting point, the effects of reinforcement configuration on the in-plane permeability of textile reinforcements, through-thickness thermal conductivity of PMCs and in-plane stiffness of unidirectional and bidirectional PMCs were quantified systematically and correlated with specific geometric parameters. Variability was quantified for each property at a constant fibre volume fraction. It was observed that variability differed strongly between properties; as such, the simulated behaviour can be related to variability levels seen in experimental measurements. The effects of the geometry of textile reinforcements on the aforementioned processing and performance properties of the textiles and PMCs made from these textiles was demonstrated and validated, but only for simple cases as thorough and credible geometric models were not available at the onset of this work. Outcomes of this work were published in a peer-reviewed journal [101]. Through this thesis it was demonstrated that predicting changes in textile geometry prior and during loading is feasible using the proposed particle-based modelling method. The particle-based modelling method relies on discrete mechanics and offers an alternative to more traditional methods based on continuum mechanics. Specifically it alleviates issues caused by large strains and management of intricate, evolving contact present in finite element simulations. The particle-based modelling method enables credible, intricate modelling of the geometry of textiles at the mesoscopic scale as well as faithful mechanical modelling under load. Changes to textile geometry and configuration due to the normal compaction pressure, stress relaxation, in-plane shear and other types of loads were successfully predicted. During simulation, particles were moved randomly until a stable state of minimum strain energy in the system was reached; as particles moved upon iteration, the configuration of fibres in the textile changed under constant boundary conditions. Then boundary conditions were altered corresponding to strains imposed on the textile, and the system was iterated again towards a new state of minimum strain energy. The Metropolis algorithm of the Monte Carlo method was adopted in this specific implementation. The method relies on a statistical approach implemented in computational algorithms. In addition to geometrical modelling, the proposed particle-based modelling method enables the prediction of major elements of the constitutive behaviour of textiles and textile reinforcements. In fact, prediction of the constitutive behaviour is integral to the prediction of the meso-scale geometry. Simulation results obtained from the proposed particle-based modelling method were validated experimentally for yarns, single-layer textiles and multi-layer textiles undergoing compaction. Validation work showed that the particle-based modelling method replicates reality very faithfully, and it also showed the suitability of including Gutowski's function along with Hertz' function for representing lateral compaction of yarns. The procedure and results were accepted in final form for publication in a peer reviewed journal [104]. The capability of the proposed particle-based modelling method towards replicating the time-dependent relaxation and reconfiguration of woven textiles subjected to compaction loading was investigated. The capability, which was demonstrated for single and double-layers of plain woven textiles, is intrinsic to the modelling method. The method is unique in the fact that in contrary to work previously reported in the literature, it models the compaction and the relaxation seamlessly in the same simulations and environment. This work is being finalised towards submission for publication in a peer reviewed journal [103]. The proposed particle-based modelling method was also used for modelling in-plane shear in woven textiles. Simulation results were validated experimentally for a single-layer plain woven textile. Validation work showed that the particle-based modelling method reproduces experimental data and published trends very well. A novel algorithm for modelling friction was introduced, leading to results being obtained from a significantly less computationally demanding procedure in these simulations. This work was submitted for publication in a peer reviewed journal [102]. Finally the thesis discusses early work towards the application of the method to carbon fibre fabrics through the description of expansion algorithm (EA) to be used in modelling textiles made of yarns featuring very large numbers of fibres. Furthermore, additional modelling work is presented towards further manufacturing process involving technical textiles, namely textile bending and punching. The latter part is presented as early steps towards future work.

Doctor of Philosophy / Department of Mechanical and Nuclear Engineering / Youqi Wang / The strength of the fabric system is based on fiber strength and fabric mechanics. Modeling a fabric system accurately requires research into fiber behavior within the yarn and yarn behavior within the fabric. Limited computer resources require new approaches to yarn modeling and fabric modeling especially in regards to ballistic impact. The fabric is discontinuous. There are many factors which require modeling the physics in order to accurately simulate and design fabric systems. Weaving yarns into fabrics can introduce fiber level damages such as surface defects and crimps through sliding friction and bending and thus add variance to the tensile strength of the fibered yarn. A Weibull distribution is an often used method to develop a statistical model and is developed to calculate the strength of the yarn. It is necessary to carefully remove the fibers from the as woven fabric and use a standard ASTM single fiber tensile test to create a Weibull distribution of tensile strength. In general in Kevlar systems the edge radius for laboratory projectiles is much larger than the actual dimeter of the fiber; however, the yarn itself can be sheared, and this fibered yarn system requires modeling. There is no direct measurement of Kevlar fiber shear strength, so combined tensile-twist test data is used to develop equations to determined shear strength. DFMA is modeling software developed to create digital fabrics in a method that accurately models yarn shape with limited computer resources using a concept of a digital fiber. The digital fiber represents multiple real fibers, so it is necessary to use the digital yarn effective bending rigidity developed with numerical simulation of experimental results. Since the yarn is composed of hundreds to thousands of fibers, the physical yarn cannot be modeled in full scale fabrics. The yarn composed of digital fibers is structurally similar to real yarns and is capable of representing the real fabric mechanics. In the process of impact, within the relatively short time frame, the distribution of stress is mostly in principal yarns at a time when the event is considered complete through penetration or projectile rebound. The hybrid mesh method represents the small number of principal yarns with high density mesh and the rest of the fabric (the non-principal yarns) with coarse mesh. With hybrid mesh, the full scale simulation of actual fabrics is possible. The projectile geometry for real threats is variant depending on the types of projectiles in use (projectiles for maximum energy transfer to the target or projectiles for high shear). The laboratory projectiles are therefore variant in order to represent threats. In this research the RCC is the threat and two standard weights are modeled with local geometry. The local laboratory projectile geometry is controlled however it is bounded by a tolerance much larger than the Kevlar fibers studied here. It does act against the fibered yarn which will shear mechanically dependent on fiber to fiber interactions and possibly fiber shear strength.

