The increasing use of fibre reinforced polymers in structural components often requires an accurate assessment of the strength of the component. The strength of composite materials is usually based on the strength of an individual lamina. This is then combined in a manner depending on the orientation of the plies within the laminate. The actual failure process is often ignored in this type of analysis. Composite failure is the result of damage accumulation from a number of failure modes, in particular, fibre failure, matrix failure and failure of the interface between the fibres and matrix. Measurement of the interfacial strength requires specialised testing techniques in order to obtain accurate characterization of the interfacial failure processes. This research uses a double torsion specimen reinforced with fibres in a number of configurations. The testing techniques developed allow the interaction of a matrix crack with fibres, resulting in the failure of the interface. Finite element analysis has been used to gain an insight into the deformation mechanisms. A compliance change analysis has been developed so that the load in the fibres can be calculated. Results from the finite element analysis confirm the analytical procedures and show that, for the fibre/resin combination tested, the interface has a lower fracture toughness than the matrix material. The interaction between the fibres and matrix shows that the mechanism of fibre bridging inhibits the propagation of matrix cracks. This produces an apparent increase in the toughness of the composite system. To confirm the failure processes occurring, the technique of acoustic emission has been used to monitor the development of the specimen failure. In line with other workers, it is shown that matrix failure produces low amplitude events and interfacial failure produces mid amplitude events. Fibre failure did not occur to any significant degree. This thesis shows how the contribution from the presence of an interface affects the fracture of composite materials and how, via the reinforced double torsion specimen, this contribution can be measured and interpreted.

This thesis explores the seismic analysis and design of post-tensioning concrete masonry (PCM) walls. Using unbonded post-tensioning, walls are vertically prestressed by means of strands or bars which are passed through vertical ducts inside the walls. As the walls are subjected to lateral displacements (in-plane loading), gaps form at the horizontal joints, reducing the system stiffness. As long as the prestressing strands are kept within the elastic limit, or at least maintain a considerable amount of the initial prestressing force, they can provide a restoring force, which will return the walls to their original alignment upon unloading. The key feature in this behaviour is attributable to the tendons being unbonded over the entire wall height, allowing for distribution of tendon strain over the entire length of the tendon. An extensive literature review found that post-tensioning of masonry has had limited application in seismic areas and that there currently are no specific code requirements for it’s use for ductile seismic design, largely as a consequence of little knowledge about the ductility capacity and energy dissipation characteristics. It was concluded that structural testing of PCM walls and concrete masonry creep and shrinkage testing were essential to advance the understanding of this construction type. Creep and shrinkage experiments confirmed that long term prestress losses are considerable in both grouted and ungrouted concrete masonry, and must be taken into account in design. It was concluded that it is essential to use high strength steel for prestressing of PCM in order to reduce long term losses. Structural testing confirmed that fully grouted unbonded post-tensioned concrete masonry is a competent material combination for ductile structural wall systems. In particular, PCM walls strengthened in the flexural compression zones with confining plates are expected to successfully withstand severe ground shaking from an earthquake. It was suggested that partially and ungrouted PCM walls may suitably be used in strength design (non-ductile). The proposed prediction method for wall in-plane behaviour was validated by experimental results. Good correlation between predictions and results was found. Displacement spectra were developed for ductile seismic design of PCM walls. These can be used to accurately estimate the displacement demand imposed on multi-storey PCM cantilever walls.

This thesis is divided into four parts. The first part discusses the problem of using prestressed concrete for earthquake resistant structures, and examines the "equivalent viscous damping" approach to the measurement of structural damping. Part two describes the reversed cyclic testing of prestressed concrete members: end moments were applied to the members to simulate earthquake loading, and measurements were made of stiffness and damping energy. An analytical method of obtaining moment - rotation curves for prestressed members, taking account of curvature concentration at the tension cracks, is presented in part three. The last part is a study of the response of an idealized prestressed concrete structure to a recorded earthquake ground motion. The test results given in part two are used to define the properties of the idealized structure.

