Robust model-order reduction of complex biological processes

This paper addresses robust model-order reduction of a high dimensional nonlinear partial differential equation (PDE) model of a complex biological process. Based on a nonlinear, distributed parameter model of the same process which was validated against experimental data of an existing, pilot-scale BNR activated sludge plant, we developed a state-space model with 154 state variables in this work. A general algorithm for robustly reducing the nonlinear PDE model is presented and based on an investigation of five state-of-the-art model-order reduction techniques, we are able to reduce the original model to a model with only 30 states without incurring pronounced modelling errors. The Singular perturbation approximation balanced truncating technique is found to give the lowest modelling errors in low frequency ranges and hence is deemed most suitable for controller design and other real-time applications. (C) 2002 Elsevier Science Ltd. All rights reserved.

Elastically restrained Bernoulli-Euler beams applied to rotary machinery modelling

Facing the lateral vibration problem of a machine rotor as a beam on elastic supports in bending, the authors deal with the free vibration of elastically restrained Bernoulli-Euler beams carrying a finite number of concentrated elements along their length. Based on Rayleigh's quotient, an iterative strategy is developed to find the approximated torsional stiffness coefficients, which allows the reconciliation between the theoretical model results and the experimental ones, obtained through impact tests. The mentioned algorithm treats the vibration of continuous beams under a determined set of boundary and continuity conditions, including different torsional stiffness coefficients and the effect of attached concentrated masses and rotational inertias, not only in the energetic terms of the Rayleigh's quotient but also on the mode shapes, considering the shape functions defined in branches. Several loading cases are examined and examples are given to illustrate the validity of the model and accuracy of the obtained natural frequencies.

Pseudo-fractional ARMA modelling using a double Levinson recursion

IET Control Theory & Applications, Vol. 1, Nº 1

Monte Carlo simulation of x-ray spectra generated by kilo-electron-volt electrons

We present a general algorithm for the simulation of x-ray spectra emitted from targets of arbitrary composition bombarded with kilovolt electron beams. Electron and photon transport is simulated by means of the general-purpose Monte Carlo code PENELOPE, using the standard, detailed simulation scheme. Bremsstrahlung emission is described by using a recently proposed algorithm, in which the energy of emitted photons is sampled from numerical cross-section tables, while the angular distribution of the photons is represented by an analytical expression with parameters determined by fitting benchmark shape functions obtained from partial-wave calculations. Ionization of K and L shells by electron impact is accounted for by means of ionization cross sections calculated from the distorted-wave Born approximation. The relaxation of the excited atoms following the ionization of an inner shell, which proceeds through emission of characteristic x rays and Auger electrons, is simulated until all vacancies have migrated to M and outer shells. For comparison, measurements of x-ray emission spectra generated by 20 keV electrons impinging normally on multiple bulk targets of pure elements, which span the periodic system, have been performed using an electron microprobe. Simulation results are shown to be in close agreement with these measurements.

Packing triangles in low degree graphs and indifference graphs

We consider the problems of finding the maximum number of vertex-disjoint triangles (VTP) and edge-disjoint triangles (ETP) in a simple graph. Both problems are NP-hard. The algorithm with the best approximation ratio known so far for these problems has ratio 3/2 + epsilon, a result that follows from a more general algorithm for set packing obtained by Hurkens and Schrijver [On the size of systems of sets every t of which have an SDR, with an application to the worst-case ratio of heuristics for packing problems, SIAM J. Discrete Math. 2(1) (1989) 68-72]. We present improvements on the approximation ratio for restricted cases of VTP and ETP that are known to be APX-hard: we give an approximation algorithm for VTP on graphs with maximum degree 4 with ratio slightly less than 1.2, and for ETP on graphs with maximum degree 5 with ratio 4/3. We also present an exact linear-time algorithm for VTP on the class of indifference graphs. (C) 2007 Elsevier B.V. All rights reserved.

