L’optimisation des transferts de chaleur est un enjeu de recherche important pour améliorer les performances des systèmes énergétiques, notamment des échangeurs de chaleur. Cette optimisation peut s’appuyer sur différents leviers, comme le choix du régime d’écoulement, ou la modification de la géométrie de l’échangeur. C’est sur ce deuxième angle que nous avons travaillé au cours de cette thèse, en proposant une méthode d’optimisation topologique. Le nombre de paramètres impliqués dans cette méthode est important (plusieurs milliers). Une méthode à gradient est donc employée. Le calcul du gradient de la fonction de coût en fonction des paramètres de design est facilité par l’emploi d’une méthode adjointe. Le problème direct est résolu à l’aide de la méthode de Lattice Boltzmann (LBM). L’implémentation aisée des conditions aux limites dans l’algorithme LBM est un bénéfice pour l’optimisation topologique. De plus, la formulation de la LBM est explicite et hautement parallélisable, notamment sur les cartes graphiques (GPU), utilisées au cours de cette thèse. Ensuite le domaine d’optimisation est composé de fluide et de solide. Leur distribution est définie par une fonction Level- Set (LSF). Cette fonction est continue et le contour zéro définit précisément l’interface fluide/solide. La méthode d’optimisation développée a été testée et validée pour différents objectifs (minimisation de la température moyenne, maximisation de l’échange de chaleur) et contraintes (limitation des pertes de charges, porosité fixée). / Heat transfer enhancement is an important research area to improve the efficiency of energy systems, especially for heat exchangers. There are different ways of optimizing such systems as the choice of flow regime or the modification on the fluid channels geometry. This thesis is focused on the latter option with the development of a topology optimization method. The number of design parameters involved in this technique is important (several thousands) so a gradient-based method is employed. The calculation of the cost function gradient with respect to the design variables is done with an adjoint-state method. The forward problem is solved with the Lattice Boltzmann Method (LBM). The simple implementation of the LBM boundary conditions is an interesting feature for topology optimization. The LBM algorithm is also highly parallelizable and GPU cards have been used in this thesis to obtain fast computational times. The efficiency of the LBM is important because the forward problem must be solved at each optimization step. Then, the optimization domain is composed of either solid of fluid elements. This material distribution is defined by a Level-Set Function (LSF). This is a continuous function in which the zero contour defines the fluid/solid interface, allowing an accurate description of both domains. The proposed optimization method has been tested for different objectives (minimization of the mean temperature, maximization of heat transfer rate) and constraints (pressure drops limitation, fixed porosity).