System-on-a-chip (SoC) designs have gained immense popularity as they provide designers with the ability of integrating all components (called IPs) of an application-specific computer system onto a single chip. However, one of the main bottlenecks of the SoC design cycle is the validation of complex designs. As system size grows, validation time increases beyond manageable limits. It is desirable that design inconsistences are found and fixed early in the design process, as validation overheads are significantly higher after IPs are integrated. This thesis presents a range of techniques for the automatic verification and design of SoCs that aim to reduce post-integration validation costs. Firstly, local module checking algorithm, a practical implementation of module checking, is presented. This technique allows for the comprehensive verification of IPs such that they guarantee the satisfaction of critical specifications regardless of the SoC they are used in. Local module checking is shown to be able to validate IPs in much lesser time on average than global module checking, and can help in handling many important validation tasks much before the integration stage. Next, a number of protocol conversion techniques that assist in the composition of IPs with incompatible protocols is presented. The inconsistencies between IP protocols, called mismatches, are bridged by the automatic generation of some extra glue-logic, called a converter. Converters generated by the proposed techniques can handle control, datawidth and clock mismatches between multiple IPs in a unified manner. These approaches ensure that the integration of IPs is correct-by-construction, such that the final system is guaranteed to satisfy key specifications without the need for further validation. Finally, a technique for automatic IP reuse using forced simulation is presented, which involves automatically generating an adaptor that guides an IP such that it satisfies desired specifications. The proposed technique can generate adaptors in many cases where existing IP techniques fail. As it is guaranteed that reused IPs satisfy desired specifications, post-integration validation costs are significantly reduced. For each proposed technique, a comprehensive set of results is presented that highlights the significance of the solution. It is noted that the proposed approaches can help automate SoC design and achieve significant savings in post-integration validation costs.

The aim of intensity modulated radiation therapy (IMRT) is to kill tumor cells while at the same time protecting the surrounding tissue and organs from the damaging effect of radiation. To achieve these goals computerized inverse planning systems are used. Given the number of beams and beam directions, beam intensity profiles that yield the best dose distribution under consideration of clinical and physical constraints are calculated. This is called beam intensity optimization problem. In this thesis, we first review existing mathematical models and computation methods for the beam intensity optimization problem. Next, we formulate the beam intensity optimization problem as a multiobjective linear programme (MOLP) with three objectives. For clinical cases this optimization problem involves thousands of variables and tens of thousands of constraints and existing methods such as multiobjective simplex methods can not handle it. The rest of the thesis is dedicated to developing methods to solve this large MOLP efficiently and to the application in the beam intensity optimization problem. Benson (1998c) argues that solving an MOLP in objective space needs less computation time than solving it in decision space if the number of objectives of the MOLP is much smaller than the number of variables. Moreover, the constraint matrix of the problem relies on the calculation of dose deposited in tissue. Since this calculation is always imprecise solving the MOLP exactly is not necessary in practice. This motivates us to develop algorithms for solving an MOLP in objective space approximately. We summarize Benson’s outer approximation algorithm for solving MOLPs in objective space and propose some small changes to improve computational performance. Then in order to approximate the true nondominated set we propose a modification of Benson’s algorithm which is called an approximation version of Benson’s algorithm. Our approximation algorithm computes an inner and an outer approximation of the nondominated set. We prove that the inner approximation provides a set of ε-nondominated points. The geometric duality theory of Heyde and L¨ohne (2006) defines a dual to an MOLP and it assures us to be able to find the nondominated set of the primal MOLP by solving its dual MOLP. Based on this we develop a dual variant of Benson’s outer approximation algorithm to solve the dual MOLP in objective space. We prove that solving the dual provides a weight set decomposition. We compare the primal algorithm and the dual algorithm on small illustrative and on radiotherapy examples. Furthermore, we propose an algorithm to solve the dual MOLP approximately but within specified tolerance. This approximate solution set can be used to calculate an approximation of the nondominated set of the primal MOLP.We show that this set is an ε-nondominated set of the original primal MOLP and provide numerical evidence that this approach can be faster than solving the primal MOLP approximately. Considering that the set of nondominated points is infinite, it is not very useful from the planners’ point of view. We address the problem of finding well distributed nondominated points for an MOLP.We propose a method which combines the global shooting and normal boundary intersection methods. By doing so, we overcome the limitation of normal boundary intersection method that parts of the nondominated set may be missed. Discrepancy analysis of the nondominated points from a geometry point of view shows that this method produces evenly distributed nondominated points. Moreover, the coverage error and the uniformity level can be measured. Finally, we apply the algorithms developed to the beam intensity optimization problem of 3D clinical cases with voxel size of 5mm and 3mm. A technique of reducing the resolution in normal tissue has been used to reduce the computation time. The results clearly illustrate the advantages of our methods.

