Factorization method and irregular inclusions in electrical impedance tomography

In electrical impedance tomography, one tries to recover the conductivity inside a physical body from boundary measurements of current and voltage. In many practically important situations, the investigated object has known background conductivity but it is contaminated by inhomogeneities. The factorization method of Andreas Kirsch provides a tool for locating such inclusions. Earlier, it has been shown that under suitable regularity conditions positive (or negative) inhomogeneities can be characterized by the factorization technique if the conductivity or one of its higher normal derivatives jumps on the boundaries of the inclusions. In this work, we use a monotonicity argument to generalize these results: We show that the factorization method provides a characterization of an open inclusion (modulo its boundary) if each point inside the inhomogeneity has an open neighbourhood where the perturbation of the conductivity is strictly positive (or negative) definite. In particular, we do not assume any regularity of the inclusion boundary or set any conditions on the behaviour of the perturbed conductivity at the inclusion boundary. Our theoretical findings are verified by two-dimensional numerical experiments.

Detecting interfaces in a parabolic-elliptic problem from surface measurements

Assuming that the heat capacity of a body is negligible outside certain inclusions the heat equation degenerates to a parabolic-elliptic interface problem. In this work we aim to detect these interfaces from thermal measurements on the surface of the body. We deduce an equivalent variational formulation for the parabolic-elliptic problem and give a new proof of the unique solvability based on Lions’s projection lemma. For the case that the heat conductivity is higher inside the inclusions, we develop an adaptation of the factorization method to this time-dependent problem. In particular this shows that the locations of the interfaces are uniquely determined by boundary measurements. The method also yields to a numerical algorithm to recover the inclusions and thus the interfaces. We demonstrate how measurement data can be simulated numerically by a coupling of a finite element method with a boundary element method, and finally we present some numerical results for the inverse problem.

A sampling method for detecting buried objects using electromagnetic scattering

We consider a simple (but fully three-dimensional) mathematical model for the electromagnetic exploration of buried, perfect electrically conducting objects within the soil underground. Moving an electric device parallel to the ground at constant height in order to generate a magnetic field, we measure the induced magnetic field within the device, and factor the underlying mathematics into a product of three operations which correspond to the primary excitation, some kind of reflection on the surface of the buried object(s) and the corresponding secondary excitation, respectively. Using this factorization we are able to give a justification of the so-called sampling method from inverse scattering theory for this particular set-up.

Model reduction of stochastic groundwater flow and transport processes

This work presents a comprehensive methodology for the reduction of analytical or numerical stochastic models characterized by uncertain input parameters or boundary conditions. The technique, based on the Polynomial Chaos Expansion (PCE) theory, represents a versatile solution to solve direct or inverse problems related to propagation of uncertainty. The potentiality of the methodology is assessed investigating different applicative contexts related to groundwater flow and transport scenarios, such as global sensitivity analysis, risk analysis and model calibration. This is achieved by implementing a numerical code, developed in the MATLAB environment, presented here in its main features and tested with literature examples. The procedure has been conceived under flexibility and efficiency criteria in order to ensure its adaptability to different fields of engineering; it has been applied to different case studies related to flow and transport in porous media. Each application is associated with innovative elements such as (i) new analytical formulations describing motion and displacement of non-Newtonian fluids in porous media, (ii) application of global sensitivity analysis to a high-complexity numerical model inspired by a real case of risk of radionuclide migration in the subsurface environment, and (iii) development of a novel sensitivity-based strategy for parameter calibration and experiment design in laboratory scale tracer transport.