A significant limitation of the conventional level set method in topology optimization is that it can not create new holes in the design domain. Therefore, the topological derivative approach is proposed to overcome this problem. In this part of the thesis, we investigated the use of the topological derivative in combination with the level set method for topology optimization of solid structures. The topological derivative can indicate the appropriate location to create new holes so that the strong dependency of the optimal topology on the initial design can be alleviated. We also develop an approach to evolve the level set function by replacing the gradient item with a Delta function in the standard Hamilton-Jacobi equation. We find that this handling can create new holes in the solid domain, grow a structure from an empty domain, and improve the convergence rate of the optimization process. The success of our approach is demonstrated by several numerical examples. / Following those methods some numerical implementation issues are discussed, and numerical examples of 2D structural topology optimization problems of minimum compliance design are given and combined with a comparative study where the efficiency, convergence and accuracy of the present methods are highlighted. Finally, conclusions are given. / In the second part of this thesis, we implement another variational level set method, the piecewise constant level set (PCLS) method. This method was first proposed by Lie-Lysaker-Tai in the interface problem field for such tasks as image segmentation and denoising problems. In this approach, by defining a piecewise density function over the whole design domain, the sensitivity of the objective function in respect to the design variable, the level set surface, can be explicitly obtained. Thus, the piecewise density function can be viewed as a bridge establishing the relationship between the implicit level set function and the performance function defined on the design domain. This proposed method retains the advantages of the implicit level set representation, such as the capability of the interface to develop sharp corners, break apart and merge together in a flexible manner. Because the PCLS method is implemented by an implicit iteration differential scheme rather than solving the Hamilton-Jacobi equation, it is not only free of the CFL condition and the reinitialization scheme, but it is also easy to implement. These favorable properties lead to a great timesaving advantage over the conventional level set method. Two other meaningful advantages are the natural nucleation property with which the proposed PCLS method need not incorporate any artificial nucleation scheme and the dependence of the initial design is greatly alleviated. / In the third part of this thesis, we apply a parametric scheme by combining the conventional level set method with radial basis functions (RBFs). This method is introduced because the conventional level set function has no analytical form then the entire design domain must be made discrete in an artificial manner using a rectilinear grid for level set processing - often through a distance transform. The classical level set method for structural topology optimization requires a careful choice of an upwind scheme, extension velocity and a reinitialization algorithm. With the versatile tool, RBF, the original problem can be converted to a parametric optimization problem. Therefore, the costly Hamilton-Jacobi PDE solving procedure can be easily replaced by a standard gradient method or another mature conventional optimization method in the parameter space such as MMA, OC, mathematic programming and so on. / Keywords: structural optimization, level set method, topological derivative, radial basis functions, piecewise constant level set method. (Abstract shortened by UMI.) / The concept of structural optimization has been more and more widely accepted in many engineering fields during the past several decades, because the optimization can result in a much more reasonable and economical structure design with even less material consumption. / Wei Peng. / "June 2007." / Adviser: Yu Michael Wang. / Source: Dissertation Abstracts International, Volume: 69-01, Section: B, page: 0640. / Thesis (Ph.D.)--Chinese University of Hong Kong, 2007. / Includes bibliographical references (p. 166-180). / Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Electronic reproduction. [Ann Arbor, MI] : ProQuest Information and Learning, [200-] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Abstracts in English and Chinese. / School code: 1307.

Another interesting area we investigate is topology optimization with geometric control. Our initial research interests in topology optimization with geometric width control were motivated by the challenge of how to get more manufacturable compliant mechanism designs for MEMS devices. Considering MEMS fabrication technologies, say LIGA, it is natural that designs with specified feature width are more preferable with respect to those with free-form geometries. We propose a variational approach to this problem. A novel quadratic energy functional is employed to govern the geometric feature width of the design. This geometry describing functional is added to the performance-describing functionals. In this way both the performance and the geometric width of the design are optimized simultaneously. The preliminary results show that this method is capable of generating strip-like (or beam) designs with specified feature width, which is a highly desirable characteristic and uniquely distinguishes the proposed method. / Compliant mechanisms are involved in many applications both in the macro and in the micro world. But for a long time, the design procedure of compliant mechanisms was rather a handicraft than a technology. The conventional way is made on an ad hoc basis which to a large extent depends on the designers intuition, experience and inspiration. The limitations of such a trial-and-error approach are obvious: it is not always guaranteed to work, especially when the design is very complicated or when topology and multi-material problems are taken into account. The practical design and application of compliant mechanisms are in need of a systematic approach to create conceptual design. Here, we take a level-set-based new approach to solve this problem. / Considering the actual requirements on reliability, we also investigate how to get conceptual designs with distributed compliance, which is the core part of this thesis. We find the intrinsic defects in the widely used spring model and prove that it will inevitably cause designs with both large output displacements and low strain energies. We will show low strain energy does not guarantee high stiffness. To evenly distribute the compliance, we propose a new method considering the "characteristic stiffness" at interested points. In this way, the strength (stiffness) at the output port of the system is involved into the objective function and optimized directly. This new method is applied to some benchmark examples of both structure optimization and compliant mechanism optimization to validate its performance. / In our proposed method, the compliant mechanism design problem is recast as an infinite dimensional optimization problem, where the design variable is the geometric shape of the compliant mechanism and the goal is to find a suitable shape in the admissible design space so that the objective functional can reach a minimum. The geometric shape of the compliant mechanism is represented as the zero level set of a one-higher dimensional level set function, and the dynamic variations of the shape are governed by the Hamilton-Jacobi partial differential equation. The application of level set methods endows the optimization process with the particular quality that topological changes of the boundary, such as merging or splitting, can be handled in a natural fashion. By making a connection between the velocity fields in the Hamilton-Jacobi partial differential equation with the shape gradient of the objective functional, we go further to transform the optimization problem into that of finding a steady state solution of the partial differential equation. / Our research follows the route from ease to difficulty, reflecting our understanding of the compliant mechanism design problem at different stages. The first problem addressed in this thesis is how to maintain the structural connectivity during the topology optimization process. De facto hinges are known to be a fairly typical phenomenon in topology optimization of compliant mechanisms; they represent highly localized compliance regions. A most adverse side effect caused by de facto hinges is that they are prone to cause a structurally disconnected design, especially to that with a low volume ratio. To solve this problem, a digital topological connectivity scheme is integrated within the level set model, which ensures connectivity while allows topology changes. This is our first step in the research process. / Chen, Shikui. / "January 2007." / Adviser: Michael Yu Wang. / Source: Dissertation Abstracts International, Volume: 68-08, Section: B, page: 5513. / Thesis (Ph.D.)--Chinese University of Hong Kong, 2007. / Includes bibliographical references (p. 149-162). / Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Electronic reproduction. [Ann Arbor, MI] : ProQuest Information and Learning, [200-] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Abstracts in English and Chinese. / School code: 1307.