The aim of intensity modulated radiation therapy (IMRT) is to kill tumor cells while at the same time protecting the surrounding tissue and organs from the damaging effect of radiation. To achieve these goals computerized inverse planning systems are used. Given the number of beams and beam directions, beam intensity profiles that yield the best dose distribution under consideration of clinical and physical constraints are calculated. This is called beam intensity optimization problem. In this thesis, we first review existing mathematical models and computation methods for the beam intensity optimization problem. Next, we formulate the beam intensity optimization problem as a multiobjective linear programme (MOLP) with three objectives. For clinical cases this optimization problem involves thousands of variables and tens of thousands of constraints and existing methods such as multiobjective simplex methods can not handle it. The rest of the thesis is dedicated to developing methods to solve this large MOLP efficiently and to the application in the beam intensity optimization problem. Benson (1998c) argues that solving an MOLP in objective space needs less computation time than solving it in decision space if the number of objectives of the MOLP is much smaller than the number of variables. Moreover, the constraint matrix of the problem relies on the calculation of dose deposited in tissue. Since this calculation is always imprecise solving the MOLP exactly is not necessary in practice. This motivates us to develop algorithms for solving an MOLP in objective space approximately. We summarize Benson’s outer approximation algorithm for solving MOLPs in objective space and propose some small changes to improve computational performance. Then in order to approximate the true nondominated set we propose a modification of Benson’s algorithm which is called an approximation version of Benson’s algorithm. Our approximation algorithm computes an inner and an outer approximation of the nondominated set. We prove that the inner approximation provides a set of ε-nondominated points. The geometric duality theory of Heyde and L¨ohne (2006) defines a dual to an MOLP and it assures us to be able to find the nondominated set of the primal MOLP by solving its dual MOLP. Based on this we develop a dual variant of Benson’s outer approximation algorithm to solve the dual MOLP in objective space. We prove that solving the dual provides a weight set decomposition. We compare the primal algorithm and the dual algorithm on small illustrative and on radiotherapy examples. Furthermore, we propose an algorithm to solve the dual MOLP approximately but within specified tolerance. This approximate solution set can be used to calculate an approximation of the nondominated set of the primal MOLP.We show that this set is an ε-nondominated set of the original primal MOLP and provide numerical evidence that this approach can be faster than solving the primal MOLP approximately. Considering that the set of nondominated points is infinite, it is not very useful from the planners’ point of view. We address the problem of finding well distributed nondominated points for an MOLP.We propose a method which combines the global shooting and normal boundary intersection methods. By doing so, we overcome the limitation of normal boundary intersection method that parts of the nondominated set may be missed. Discrepancy analysis of the nondominated points from a geometry point of view shows that this method produces evenly distributed nondominated points. Moreover, the coverage error and the uniformity level can be measured. Finally, we apply the algorithms developed to the beam intensity optimization problem of 3D clinical cases with voxel size of 5mm and 3mm. A technique of reducing the resolution in normal tissue has been used to reduce the computation time. The results clearly illustrate the advantages of our methods.