Numerical analysis of the relation between interactions and structure in a molecular fluid

Coarse graining is a popular technique used in physics to speed up the computer simulation of molecular fluids. An essential part of this technique is a method that solves the inverse problem of determining the interaction potential or its parameters from the given structural data. Due to discrepancies between model and reality, the potential is not unique, such that stability of such method and its convergence to a meaningful solution are issues.rnrnIn this work, we investigate empirically whether coarse graining can be improved by applying the theory of inverse problems from applied mathematics. In particular, we use the singular value analysis to reveal the weak interaction parameters, that have a negligible influence on the structure of the fluid and which cause non-uniqueness of the solution. Further, we apply a regularizing Levenberg-Marquardt method, which is stable against the mentioned discrepancies. Then, we compare it to the existing physical methods - the Iterative Boltzmann Inversion and the Inverse Monte Carlo method, which are fast and well adapted to the problem, but sometimes have convergence problems.rnrnFrom analysis of the Iterative Boltzmann Inversion, we elaborate a meaningful approximation of the structure and use it to derive a modification of the Levenberg-Marquardt method. We engage the latter for reconstruction of the interaction parameters from experimental data for liquid argon and nitrogen. We show that the modified method is stable, convergent and fast. Further, the singular value analysis of the structure and its approximation allows to determine the crucial interaction parameters, that is, to simplify the modeling of interactions. Therefore, our results build a rigorous bridge between the inverse problem from physics and the powerful solution tools from mathematics. rn

Can You Hear the Shape of a Drum?

If you had perfect pitch and listened to a recording of the sounds a drum made when struck, could you determine the shape of the drum? This question is an example of an inverse problem; inverse problems arise in medical imaging, oil prospecting, spectroscopy, and many other fields. We’ll first discuss the analogous question in the simpler setting of plucking a string. Then we’ll tackle the problem for drums and see that there are some surprises. Finally, I will give a brief indication of how this problem relates to some of my recent research. The emphasis will be on the ideas rather than on the technical details, so there will be pretty pictures instead of equations.

Confidence Bands for Convex Median Curves using Sign-Tests

Suppose that one observes pairs (x1,Y1), (x2,Y2), ..., (xn,Yn), where x1 < x2 < ... < xn are fixed numbers while Y1, Y2, ..., Yn are independent random variables with unknown distributions. The only assumption is that Median(Yi) = f(xi) for some unknown convex or concave function f. We present a confidence band for this regression function f using suitable multiscale sign tests. While the exact computation of this band seems to require O(n4) steps, good approximations can be obtained in O(n2) steps. In addition the confidence band is shown to have desirable asymptotic properties as the sample size n tends to infinity.

Marshall's Lemma for Convex Density Estimation

Marshall's (1970) lemma is an analytical result which implies root-n-consistency of the distribution function corresponding to the Grenander (1956) estimator of a non-decreasing probability density. The present paper derives analogous results for the setting of convex densities on [0,\infty).

KrigInv: An efficient and user-friendly implementation of batch-sequential inversion strategies based on kriging

Several strategies relying on kriging have recently been proposed for adaptively estimating contour lines and excursion sets of functions under severely limited evaluation budget. The recently released R package KrigInv 3 is presented and offers a sound implementation of various sampling criteria for those kinds of inverse problems. KrigInv is based on the DiceKriging package, and thus benefits from a number of options concerning the underlying kriging models. Six implemented sampling criteria are detailed in a tutorial and illustrated with graphical examples. Different functionalities of KrigInv are gradually explained. Additionally, two recently proposed criteria for batch-sequential inversion are presented, enabling advanced users to distribute function evaluations in parallel on clusters or clouds of machines. Finally, auxiliary problems are discussed. These include the fine tuning of numerical integration and optimization procedures used within the computation and the optimization of the considered criteria.

Hybrid topological derivative and gradient-based methods for electrical impedance tomography

We present a technique to reconstruct the electromagnetic properties of a medium or a set of objects buried inside it from boundary measurements when applying electric currents through a set of electrodes. The electromagnetic parameters may be recovered by means of a gradient method without a priori information on the background. The shape, location and size of objects, when present, are determined by a topological derivative-based iterative procedure. The combination of both strategies allows improved reconstructions of the objects and their properties, assuming a known background.