Método de integração em dimensão negativa em teoria quântica de campos

This work is a review of the Negative Dimension Integration Method as a powerful tool for the computation of the radiative corrections present in Quantum Field Perturbation Theory. This method is applicable in the context of Dimensional Regularization and it provides exact solutions for Feynman integrals with both dimensional parameter and propagator exponents generalized. These solutions are presentedintheformoflinearcombinationsofhypergeometricfunctionswhosedomains of convergence are related to the analytic structure of the Feynman Integral. Each solution is connected to the others trough analytic continuations. Besides presenting and discussing the general algorithm of the method in a detailed way, we offer concrete applications to scalar one-loop and two-loop integrals as well as to the one-loop renormalizationofQuantumElectrodynamics (QED)

HIGH PERFORMANCE, LOW COST SUBSPACE DECOMPOSITION AND POLYNOMIAL ROOTING FOR REAL TIME DIRECTION OF ARRIVAL ESTIMATION: ANALYSIS AND IMPLEMENTATION

This thesis develops high performance real-time signal processing modules for direction of arrival (DOA) estimation for localization systems. It proposes highly parallel algorithms for performing subspace decomposition and polynomial rooting, which are otherwise traditionally implemented using sequential algorithms. The proposed algorithms address the emerging need for real-time localization for a wide range of applications. As the antenna array size increases, the complexity of signal processing algorithms increases, making it increasingly difficult to satisfy the real-time constraints. This thesis addresses real-time implementation by proposing parallel algorithms, that maintain considerable improvement over traditional algorithms, especially for systems with larger number of antenna array elements. Singular value decomposition (SVD) and polynomial rooting are two computationally complex steps and act as the bottleneck to achieving real-time performance. The proposed algorithms are suitable for implementation on field programmable gated arrays (FPGAs), single instruction multiple data (SIMD) hardware or application specific integrated chips (ASICs), which offer large number of processing elements that can be exploited for parallel processing. The designs proposed in this thesis are modular, easily expandable and easy to implement. Firstly, this thesis proposes a fast converging SVD algorithm. The proposed method reduces the number of iterations it takes to converge to correct singular values, thus achieving closer to real-time performance. A general algorithm and a modular system design are provided making it easy for designers to replicate and extend the design to larger matrix sizes. Moreover, the method is highly parallel, which can be exploited in various hardware platforms mentioned earlier. A fixed point implementation of proposed SVD algorithm is presented. The FPGA design is pipelined to the maximum extent to increase the maximum achievable frequency of operation. The system was developed with the objective of achieving high throughput. Various modern cores available in FPGAs were used to maximize the performance and details of these modules are presented in detail. Finally, a parallel polynomial rooting technique based on Newton’s method applicable exclusively to root-MUSIC polynomials is proposed. Unique characteristics of root-MUSIC polynomial’s complex dynamics were exploited to derive this polynomial rooting method. The technique exhibits parallelism and converges to the desired root within fixed number of iterations, making this suitable for polynomial rooting of large degree polynomials. We believe this is the first time that complex dynamics of root-MUSIC polynomial were analyzed to propose an algorithm. In all, the thesis addresses two major bottlenecks in a direction of arrival estimation system, by providing simple, high throughput, parallel algorithms.

Development and improvement of protection-system algorithms for ground-fault detection and location in synchronous machines