The aim of intensity modulated radiation therapy (IMRT) is to kill tumor cells while at the same time protecting the surrounding tissue and organs from the damaging effect of radiation. To achieve these goals computerized inverse planning systems are used. Given the number of beams and beam directions, beam intensity profiles that yield the best dose distribution under consideration of clinical and physical constraints are calculated. This is called beam intensity optimization problem. In this thesis, we first review existing mathematical models and computation methods for the beam intensity optimization problem. Next, we formulate the beam intensity optimization problem as a multiobjective linear programme (MOLP) with three objectives. For clinical cases this optimization problem involves thousands of variables and tens of thousands of constraints and existing methods such as multiobjective simplex methods can not handle it. The rest of the thesis is dedicated to developing methods to solve this large MOLP efficiently and to the application in the beam intensity optimization problem. Benson (1998c) argues that solving an MOLP in objective space needs less computation time than solving it in decision space if the number of objectives of the MOLP is much smaller than the number of variables. Moreover, the constraint matrix of the problem relies on the calculation of dose deposited in tissue. Since this calculation is always imprecise solving the MOLP exactly is not necessary in practice. This motivates us to develop algorithms for solving an MOLP in objective space approximately. We summarize Benson’s outer approximation algorithm for solving MOLPs in objective space and propose some small changes to improve computational performance. Then in order to approximate the true nondominated set we propose a modification of Benson’s algorithm which is called an approximation version of Benson’s algorithm. Our approximation algorithm computes an inner and an outer approximation of the nondominated set. We prove that the inner approximation provides a set of ε-nondominated points. The geometric duality theory of Heyde and L¨ohne (2006) defines a dual to an MOLP and it assures us to be able to find the nondominated set of the primal MOLP by solving its dual MOLP. Based on this we develop a dual variant of Benson’s outer approximation algorithm to solve the dual MOLP in objective space. We prove that solving the dual provides a weight set decomposition. We compare the primal algorithm and the dual algorithm on small illustrative and on radiotherapy examples. Furthermore, we propose an algorithm to solve the dual MOLP approximately but within specified tolerance. This approximate solution set can be used to calculate an approximation of the nondominated set of the primal MOLP.We show that this set is an ε-nondominated set of the original primal MOLP and provide numerical evidence that this approach can be faster than solving the primal MOLP approximately. Considering that the set of nondominated points is infinite, it is not very useful from the planners’ point of view. We address the problem of finding well distributed nondominated points for an MOLP.We propose a method which combines the global shooting and normal boundary intersection methods. By doing so, we overcome the limitation of normal boundary intersection method that parts of the nondominated set may be missed. Discrepancy analysis of the nondominated points from a geometry point of view shows that this method produces evenly distributed nondominated points. Moreover, the coverage error and the uniformity level can be measured. Finally, we apply the algorithms developed to the beam intensity optimization problem of 3D clinical cases with voxel size of 5mm and 3mm. A technique of reducing the resolution in normal tissue has been used to reduce the computation time. The results clearly illustrate the advantages of our methods.

