Topology optimisation techniques help designers to nd the best layout of structural members. When followed by shape and sizing optimisation, these techniques result in far greater savings than shape and sizing optimisation alone. During the last three decades extensive research has been carried out in the topology optimisation area. Consequently topology optimisation techniques have been considerably improved and successfully applied to a range of physical problems. These techniques are now regarded as invaluable tools in mechanical, aerostructural and structural design. In spite of great potential in geomechanical problems, however, the application of topology optimisation techniques in this eld has not been studied thoroughly. This thesis explores the state-of-the-art topology and shape optimisation methods in excavation design. The main problems of concern in this thesis are to nd the optimum shape of an underground opening and to optimise the reinforcement distribution around it. To tackle these problems, new formulations for some topology optimisation techniques are proposed in this thesis to match the requirements in excavation problems. Although linear elastic material models have limited applications in excavation design, these models are used in the rst part of this thesis to introduce the proposed optimisation technique and to verify it. Simultaneous shape and reinforcement optimisation is considered as well. Using the proposed optimisation techniques, it is shown that the computational effort needed for this mixed optimisation problem is almost the same as the effort required to solve each of shape or reinforcement optimisation problems alone. In the next part of this thesis, reinforcement optimisation of tunnels in massive rocks is addressed where the behaviour of the rock mass is in uenced by few major discontinuities. Although discontinuities exist in the majority of rock masses, due to its complexities, optimising the excavations in these types of rocks has not been considered by any other researcher before. A method for reinforcement optimisation of tunnels in such rock masses is proposed in this thesis and its capability is demonstrated by means of numerical examples. Lastly, shape optimisation of excavations in elasto-plastic soil is addressed. In this problem the excavation sequence is also taken into account. A stressbased parameter is dened to evaluate the efficiency of the soil elements assuming Mohr-Coulomb material model. Some examples are solved to illustrate and verify the application of the proposed technique. Being one of the rst theses on the topic, this work concentrates on the theoretical background and the possibility of applying topology optimisation techniques in excavation designs. It has been demonstrated that a properly tailored topology optimisation technique can be applied to nd both the optimum shape and the optimum reinforcement design of openings. Optimising the excavations in various types of grounds including elastic homogeneous rock masses, massive rocks, and elasto-plastic soil and rock media have been considered. Different objective functions, namely, mean compliance, oor heave, and tunnel convergence have been selected and successfully minimised using the proposed techniques. The results obtained in this thesis illustrate that the proposed topology optimisation techniques are very useful for improving excavation designs.

Pharmaceutical tablets have been the most dominant form for drug delivery and they need to be strong enough to withstand external stresses due to packaging and loading conditions before use. The strength of the produced tablets, which is characterised by their compressibility and compactibility, is usually deter-mined through a physical prototype. This process is sometimes quite expensive and time consuming. Therefore, simulating this process before hand can over-come this problem. A technique for shape modelling of pharmaceutical tablets based on the use of Partial Differential Equations is presented in this thesis. The volume and the sur-face area of the generated parametric tablet in various shapes have been es-timated numerically. This work also presents an extended formulation of the PDE method to a higher dimensional space by increasing the number of pa-rameters responsible for describing the surface in order to generate a solid tab-let. The shape and size of the generated solid tablets can be changed by ex-ploiting the analytic expressions relating the coefficients associated with the PDE method. The solution of the axisymmetric boundary value problem for a finite cylinder subject to a uniform axial load has been utilised in order to model a displace-ment component of a compressed PDE-based representation of a flat-faced round tablet. The simulation results, which are analysed using the Heckel model, show that the developed model is capable of predicting the compressibility of pharmaceutical powders since it fits the experimental data accurately. The opti-mal design of pharmaceutical tablets with particular volume and maximum strength has been obtained using an automatic design optimisation which is performed by combining the PDE method and a standard method for numerical optimisation.