Analysis of the static and dynamic behaviour of hydraulic fills

En la actualidad existe un gran conocimiento en la caracterización de rellenos hidráulicos, tanto en su caracterización estática, como dinámica. Sin embargo, son escasos en la literatura estudios más generales y globales de estos materiales, muy relacionados con sus usos y principales problemáticas en obras portuarias y mineras. Los procedimientos semi‐empíricos para la evaluación del efecto silo en las celdas de cajones portuarios, así como para el potencial de licuefacción de estos suelos durantes cargas instantáneas y terremotos, se basan en estudios donde la influencia de los parámetros que los rigen no se conocen en gran medida, dando lugar a resultados con considerable dispersión. Este es el caso, por ejemplo, de los daños notificados por el grupo de investigación del Puerto de Barcelona, la rotura de los cajones portuarios en el Puerto de Barcelona en 2007. Por estos motivos y otros, se ha decidido desarrollar un análisis para la evaluación de estos problemas mediante la propuesta de una metodología teórico‐numérica y empírica. El enfoque teórico‐numérico desarrollado en el presente estudio se centra en la determinación del marco teórico y las herramientas numéricas capaces de solventar los retos que presentan estos problemas. La complejidad del problema procede de varios aspectos fundamentales: el comportamiento no lineal de los suelos poco confinados o flojos en procesos de consolidación por preso propio; su alto potencial de licuefacción; la caracterización hidromecánica de los contactos entre estructuras y suelo (camino preferencial para el flujo de agua y consolidación lateral); el punto de partida de los problemas con un estado de tensiones efectivas prácticamente nulo. En cuanto al enfoque experimental, se ha propuesto una metodología de laboratorio muy sencilla para la caracterización hidromecánica del suelo y las interfaces, sin la necesidad de usar complejos aparatos de laboratorio o procedimientos excesivamente complicados. Este trabajo incluye por tanto un breve repaso a los aspectos relacionados con la ejecución de los rellenos hidráulicos, sus usos principales y los fenómenos relacionados, con el fin de establecer un punto de partida para el presente estudio. Este repaso abarca desde la evolución de las ecuaciones de consolidación tradicionales (Terzaghi, 1943), (Gibson, English & Hussey, 1967) y las metodologías de cálculo (Townsend & McVay, 1990) (Fredlund, Donaldson and Gitirana, 2009) hasta las contribuciones en relación al efecto silo (Ranssen, 1985) (Ravenet, 1977) y sobre el fenómeno de la licuefacción (Casagrande, 1936) (Castro, 1969) (Been & Jefferies, 1985) (Pastor & Zienkiewicz, 1986). Con motivo de este estudio se ha desarrollado exclusivamente un código basado en el método de los elementos finitos (MEF) empleando el programa MATLAB. Para ello, se ha esablecido un marco teórico (Biot, 1941) (Zienkiewicz & Shiomi, 1984) (Segura & Caron, 2004) y numérico (Zienkiewicz & Taylor, 1989) (Huerta & Rodríguez, 1992) (Segura & Carol, 2008) para resolver problemas de consolidación multidimensional con condiciones de contorno friccionales, y los correspondientes modelos constitutivos (Pastor & Zienkiewicz, 1986) (Fiu & Liu, 2011). Asimismo, se ha desarrollado una metodología experimental a través de una serie de ensayos de laboratorio para la calibración de los modelos constitutivos y de la caracterización de parámetros índice y de flujo (Castro, 1969) (Bahda 1997) (Been & Jefferies, 2006). Para ello se han empleado arenas de Hostun como material (relleno hidráulico) de referencia. Como principal aportación se incluyen una serie de nuevos ensayos de corte directo para la caracterización hidromecánica de la interfaz suelo – estructura de hormigón, para diferentes tipos de encofrados y rugosidades. Finalmente, se han diseñado una serie de algoritmos específicos para la resolución del set de ecuaciones diferenciales de gobierno que definen este problema. Estos algoritmos son de gran importancia en este problema para tratar el procesamiento transitorio de la consolidación de los rellenos hidráulicos, y de otros efectos relacionados con su implementación en celdas de cajones, como el efecto silo y la licuefacciones autoinducida. Para ello, se ha establecido un modelo 2D axisimétrico, con formulación acoplada u‐p para elementos continuos y elementos interfaz (de espesor cero), que tratan de simular las condiciones de estos rellenos hidráulicos cuando se colocan en las celdas portuarias. Este caso de estudio hace referencia clara a materiales granulares en estado inicial muy suelto y con escasas tensiones efectivas, es decir, con prácticamente todas las sobrepresiones ocasionadas por el proceso de autoconsolidación (por peso propio). Por todo ello se requiere de algoritmos numéricos específicos, así como de modelos constitutivos particulares, para los elementos del continuo y para los elementos interfaz. En el caso de la simulación de diferentes procedimientos de puesta en obra de los rellenos se ha requerido la modificacion de los algoritmos empleados para poder así representar numéricamente la puesta en obra de estos materiales, además de poder realizar una comparativa de los resultados para los distintos procedimientos. La constante actualización de los parámetros del suelo, hace también de este algoritmo una potente herramienta que permite establecer un interesante juego de perfiles de variables, tales como la densidad, el índice de huecos, la fracción de sólidos, el exceso de presiones, y tensiones y