Wireless Sensor Networks (WSNs), consisting of many small sensing devices working in concert, have the potential to revolutionise every aspect of our lives. Although the technology is still in its infancy offers an unlimited number of possible applications, ranging from military surveillance to environmental monitoring. These WSNs are prone to physical sensor failures due to environmental conditions such bio fouling and an adverse ambient environment, as well as threats that arise from their operation in an open environment. Consequently, reliability and fault-tolerance techniques become a critical aspect of the research associated with WSNs. In mission critical applications, such as the monitoring of enemy troops, unreliable or faulty information produced by WSNs could potentially lead to fatal outcomes. In such applications, it necessary to receive both a correct notification of event occurrences and uncorrupted data. Developing a fault-tolerance system for WSNs is a challenging task. New self-configuration, self-recognition and self-organisation techniques are needed due to unique aspects of the operation of WSNs. Our current understanding of WSNs leads to an immunologically inspired solution to the design of a fault-tolerant network. One of the main roles of the Natural Immune System(NIS) is the recognition of self and the elimination of non-self proteins. Hence, in order to have an immune system equivalent for a sensor network, we must have a clear and stable definition of what constitutes the Self and the Non-Self Sets in a sensor network. This thesis explores two different approaches to modelling, collection and representation of the Self-Set in distributed sensor networks. We approach this problem, of identifying what constitutes the Self-Set in terms of sensor readings, using pattern recognition techniques from the machine learning field that leverages a small number of past observations of sensor nodes. We have chosen Competitive Learning Neural Network (CLNN) for the construction of the Self-Set. We define and evaluate two approaches for the aggregation of the Self-Set across multiple sensors in a WSN. The first approach is the Graph Theory Based Aggregation (GTBA) which consists of two main parts, namely: classification of the sensor readings by means of CLNN, which provides the multimodal view data and GTBA of the CLNN output, which takes intersections of intervals produced by CLNN. In this thesis we define and evaluate two different interpretations of GTBA, namely: Midpoint Intersection (MPI): one that considers the midpoint of intervals. Midpoint Free Intersection (MFI): one that does not take the midpoints into account but assigns the confidence levels to each of the resulted intersections. We evaluated both interpretations on three different types of phenomena and have shown that the second interpretation, MFI, consistently produced more precise representations of the environment under observation. However, MFI produced a very strict representation of the phenomenon, which consequently led to a large number of systems' retraining. Hence, we defined and evaluated a technique which produced a more relaxed representation of the Self-Set and at the same time preserved the finer variation in the phenomenon. The second approach is based on unsupervised learning. We define and evaluate three related unsupervised learning procedures ?? Divergence and Merging (DMP), Suboptimal Clustering (SOC), and Simple Clustering (SC) for the collection of the Self-Set. We explore the design tradeoffs in unsupervised learning schemes with respect to the clustering quality. We implement and evaluate these related unsupervised learning procedures on a realworld data set. The outcome of these experiments show that, out of the three unsupervised learning procedures studied in this thesis, the Suboptimal Clustering procedure appears to be the most suitable for the classification of sensor readings, provided that the amount of free memory is large enough to store and recluster an entire training set. We evaluate aggregation of the Self-Set produced by means of the distributed implementation of the unsupervised learning procedures. The aggregation is based on extended unsupervised learning and we evaluate the possibilities of the autonomous retraining of the system. Our experiments show that, in a naturally slowly changing environment, 40% of nodes reporting deviations is a large enough number to reinitialise the retraining of the system. The final conclusion is that it is possible to have a distributed implementation of the unsupervised procedure that produces an almost identical representation of the environment, which makes unsupervised learning suitable for a large number of sensor network architectures.