Engineering design often requires the optimisation of multiple objectives, and becomes significantly more difficult and time consuming when the response surfaces are multimodal, rather than unimodal. A surrogate model, also known as a metamodel, can be used to replace expensive computer simulations, accelerating single and multi-objective optimisation and the exploration of new design concepts. The main research focus of this work is to investigate the use of a neural network surrogate model to improve optimisation of multimodal surfaces. Several significant contributions derive from evaluating the Cascade Correlation neural network as the basis of a surrogate model. The contributions to the neural network community ultimately outnumber those to the optimisation community. The effects of training this surrogate on multimodal test functions are explored. The Cascade Correlation neural network is shown to map poorly such response surfaces. A hypothesis for this weakness is formulated and tested. A new subdivision technique is created that addresses this problem; however, this new technique requires excessively large datasets upon which to train. The primary conclusion of this work is that Cascade Correlation neural networks form an unreliable basis for a surrogate model, despite successes reported in the literature. A further contribution of this work is the enhancement of an open source optimisation toolkit, achieved by the first integration of a truly multi-objective optimisation algorithm.

Pharmaceutical tablets have been the most dominant form for drug delivery and they need to be strong enough to withstand external stresses due to packaging and loading conditions before use. The strength of the produced tablets, which is characterised by their compressibility and compactibility, is usually deter-mined through a physical prototype. This process is sometimes quite expensive and time consuming. Therefore, simulating this process before hand can over-come this problem. A technique for shape modelling of pharmaceutical tablets based on the use of Partial Differential Equations is presented in this thesis. The volume and the sur-face area of the generated parametric tablet in various shapes have been es-timated numerically. This work also presents an extended formulation of the PDE method to a higher dimensional space by increasing the number of pa-rameters responsible for describing the surface in order to generate a solid tab-let. The shape and size of the generated solid tablets can be changed by ex-ploiting the analytic expressions relating the coefficients associated with the PDE method. The solution of the axisymmetric boundary value problem for a finite cylinder subject to a uniform axial load has been utilised in order to model a displace-ment component of a compressed PDE-based representation of a flat-faced round tablet. The simulation results, which are analysed using the Heckel model, show that the developed model is capable of predicting the compressibility of pharmaceutical powders since it fits the experimental data accurately. The opti-mal design of pharmaceutical tablets with particular volume and maximum strength has been obtained using an automatic design optimisation which is performed by combining the PDE method and a standard method for numerical optimisation.

Engineering design often requires the optimisation of multiple objectives, and becomes significantly more difficult and time consuming when the response surfaces are multimodal, rather than unimodal. A surrogate model, also known as a metamodel, can be used to replace expensive computer simulations, accelerating single and multi-objective optimisation and the exploration of new design concepts. The main research focus of this work is to investigate the use of a neural network surrogate model to improve optimisation of multimodal surfaces. Several significant contributions derive from evaluating the Cascade Correlation neural network as the basis of a surrogate model. The contributions to the neural network community ultimately outnumber those to the optimisation community. The effects of training this surrogate on multimodal test functions are explored. The Cascade Correlation neural network is shown to map poorly such response surfaces. A hypothesis for this weakness is formulated and tested. A new subdivision technique is created that addresses this problem; however, this new technique requires excessively large datasets upon which to train. The primary conclusion of this work is that Cascade Correlation neural networks form an unreliable basis for a surrogate model, despite successes reported in the literature. A further contribution of this work is the enhancement of an open source optimisation toolkit, achieved by the first integration of a truly multi-objective optimisation algorithm.