deformaciones. En definitiva, el modelo otorga un mejor entendimiento del efecto silo, término comúnmente usado para definir el fenómeno transitorio del gradiente de presiones laterales en las estructuras de contención en forma de silo. Finalmente se incluyen una serie de comparativas entre los resultados del modelo y de diferentes estudios de la literatura técnica, tanto para el fenómeno de las consolidaciones por preso propio (Fredlund, Donaldson & Gitirana, 2009) como para el estudio del efecto silo (Puertos del Estado, 2006, EuroCódigo (2006), Japan Tech, Stands. (2009), etc.). Para concluir, se propone el diseño de un prototipo de columna de decantación con paredes friccionales, como principal propuesta de futura línea de investigación. Wide research is nowadays available on the characterization of hydraulic fills in terms of either static or dynamic behavior. However, reported comprehensive analyses of these soils when meant for port or mining works are scarce. Moreover, the semi‐empirical procedures for assessing the silo effect on cells in floating caissons, and the liquefaction potential of these soils during sudden loads or earthquakes are based on studies where the underlying influence parameters are not well known, yielding results with significant scatter. This is the case, for instance, of hazards reported by the Barcelona Liquefaction working group, with the failure of harbor walls in 2007. By virtue of this, a complex approach has been undertaken to evaluate the problem by a proposal of numerical and laboratory methodology. Within a theoretical and numerical scope, the study is focused on the numerical tools capable to face the different challenges of this problem. The complexity is manifold; the highly non‐linear behavior of consolidating soft soils; their potentially liquefactable nature, the significance of the hydromechanics of the soil‐structure contact, the discontinuities as preferential paths for water flow, setting “negligible” effective stresses as initial conditions. Within an experimental scope, a straightforward laboratory methodology is introduced for the hydromechanical characterization of the soil and the interface without the need of complex laboratory devices or cumbersome procedures. Therefore, this study includes a brief overview of the hydraulic filling execution, main uses (land reclamation, filled cells, tailing dams, etc.) and the underlying phenomena (self‐weight consolidation, silo effect, liquefaction, etc.). It comprises from the evolution of the traditional consolidation equations (Terzaghi, 1943), (Gibson, English, & Hussey, 1967) and solving methodologies (Townsend & McVay, 1990) (Fredlund, Donaldson and Gitirana, 2009) to the contributions in terms of silo effect (Ranssen, 1895) (Ravenet, 1977) and liquefaction phenomena (Casagrande, 1936) (Castro, 1969) (Been & Jefferies, 1985) (Pastor & Zienkiewicz, 1986). The novelty of the study lies on the development of a Finite Element Method (FEM) code, exclusively formulated for this problem. Subsequently, a theoretical (Biot, 1941) (Zienkiewicz and Shiomi, 1984) (Segura and Carol, 2004) and numerical approach (Zienkiewicz and Taylor, 1989) (Huerta, A. & Rodriguez, A., 1992) (Segura, J.M. & Carol, I., 2008) is introduced for multidimensional consolidation problems with frictional contacts and the corresponding constitutive models (Pastor & Zienkiewicz, 1986) (Fu & Liu, 2011). An experimental methodology is presented for the laboratory test and material characterization (Castro 1969) (Bahda 1997) (Been & Jefferies 2006) using Hostun sands as reference hydraulic fill. A series of singular interaction shear tests for the interface calibration is included. Finally, a specific model algorithm for the solution of the set of differential equations governing the problem is presented. The process of consolidation and settlements involves a comprehensive simulation of the transient process of decantation and the build‐up of the silo effect in cells and certain phenomena related to self‐compaction and liquefaction. For this, an implementation of a 2D axi‐syimmetric coupled model with continuum and interface elements, aimed at simulating conditions and self‐weight consolidation of hydraulic fills once placed into floating caisson cells or close to retaining structures. This basically concerns a loose granular soil with a negligible initial effective stress level at the onset of the process. The implementation requires a specific numerical algorithm as well as specific constitutive models for both the continuum and the interface elements. The simulation of implementation procedures for the fills has required the modification of the algorithm so that a numerical representation of these procedures is carried out. A comparison of the results for the different procedures is interesting for the global analysis. Furthermore, the continuous updating of the model provides an insightful logging of variable profiles such as density, void ratio and solid fraction profiles, total and excess pore pressure, stresses and strains. This will lead to a better understanding of complex phenomena such as the transient gradient in lateral pressures due to silo effect in saturated soils. Interesting model and literature comparisons for the self‐weight consolidation (Fredlund, Donaldson, & Gitirana, 2009) and the silo effect results (Puertos del Estado (2006), EuroCode (2006), Japan Tech, Stands. (2009)). This study closes with the design of a decantation column prototype with frictional walls as the main future line of research.