Uno de los defectos más frecuentes en los generadores síncronos son los defectos a tierra tanto en el devanado estatórico, como de excitación. Se produce un defecto cuando el aislamiento eléctrico entre las partes activas de cualquiera de estos devanados y tierra se reduce considerablemente o desaparece. La detección de los defectos a tierra en ambos devanados es un tema ampliamente estudiado a nivel industrial. Tras la detección y confirmación de la existencia del defecto, dicha falta debe ser localizada a lo largo del devanado para su reparación, para lo que habitualmente el rotor debe ser extraído del estator. Esta operación resulta especialmente compleja y cara. Además, el hecho de limitar la corriente de defecto en ambos devanados provoca que el defecto no sea localizable visualmente, pues apenas existe daño en el generador. Por ello, se deben aplicar técnicas muy laboriosas para localizar exactamente el defecto y poder así reparar el devanado. De cara a reducir el tiempo de reparación, y con ello el tiempo en que el generador esta fuera de servicio, cualquier información por parte del relé de protección acerca de la localización del defecto resultaría de gran utilidad. El principal objetivo de esta tesis doctoral ha sido el desarrollo de nuevos algoritmos que permitan la estimación de la localización de los defectos a tierra tanto en el devanado rotórico como estatórico de máquinas síncronas. Respecto al devanado de excitación, se ha presentado un nuevo método de localización de defectos a tierra para generadores con excitación estática. Este método permite incluso distinguir si el defecto se ha producido en el devanado de excitación, o en cualquiera de los componentes del sistema de excitación, esto es, transformador de excitación, conductores de alimentación del rectificador controlado, etc. En caso de defecto a tierra en del devanado rotórico, este método proporciona una estimación de su localización. Sin embargo, para poder obtener la localización del defecto, se precisa conocer el valor de resistencia de defecto. Por ello, en este trabajo se presenta además un nuevo método para la estimación de este parámetro de forma precisa. Finalmente, se presenta un nuevo método de detección de defectos a tierra, basado en el criterio direccional, que complementa el método de localización, permitiendo tener en cuenta la influencia de las capacidades a tierra del sistema. Estas capacidades resultan determinantes a la hora de localizar el defecto de forma adecuada. En relación con el devanado estatórico, en esta tesis doctoral se presenta un nuevo algoritmo de localización de defectos a tierra para generadores que dispongan de la protección de faltas a tierra basada en la inyección de baja frecuencia. Se ha propuesto un método general, que tiene en cuenta todos los parámetros del sistema, así como una versión simplificada del método para generadores con capacidades a tierra muy reducida, que podría resultar de fácil implementación en relés de protección comercial. Los algoritmos y métodos presentados se han validado mediante ensayos experimentales en un generador de laboratorio de 5 kVA, así como en un generador comercial de 106 MVA con resultados satisfactorios y prometedores. ABSTRACT One of the most common faults in synchronous generators is the ground fault in both the stator winding and the excitation winding. In case of fault, the insulation level between the active part of any of these windings and ground lowers considerably, or even disappears. The detection of ground faults in both windings is a very researched topic. The fault current is typically limited intentionally to a reduced level. This allows to detect easily the ground faults, and therefore to avoid damage in the generator. After the detection and confirmation of the existence of a ground fault, it should be located along the winding in order to repair of the machine. Then, the rotor has to be extracted, which is a very complex and expensive operation. Moreover, the fact of limiting the fault current makes that the insulation failure is not visually detectable, because there is no visible damage in the generator. Therefore, some laborious techniques have to apply to locate accurately the fault. In order to reduce the repair time, and therefore the time that the generator is out of service, any information about the approximate location of the fault would be very useful. The main objective of this doctoral thesis has been the development of new algorithms and methods to estimate the location of ground faults in the stator and in the rotor winding of synchronous generators. Regarding the excitation winding, a new location method of ground faults in excitation winding of synchronous machines with static excitation has been presented. This method allows even to detect if the fault is at the excitation winding, or in any other component of the excitation system: controlled rectifier, excitation transformer, etc. In case of ground fault in the rotor winding, this method provides an estimation of the fault location. However, in order to calculate the location, the value of fault resistance is necessary. Therefore, a new fault-resistance estimation algorithm is presented in this text. Finally, a new fault detection algorithm based on directional criterion is described to complement the fault location method. This algorithm takes into account the influence of the capacitance-to-ground of the system, which has a remarkable impact in the accuracy of the fault location. Regarding the stator winding, a new fault-location algorithm has been presented for stator winding of synchronous generators. This algorithm is applicable to generators with ground-fault protection based in low-frequency injection. A general algorithm, which takes every parameter of the system into account, has been presented. Moreover, a simplified version of the algorithm has been proposed for generators with especially low value of capacitance to ground. This simplified algorithm might be easily implementable in protective relays. The proposed methods and algorithms have been tested in a 5 kVA laboratory generator, as well as in a 106 MVA synchronous generator with satisfactory and promising results.

The exact sample complexity of PAC-learning problems with unit VC dimension

The Vapnik-Chervonenkis (VC) dimension is a combinatorial measure of a certain class of machine learning problems, which may be used to obtain upper and lower bounds on the number of training examples needed to learn to prescribed levels of accuracy. Most of the known bounds apply to the Probably Approximately Correct (PAC) framework, which is the framework within which we work in this paper. For a learning problem with some known VC dimension, much is known about the order of growth of the sample-size requirement of the problem, as a function of the PAC parameters. The exact value of sample-size requirement is however less well-known, and depends heavily on the particular learning algorithm being used. This is a major obstacle to the practical application of the VC dimension. Hence it is important to know exactly how the sample-size requirement depends on VC dimension, and with that in mind, we describe a general algorithm for learning problems having VC dimension 1. Its sample-size requirement is minimal (as a function of the PAC parameters), and turns out to be the same for all non-trivial learning problems having VC dimension 1. While the method used cannot be naively generalised to higher VC dimension, it suggests that optimal algorithm-dependent bounds may improve substantially on current upper bounds.