Seeking to quench airliners’ unending thirst for lightweight, reliable and more comfortable seating solutions, designers are developing a new generation of slim economy – class seats. Challenge in front of the designers is to carve out additional “living space”, as well as to give a “lie – flat” experience to air travellers with strict adherence to safety regulations. Present research tries to address all these industry needs through an innovative and novel “Sleep Seat”. A generous angle of recline (40 degree), movement of “Seat Pan” along the gradient, fixed outer shell of backrest, and unique single “Forward Beam” design distinguishes “Sleep Seat” form current generation seats. It is an ultralightweight design weighing 8kg (typical seat weight is 11kg). It satisfies “Generic Requirements (GR2)” which ensures “Comfort in Air”. It will be a “16g” seat, means it can sustain the “Emergency landing” loads as specified by “Certification Specifications (CS 25.561 and CS 25.562)”. For present research, only CS 25.561 has been considered. Since, the design of “Sleep Seat” is still in its conceptual phase, it is not possible to build the prototypes and their physical testing, due to costs and time involved. “Finite Element Analysis (FEA)” is a useful tool to predict the response of the structure when subjected to real life loads. Hence, the aim of research being undertaken is to develop a detailed FE model of the complete seat structure, which will help designers to identify potential weak areas and to compare different design concepts virtually, thereby reducing the development cycle time. In order to avoid handling of large number of design variables; major load carrying members (called Primary Load Path) i.e. Forward beam and leg; are designed for the most critical “Forward 9g” loads; using FEA results as a basis. A robust framework to verify the FEA results is developed. “Sequential Model Development Approach”; which builds the final, detailed FE model starting from preliminary model (by continuously updating the FE model by addition of details that are backed up by pilot studies); resulted in a FE model which could predict the stress induced in each of the components for applied CS 25.561 loads along with “Seat Interface Loads”. The “Interface Load” is the force exerted by the seat design on the floor and is one of the main contributing factors in seat design. “Optistruct” is used as a solver for linear static FEA, whereas “Abaqus / Standard” is used for non-linear FEA. Stepwise methodologies for mesh sensitivity study, modelling of bolt-preload, representing bolted joint in FEA, preventing rigid body motion, and obtaining a converged solution for non-linear FEA are developed during this research. Free-Shape Optimisation is used to arrive at a final design of Seat-leg. All the findings and steps taken during this are well documented in this report. Finally, a detailed FE model (involving all the three non-linearities : Contact, material and geometric) of the complete seat structure was analysed for the loads taken from CS 25.561, and it was found that design of “Forward beam” and leg are safe against CS 25.561. Therefore, all the aims and objectives outlined for this research were accomplished. For future work, first area to look for, would be validation of present FEA results by experimental testing. FE model to simulate dynamic loads CS 25.562 can be developed followed by design improvements and optimisation.

Doctor of Philosophy / We consider the problem of minimising the eigenvalues of the Laplacian with Robin boundary conditions $\frac{\partial u}{\partial \nu} + \alpha u = 0$ and generalised Wentzell boundary conditions $\Delta u + \beta \frac{\partial u}{\partial \nu} + \gamma u = 0$ with respect to the domain $\Omega \subset \mathbb R^N$ on which the problem is defined. For the Robin problem, when $\alpha > 0$ we extend the Faber-Krahn inequality of Daners [Math. Ann. 335 (2006), 767--785], which states that the ball minimises the first eigenvalue, to prove that the minimiser is unique amongst domains of class $C^2$. The method of proof uses a functional of the level sets to estimate the first eigenvalue from below, together with a rearrangement of the ball's eigenfunction onto the domain $\Omega$ and the usual isoperimetric inequality. We then prove that the second eigenvalue attains its minimum only on the disjoint union of two equal balls, and set the proof up so it works for the Robin $p$-Laplacian. For the higher eigenvalues, we show that it is in general impossible for a minimiser to exist independently of $\alpha > 0$. When $\alpha < 0$, we prove that every eigenvalue behaves like $-\alpha^2$ as $\alpha \to -\infty$, provided only that $\Omega$ is bounded with $C^1$ boundary. This generalises a result of Lou and Zhu [Pacific J. Math. 214 (2004), 323--334] for the first eigenvalue. For the Wentzell problem, we (re-)prove general operator properties, including for the less-studied case $\beta < 0$, where the problem is ill-posed in some sense. In particular, we give a new proof of the compactness of the resolvent and the structure of the spectrum, at least if $\partial \Omega$ is smooth. We prove Faber-Krahn-type inequalities in the general case $\beta, \gamma \neq 0$, based on the Robin counterpart, and for the ``best'' case $\beta, \gamma > 0$ establish a type of equivalence property between the Wentzell and Robin minimisers for all eigenvalues. This yields a minimiser of the second Wentzell eigenvalue. We also prove a Cheeger-type inequality for the first eigenvalue in this case.