An advanced regularization methodology for use in watershed model calibration

A calibration methodology based on an efficient and stable mathematical regularization scheme is described. This scheme is a variant of so-called Tikhonov regularization in which the parameter estimation process is formulated as a constrained minimization problem. Use of the methodology eliminates the need for a modeler to formulate a parsimonious inverse problem in which a handful of parameters are designated for estimation prior to initiating the calibration process. Instead, the level of parameter parsimony required to achieve a stable solution to the inverse problem is determined by the inversion algorithm itself. Where parameters, or combinations of parameters, cannot be uniquely estimated, they are provided with values, or assigned relationships with other parameters, that are decreed to be realistic by the modeler. Conversely, where the information content of a calibration dataset is sufficient to allow estimates to be made of the values of many parameters, the making of such estimates is not precluded by preemptive parsimonizing ahead of the calibration process. White Tikhonov schemes are very attractive and hence widely used, problems with numerical stability can sometimes arise because the strength with which regularization constraints are applied throughout the regularized inversion process cannot be guaranteed to exactly complement inadequacies in the information content of a given calibration dataset. A new technique overcomes this problem by allowing relative regularization weights to be estimated as parameters through the calibration process itself. The technique is applied to the simultaneous calibration of five subwatershed models, and it is demonstrated that the new scheme results in a more efficient inversion, and better enforcement of regularization constraints than traditional Tikhonov regularization methodologies. Moreover, it is argued that a joint calibration exercise of this type results in a more meaningful set of parameters than can be achieved by individual subwatershed model calibration. (c) 2005 Elsevier B.V. All rights reserved.