Technical textiles are increasingly being engineered and used in challenging applications, in areas such as safety, biomedical devices, architecture and others, where they must meet stringent demands including excellent and predictable load bearing capabilities. They also form the bases for one of the most widespread group of composite materials, fibre reinforced polymer-matrix composites (PMCs), which comprise materials made of stiff and strong fibres generally available in textile form and selected for their structural potential, combined with a polymer matrix that gives parts their shape. Manufacturing processes for PMCs and technical textiles, as well as parts and advanced textile structures must be engineered, ideally through simulation, and therefore diverse properties of the textiles, textile reinforcements and PMC materials must be available for predictive simulation. Knowing the detailed geometry of technical textiles is essential to predicting accurately the processing and performance properties of textiles and PMC parts. In turn, the geometry taken by a textile or a reinforcement textile is linked in an intricate manner to its constitutive behaviour. This thesis proposes, investigates and validates a general numerical tool for the integrated and comprehensive analysis of textile geometry and constitutive behaviour as required toward engineering applications featuring technical textiles and textile reinforcements. The tool shall be general with regards to the textiles modelled and the loading cases applied. Specifically, the work aims at fulfilling the following objectives: 1) developing and implementing dedicated simulation software for modelling textiles subjected to various load cases; 2) providing, through simulation, geometric descriptions for different textiles subjected to different load cases namely compaction, relaxation and shear; 3) predicting the constitutive behaviour of the textiles undergoing said load cases; 4) identifying parameters affecting the textile geometry and constitutive behaviour under evolving loading; 5) validating simulation results with experimental trials; and 6) demonstrating the applicability of the simulation procedure to textile reinforcements featuring large numbers of small fibres as used in PMCs. As a starting point, the effects of reinforcement configuration on the in-plane permeability of textile reinforcements, through-thickness thermal conductivity of PMCs and in-plane stiffness of unidirectional and bidirectional PMCs were quantified systematically and correlated with specific geometric parameters. Variability was quantified for each property at a constant fibre volume fraction. It was observed that variability differed strongly between properties; as such, the simulated behaviour can be related to variability levels seen in experimental measurements. The effects of the geometry of textile reinforcements on the aforementioned processing and performance properties of the textiles and PMCs made from these textiles was demonstrated and validated, but only for simple cases as thorough and credible geometric models were not available at the onset of this work. Outcomes of this work were published in a peer-reviewed journal [101]. Through this thesis it was demonstrated that predicting changes in textile geometry prior and during loading is feasible using the proposed particle-based modelling method. The particle-based modelling method relies on discrete mechanics and offers an alternative to more traditional methods based on continuum mechanics. Specifically it alleviates issues caused by large strains and management of intricate, evolving contact present in finite element simulations. The particle-based modelling method enables credible, intricate modelling of the geometry of textiles at the mesoscopic scale as well as faithful mechanical modelling under load. Changes to textile geometry and configuration due to the normal compaction pressure, stress relaxation, in-plane shear and other types of loads were successfully predicted. During simulation, particles were moved randomly until a stable state of minimum strain energy in the system was reached; as particles moved upon iteration, the configuration of fibres in the textile changed under constant boundary conditions. Then boundary conditions were altered corresponding to strains imposed on the textile, and the system was iterated again towards a new state of minimum strain energy. The Metropolis algorithm of the Monte Carlo method was adopted in this specific implementation. The method relies on a statistical approach implemented in computational algorithms. In addition to geometrical modelling, the proposed particle-based modelling method enables the prediction of major elements of the constitutive behaviour of textiles and textile reinforcements. In fact, prediction of the constitutive behaviour is integral to the prediction of the meso-scale geometry. Simulation results obtained from the proposed particle-based modelling method were validated experimentally for yarns, single-layer textiles and multi-layer textiles undergoing compaction. Validation work showed that the particle-based modelling method replicates reality very faithfully, and it also showed the suitability of including Gutowski's function along with Hertz' function for representing lateral compaction of yarns. The procedure and results were accepted in final form for publication in a peer reviewed journal [104]. The capability of the proposed particle-based modelling method towards replicating the time-dependent relaxation and reconfiguration of woven textiles subjected to compaction loading was investigated. The capability, which was demonstrated for single and double-layers of plain woven textiles, is intrinsic to the modelling method. The method is unique in the fact that in contrary to work previously reported in the literature, it models the compaction and the relaxation seamlessly in the same simulations and environment. This work is being finalised towards submission for publication in a peer reviewed journal [103]. The proposed particle-based modelling method was also used for modelling in-plane shear in woven textiles. Simulation results were validated experimentally for a single-layer plain woven textile. Validation work showed that the particle-based modelling method reproduces experimental data and published trends very well. A novel algorithm for modelling friction was introduced, leading to results being obtained from a significantly less computationally demanding procedure in these simulations. This work was submitted for publication in a peer reviewed journal [102]. Finally the thesis discusses early work towards the application of the method to carbon fibre fabrics through the description of expansion algorithm (EA) to be used in modelling textiles made of yarns featuring very large numbers of fibres. Furthermore, additional modelling work is presented towards further manufacturing process involving technical textiles, namely textile bending and punching. The latter part is presented as early steps towards future work.