The aim of intensity modulated radiation therapy (IMRT) is to kill tumor cells while at the same time protecting the surrounding tissue and organs from the damaging effect of radiation. To achieve these goals computerized inverse planning systems are used. Given the number of beams and beam directions, beam intensity profiles that yield the best dose distribution under consideration of clinical and physical constraints are calculated. This is called beam intensity optimization problem. In this thesis, we first review existing mathematical models and computation methods for the beam intensity optimization problem. Next, we formulate the beam intensity optimization problem as a multiobjective linear programme (MOLP) with three objectives. For clinical cases this optimization problem involves thousands of variables and tens of thousands of constraints and existing methods such as multiobjective simplex methods can not handle it. The rest of the thesis is dedicated to developing methods to solve this large MOLP efficiently and to the application in the beam intensity optimization problem. Benson (1998c) argues that solving an MOLP in objective space needs less computation time than solving it in decision space if the number of objectives of the MOLP is much smaller than the number of variables. Moreover, the constraint matrix of the problem relies on the calculation of dose deposited in tissue. Since this calculation is always imprecise solving the MOLP exactly is not necessary in practice. This motivates us to develop algorithms for solving an MOLP in objective space approximately. We summarize Benson’s outer approximation algorithm for solving MOLPs in objective space and propose some small changes to improve computational performance. Then in order to approximate the true nondominated set we propose a modification of Benson’s algorithm which is called an approximation version of Benson’s algorithm. Our approximation algorithm computes an inner and an outer approximation of the nondominated set. We prove that the inner approximation provides a set of ε-nondominated points. The geometric duality theory of Heyde and L¨ohne (2006) defines a dual to an MOLP and it assures us to be able to find the nondominated set of the primal MOLP by solving its dual MOLP. Based on this we develop a dual variant of Benson’s outer approximation algorithm to solve the dual MOLP in objective space. We prove that solving the dual provides a weight set decomposition. We compare the primal algorithm and the dual algorithm on small illustrative and on radiotherapy examples. Furthermore, we propose an algorithm to solve the dual MOLP approximately but within specified tolerance. This approximate solution set can be used to calculate an approximation of the nondominated set of the primal MOLP.We show that this set is an ε-nondominated set of the original primal MOLP and provide numerical evidence that this approach can be faster than solving the primal MOLP approximately. Considering that the set of nondominated points is infinite, it is not very useful from the planners’ point of view. We address the problem of finding well distributed nondominated points for an MOLP.We propose a method which combines the global shooting and normal boundary intersection methods. By doing so, we overcome the limitation of normal boundary intersection method that parts of the nondominated set may be missed. Discrepancy analysis of the nondominated points from a geometry point of view shows that this method produces evenly distributed nondominated points. Moreover, the coverage error and the uniformity level can be measured. Finally, we apply the algorithms developed to the beam intensity optimization problem of 3D clinical cases with voxel size of 5mm and 3mm. A technique of reducing the resolution in normal tissue has been used to reduce the computation time. The results clearly illustrate the advantages of our methods.

Liquid Composite Moulding (LCM) processes are commonly used techniques for the manufacture of fibre reinforced plastic components. The range of LCM processes addresses manufacturing scenarios from low to high volume. This study explores the potential of natural fibres as reinforcement for LCM preforms, considering discontinuous fibre mats produced using several novel methods. Modified paper manufacturing techniques were employed to produce two types of wet formed wood fibre mats, the other two being manufactured using dry methods. The natural fibre reinforcements considered in this study have been characterised with regard to permeability and compaction response, such that their application to a wide range of LCM processes can be evaluated. Reinforcement permeability and compaction response data are required to simulate LCM processes. Permeability of all four types of wood fibre mats was measured as a function of fibre volume fraction. The dry compaction response of these mats has been investigated, with the compression loads being measured up to a fibre volume fraction of 0.4. A complex non-elastic compression response was observed which has significant influence on forces generated within moulds. Saturated compaction tests were also carried out, the samples infiltrated with a test fluid (mineral oil). These results were compared to typical glass fibre mats used for LCM processes. It was found that the wood fibre mats have permeability two orders of magnitude lower, and required significantly larger force to compact to at similar fibre volume fractions as compared to glass fibre reinforcements. Model fluids are used extensively for LCM characterisation experiments because of ease of handling and chemical stability. The influence of test fluid type on permeability and compaction response of four manufactured wood fibre mats has been determined using two different test fluids, a water based polar fluid (glucose syrup) and a non polar fluid (mineral oil). It was found that glucose syrup caused fibre softening and swelling which reduced the required compaction loads, and permeability of a wood fibre mat. On the other hand, mineral oil did not cause any fibre softening and swelling. The effect of geometrical parameters such as reinforcing fibre bundle diameter and length on characterisation was also determined. Six different types of flax fibre yarn mats were manufactured. A series of compaction tests were carried out on both dry and saturated samples. Saturated permeability was also measured at a number of fibre volume fraction levels. The fibre bundling reduced compaction forces and increased permeability of a mat. Composite panels were manufactured using an epoxy resin to visualise the penetration of resin into yarns and fibre cells. The reinforcement permeability and compaction response data were used to model two LCM variants, Resin Transfer Moulding (RTM), and Injection Compression Moulding (I/CM). A consolidation model approach was applied to simulate both RTM and I/CM processes, addressing a simple mould geometry. The RTM and I/CM clamping force traces, flow rates, and gate pressures were also measured. The simulation results have been compared with experiments completed for wood and glass fibre reinforcements at three different fibre volume fractions. It was found that at similar fibre volume fractions, the wood fibre mats produced longer mould filling times, and required larger forces to compact.