One Class of Stochastic Local Search Algorithms

Accelerated probabilistic modeling algorithms, presenting stochastic local search (SLS) technique, are considered. General algorithm scheme and specific combinatorial optimization method, using “golden section” rule (GS-method), are given. Convergence rates using Markov chains are received. An overview of current combinatorial optimization techniques is presented.

The optical properties of metamaterials

This thesis studies the parametric investigation, polarisation dependence and characterization of fishnet structure at near infrared wavelengths. Detailed simulations are performed to understand the behaviour of the structure at near infrared and optical wavelengths. Simulations are performed to obtain negative refractive index of the fishnet structure formed from nanoimprint lithography (NIL) by taking into account the effect of substrate and polymethyl methacrylate (PMMA) beneath it. Two different structures have been designed and fabricated of varying dimensions using NIL and their resonant wavelength measured in the near infrared at 1.45 µm and 1.88 µm. Simulations suggest that a negative refractive index real part with the magnitude -0.24 is found at 1.53 µm and this decrease to a maximum magnitude of -0.57 at 1.9 µm. The PMMA and suppressed pillars are here responsible for the increasing material losses and limiting the value of negative refractive index. An analytical approach has been suggested to characterise fishnet structures at oblique incidence. The expressions for an absorbing medium are rewritten for an alternative definition of refractive index. The expressions are initially validated for a dielectric slab and a metal film. These results provide the possibility that this proposal may yield a general algorithm for obtaining the complex reflection and transmission coefficients for artificial structures. FDTD simulations have been extensively used in this thesis to understand the optical metamaterials and their characterization.

Methods for integrating machine learning and constrained optimization

In the framework of industrial problems, the application of Constrained Optimization is known to have overall very good modeling capability and performance and stands as one of the most powerful, explored, and exploited tool to address prescriptive tasks. The number of applications is huge, ranging from logistics to transportation, packing, production, telecommunication, scheduling, and much more. The main reason behind this success is to be found in the remarkable effort put in the last decades by the OR community to develop realistic models and devise exact or approximate methods to solve the largest variety of constrained or combinatorial optimization problems, together with the spread of computational power and easily accessible OR software and resources. On the other hand, the technological advancements lead to a data wealth never seen before and increasingly push towards methods able to extract useful knowledge from them; among the data-driven methods, Machine Learning techniques appear to be one of the most promising, thanks to its successes in domains like Image Recognition, Natural Language Processes and playing games, but also the amount of research involved. The purpose of the present research is to study how Machine Learning and Constrained Optimization can be used together to achieve systems able to leverage the strengths of both methods: this would open the way to exploiting decades of research on resolution techniques for COPs and constructing models able to adapt and learn from available data. In the first part of this work, we survey the existing techniques and classify them according to the type, method, or scope of the integration; subsequently, we introduce a novel and general algorithm devised to inject knowledge into learning models through constraints, Moving Target. In the last part of the thesis, two applications stemming from real-world projects and done in collaboration with Optit will be presented.

An exact conserving algorithm for nonlinear dynamics with rotational DOFs and general hyperelasticity. Part 2: shells

Following the approach developed for rods in Part 1 of this paper (Pimenta et al. in Comput. Mech. 42:715-732, 2008), this work presents a fully conserving algorithm for the integration of the equations of motion in nonlinear shell dynamics. We begin with a re-parameterization of the rotation field in terms of the so-called Rodrigues rotation vector, allowing for an extremely simple update of the rotational variables within the scheme. The weak form is constructed via non-orthogonal projection, the time-collocation of which ensures exact conservation of momentum and total energy in the absence of external forces. Appealing is the fact that general hyperelastic materials (and not only materials with quadratic potentials) are permitted in a totally consistent way. Spatial discretization is performed using the finite element method and the robust performance of the scheme is demonstrated by means of numerical examples.

An exact conserving algorithm for nonlinear dynamics with rotational DOFs and general hyperelasticity. Part 1: Rods

A fully conserving algorithm is developed in this paper for the integration of the equations of motion in nonlinear rod dynamics. The starting point is a re-parameterization of the rotation field in terms of the so-called Rodrigues rotation vector, which results in an extremely simple update of the rotational variables. The weak form is constructed with a non-orthogonal projection corresponding to the application of the virtual power theorem. Together with an appropriate time-collocation, it ensures exact conservation of momentum and total energy in the absence of external forces. Appealing is the fact that nonlinear hyperelastic materials (and not only materials with quadratic potentials) are permitted without any prejudice on the conservation properties. Spatial discretization is performed via the finite element method and the performance of the scheme is assessed by means of several numerical simulations.