Doctor of Philosphy (PhD) / The term boset was coined by Patrick Jordan, both as an abbreviation of biordered set, and as a generalisation of poset, itself an abbreviation of partially ordered set. A boset is a set equipped with a partial multiplication and two intertwining reflexive and transitive arrow relations which satisfy certain axioms. When the arrow relations coincide the boset becomes a poset. Bosets were invented by Nambooripad (in the 1970s) who developed his own version of the theory of fundamental regular semigroups, including the classical theory of fundamental inverse semigroups using semilattices, due to Munn (in the 1960s). A semigroup is fundamental if it cannot be shrunk homomorphically without collapsing its skeleton of idempotents, which is a boset. Nambooripad constructed the maximum fundamental regular semigroup with a given boset of idempotents. Fundamental semigroups and bosets are natural candidates for basic building blocks in semigroup theory because every semigroup is a coextension of a fundamental semigroup in which the boset of idempotents is undisturbed. Recently Jordan reproved Nambooripad's results using a new construction based on arbitrary bosets. In this thesis we prove that this construction is always fundamental, which was previously known only for regular bosets, and also that it possesses a certain maximality property with respect to semigroups which are generated by regular elements. For nonregular bosets this constuction may be regular or nonregular. We introduce a class of bosets, called sawtooth bosets, which contain many regular and nonregular examples, and correct a criterion of Jordan's for the regularity of this construction for sawtooth bosets with two teeth. We also introduce a subclass, called cyclic sawtooth bosets, also containing many regular and nonregular examples, for which the construction is always regular.

On a bounded domain of $IR^N$, we are interested in the nonlinear Schrödinger problem $-Delta u + V(x)u = vert uvert^{p-2}u$ submitted to the Dirichlet boundary conditions or Neumann boundary conditions. This equation has many interests in astrophysics and quantum mechanics. Depending on the domain and the potential $V$, we are studying numerically (by making and computing algorithms) and theoretically the structure of ground state (resp. least energy nodal) solution, i.e. one-signed (resp. sign-changing) solutions with minimal energy. We prove some symmetry and symmetry breaking results and make a lot of conjectures. We also pay attention to the $p$-Laplacian case and we change the nonlinearity $vert uvert^{p-2}u$.