The cost of uniqueness in groundwater model calibration

Calibration of a groundwater model requires that hydraulic properties be estimated throughout a model domain. This generally constitutes an underdetermined inverse problem, for which a Solution can only be found when some kind of regularization device is included in the inversion process. Inclusion of regularization in the calibration process can be implicit, for example through the use of zones of constant parameter value, or explicit, for example through solution of a constrained minimization problem in which parameters are made to respect preferred values, or preferred relationships, to the degree necessary for a unique solution to be obtained. The cost of uniqueness is this: no matter which regularization methodology is employed, the inevitable consequence of its use is a loss of detail in the calibrated field. This, ill turn, can lead to erroneous predictions made by a model that is ostensibly well calibrated. Information made available as a by-product of the regularized inversion process allows the reasons for this loss of detail to be better understood. In particular, it is easily demonstrated that the estimated value for an hydraulic property at any point within a model domain is, in fact, a weighted average of the true hydraulic property over a much larger area. This averaging process causes loss of resolution in the estimated field. Where hydraulic conductivity is the hydraulic property being estimated, high averaging weights exist in areas that are strategically disposed with respect to measurement wells, while other areas may contribute very little to the estimated hydraulic conductivity at any point within the model domain, this possibly making the detection of hydraulic conductivity anomalies in these latter areas almost impossible. A study of the post-calibration parameter field covariance matrix allows further insights into the loss of system detail incurred through the calibration process to be gained. A comparison of pre- and post-calibration parameter covariance matrices shows that the latter often possess a much smaller spectral bandwidth than the former. It is also demonstrated that, as all inevitable consequence of the fact that a calibrated model cannot replicate every detail of the true system, model-to-measurement residuals can show a high degree of spatial correlation, a fact which must be taken into account when assessing these residuals either qualitatively, or quantitatively in the exploration of model predictive uncertainty. These principles are demonstrated using a synthetic case in which spatial parameter definition is based oil pilot points, and calibration is Implemented using both zones of piecewise constancy and constrained minimization regularization. (C) 2005 Elsevier Ltd. All rights reserved.

Mixture density networks

Minimization of a sum-of-squares or cross-entropy error function leads to network outputs which approximate the conditional averages of the target data, conditioned on the input vector. For classifications problems, with a suitably chosen target coding scheme, these averages represent the posterior probabilities of class membership, and so can be regarded as optimal. For problems involving the prediction of continuous variables, however, the conditional averages provide only a very limited description of the properties of the target variables. This is particularly true for problems in which the mapping to be learned is multi-valued, as often arises in the solution of inverse problems, since the average of several correct target values is not necessarily itself a correct value. In order to obtain a complete description of the data, for the purposes of predicting the outputs corresponding to new input vectors, we must model the conditional probability distribution of the target data, again conditioned on the input vector. In this paper we introduce a new class of network models obtained by combining a conventional neural network with a mixture density model. The complete system is called a Mixture Density Network, and can in principle represent arbitrary conditional probability distributions in the same way that a conventional neural network can represent arbitrary functions. We demonstrate the effectiveness of Mixture Density Networks using both a toy problem and a problem involving robot inverse kinematics.

Theoretical foundations of neural networks

Neural networks have often been motivated by superficial analogy with biological nervous systems. Recently, however, it has become widely recognised that the effective application of neural networks requires instead a deeper understanding of the theoretical foundations of these models. Insight into neural networks comes from a number of fields including statistical pattern recognition, computational learning theory, statistics, information geometry and statistical mechanics. As an illustration of the importance of understanding the theoretical basis for neural network models, we consider their application to the solution of multi-valued inverse problems. We show how a naive application of the standard least-squares approach can lead to very poor results, and how an appreciation of the underlying statistical goals of the modelling process allows the development of a more general and more powerful formalism which can tackle the problem of multi-modality.