The aim of intensity modulated radiation therapy (IMRT) is to kill tumor cells while at the same time protecting the surrounding tissue and organs from the damaging effect of radiation. To achieve these goals computerized inverse planning systems are used. Given the number of beams and beam directions, beam intensity profiles that yield the best dose distribution under consideration of clinical and physical constraints are calculated. This is called beam intensity optimization problem. In this thesis, we first review existing mathematical models and computation methods for the beam intensity optimization problem. Next, we formulate the beam intensity optimization problem as a multiobjective linear programme (MOLP) with three objectives. For clinical cases this optimization problem involves thousands of variables and tens of thousands of constraints and existing methods such as multiobjective simplex methods can not handle it. The rest of the thesis is dedicated to developing methods to solve this large MOLP efficiently and to the application in the beam intensity optimization problem. Benson (1998c) argues that solving an MOLP in objective space needs less computation time than solving it in decision space if the number of objectives of the MOLP is much smaller than the number of variables. Moreover, the constraint matrix of the problem relies on the calculation of dose deposited in tissue. Since this calculation is always imprecise solving the MOLP exactly is not necessary in practice. This motivates us to develop algorithms for solving an MOLP in objective space approximately. We summarize Benson’s outer approximation algorithm for solving MOLPs in objective space and propose some small changes to improve computational performance. Then in order to approximate the true nondominated set we propose a modification of Benson’s algorithm which is called an approximation version of Benson’s algorithm. Our approximation algorithm computes an inner and an outer approximation of the nondominated set. We prove that the inner approximation provides a set of ε-nondominated points. The geometric duality theory of Heyde and L¨ohne (2006) defines a dual to an MOLP and it assures us to be able to find the nondominated set of the primal MOLP by solving its dual MOLP. Based on this we develop a dual variant of Benson’s outer approximation algorithm to solve the dual MOLP in objective space. We prove that solving the dual provides a weight set decomposition. We compare the primal algorithm and the dual algorithm on small illustrative and on radiotherapy examples. Furthermore, we propose an algorithm to solve the dual MOLP approximately but within specified tolerance. This approximate solution set can be used to calculate an approximation of the nondominated set of the primal MOLP.We show that this set is an ε-nondominated set of the original primal MOLP and provide numerical evidence that this approach can be faster than solving the primal MOLP approximately. Considering that the set of nondominated points is infinite, it is not very useful from the planners’ point of view. We address the problem of finding well distributed nondominated points for an MOLP.We propose a method which combines the global shooting and normal boundary intersection methods. By doing so, we overcome the limitation of normal boundary intersection method that parts of the nondominated set may be missed. Discrepancy analysis of the nondominated points from a geometry point of view shows that this method produces evenly distributed nondominated points. Moreover, the coverage error and the uniformity level can be measured. Finally, we apply the algorithms developed to the beam intensity optimization problem of 3D clinical cases with voxel size of 5mm and 3mm. A technique of reducing the resolution in normal tissue has been used to reduce the computation time. The results clearly illustrate the advantages of our methods.