With the relentless shrinking of the minimum feature size of VLSI Integrated Circuits (ICs), reduction in operating voltages and increase in operating frequencies, VLSI circuits are becoming more vulnerable to radiation strikes. As a result, this problem is now important not only for space and military electronics but also for consumer ICs. Thus, the design of radiation-hardened circuits has received significant attention in recent times. This thesis addresses the radiation hardening issue for VLSI ICs. In particular, circuit techniques are presented to protect against Single Event Transients (SETs). Radiation hardening has long been an area of research for memories for space and military ICs. In a memory, the stored state can ip as a result of a radiation strike. Such bit reversals in case of memories are known as Single Event Upsets (SEUs). With the feature sizes of VLSI ICs becoming smaller, radiation-induced glitches have become a source of concern in combinational circuits also. In combinational circuits, if a glitch due to a radiation event occurs at the time the circuit outputs are being sampled, it could lead to the propagation of a faulty value. The current or voltage glitches on the nodes of a combinational circuit are known as SETs. When an SET occurring on a node of a logic network is propagated through the gates of the network and is captured by a latch as a logic error, it is transformed to an SEU. The approach presented in this thesis makes use of Code Word State Preserving (CWSP) elements at each ip-op of the design, along with additional logic to trigger a recomputation in case a SET induced error is detected. The combinational part of the design is left unaltered. The CWSP element provides 100% SET protection for glitch widths up to min{(Dmin-D1)/2, (Dmax-D2)/2}, where Dmin and Dmax are the minimum and maximum circuit delay respectively. D1 and D2 are extra delays associated with the proposed SET protection circuit. The CWSP circuit has two inputs - the flip flop output signal and the same signal delayed by a quantity 6. In case an SET error is detected at the end of a clock period i, then the computation is repeated in clock period i+1, using the correct output value, which was captured by the CWSP element in the ith clock period. Unlike previous approaches, the CWSP element is i) in a secondary computational path and ii) the CWSP logic is designed to minimally impact the critical delay path of the design. It was found through SPICE simulations that the delay penalty of the proposed approach (averaged over several designs) is less than 1%. Thus, the proposed technique is applicable for high-speed designs, where the additional delay associated with the SET protection must be kept at a minimum.

In the microprocessor development, the high performance microprocessor applies the x86 complex instruction set is used widely. And the signal-core architecture towards slowly to multi-core one .But the variable-length instruction still creates the difficulties in instruction fetching and affects the whole executive- performance. There has the mechanism which supported the split-line and fetched fleetly the variable-lengths instruction. It has the problem in high time and hardware complexity, because it was accomplished with additional hardware. Accordingly, this paper proposed a fixed-length instruction set with design in compatible and extended x86 instruction set used the fixed-length instruction form to solve the difficulties in fetching the variable-length instructions. We considered the factor an overall arrangement of memory space and decided the length 4 bytes and 8 bytes to formulate the fixed-length instruction set. And we used the following six transitionary rules to complete the formulation for the coded form of the fixed-length instructions.(1)We used the auxiliary registers to save the value to decrease the data dependence between the original registers.(2)If it could use a few instructions to complete the translation with the original registers, we used the original registers to do it.(3)The complex case instructions were coded with eight bytes.(4)It did sign-extension by itself when displacement and immediate were moved to the auxiliary registers.(5)The auxiliary registers with the diacritic prefix were only coded in the r/m field or the index field.(6)One of displacement field and immediate field was moved first when its length was longer. And we considered the hardware complexity of saving memory space and fetching instructions, we analyzed the categories of instruction packages to compress the program space to decrease the space loss which the fixed-lengths of instructions created. In the case of verifiable and experimental framework, the CINT2006 was used to be benchmarks. And the function which translated the fixed-length instructions was succeeded to execute. It was successful to achieve the purpose the program space was compressed efficiently in the instruction package mechanism.

Schlüsselseiten aus dem Exemplar der SLUB Dresden: Diss.jur.civ.447,26.

Universiẗat, Diss.--Paderborn, 1985.