Doctor of Philosophy / We consider the problem of minimising the eigenvalues of the Laplacian with Robin boundary conditions $\frac{\partial u}{\partial \nu} + \alpha u = 0$ and generalised Wentzell boundary conditions $\Delta u + \beta \frac{\partial u}{\partial \nu} + \gamma u = 0$ with respect to the domain $\Omega \subset \mathbb R^N$ on which the problem is defined. For the Robin problem, when $\alpha > 0$ we extend the Faber-Krahn inequality of Daners [Math. Ann. 335 (2006), 767--785], which states that the ball minimises the first eigenvalue, to prove that the minimiser is unique amongst domains of class $C^2$. The method of proof uses a functional of the level sets to estimate the first eigenvalue from below, together with a rearrangement of the ball's eigenfunction onto the domain $\Omega$ and the usual isoperimetric inequality. We then prove that the second eigenvalue attains its minimum only on the disjoint union of two equal balls, and set the proof up so it works for the Robin $p$-Laplacian. For the higher eigenvalues, we show that it is in general impossible for a minimiser to exist independently of $\alpha > 0$. When $\alpha < 0$, we prove that every eigenvalue behaves like $-\alpha^2$ as $\alpha \to -\infty$, provided only that $\Omega$ is bounded with $C^1$ boundary. This generalises a result of Lou and Zhu [Pacific J. Math. 214 (2004), 323--334] for the first eigenvalue. For the Wentzell problem, we (re-)prove general operator properties, including for the less-studied case $\beta < 0$, where the problem is ill-posed in some sense. In particular, we give a new proof of the compactness of the resolvent and the structure of the spectrum, at least if $\partial \Omega$ is smooth. We prove Faber-Krahn-type inequalities in the general case $\beta, \gamma \neq 0$, based on the Robin counterpart, and for the ``best'' case $\beta, \gamma > 0$ establish a type of equivalence property between the Wentzell and Robin minimisers for all eigenvalues. This yields a minimiser of the second Wentzell eigenvalue. We also prove a Cheeger-type inequality for the first eigenvalue in this case.

This thesis describes the variety and the width of issues connected with crankshafts of combustion engines. It distinguishes different factors, which influence the shape and size of these crankshafts and the selected findings are then transferred into the design of crankshaft for the particular engine. The thesis also presents coherent description of application of topology optimization to the particular crankshaft and also different, more complex design procedures leading to the optimal shape of crankshaft are afterwards suggested. Quality of such a design is then verified by comparative analysis of fatigue of the crankshaft.

Hydrodynamic vortex separators (HDVSs) are used in wastewater treatment to separate solids from wastewater. The aim of this research is to devise a CFD-based methodology that optimises their performance through modification of their design. A validation study is performed to assess whether OpenFOAM can be used to reliably model the flow of water in an HDVS. The results of the simulations are compared with experimental readings, showing a good fit when the appropriate boundary layer height and turbulence model are used. The continuous adjoint method is employed to derive the adjoint equations, associated with the drift flux equations used to model the flow of wastewater. They are specialised to the typical boundary conditions of ducted flows and are coded using OpenFOAM. An optimal design is found for boundary conditions, corresponding to typical values used in practice, and is shown to improve the performance of a simplified initial design by 40%. This optimal design is subsequently subjected to a different hydraulic loading rate and dispersed-phase volume fraction at the inlet, to assess the performance variation in these circumstances. Though the optimal design removes all the solids when the dispersed-phase fraction is reduced at the inlet, initial results suggest that the design is sensitive to hydraulic loading rate and further tests are recommended before drawing more explicit conclusions. This is the first time the adjoint drift flux equations have been derived. It is also the first time they have been coded and applied to an HDVS to optimise its performance. The methodology developed in this thesis could be applied to any device that separates solids from liquid or two immiscible liquids, in order to optimise its performance.