Efficient algorithms for dynamic probabilistic safety assessment. An application to convex damage domain.

The design of nuclear power plant has to follow a number of regulations aimed at limiting the risks inherent in this type of installation. The goal is to prevent and to limit the consequences of any possible incident that might threaten the public or the environment. To verify that the safety requirements are met a safety assessment process is followed. Safety analysis is as key component of a safety assessment, which incorporates both probabilistic and deterministic approaches. The deterministic approach attempts to ensure that the various situations, and in particular accidents, that are considered to be plausible, have been taken into account, and that the monitoring systems and engineered safety and safeguard systems will be capable of ensuring the safety goals. On the other hand, probabilistic safety analysis tries to demonstrate that the safety requirements are met for potential accidents both within and beyond the design basis, thus identifying vulnerabilities not necessarily accessible through deterministic safety analysis alone. Probabilistic safety assessment (PSA) methodology is widely used in the nuclear industry and is especially effective in comprehensive assessment of the measures needed to prevent accidents with small probability but severe consequences. Still, the trend towards a risk informed regulation (RIR) demanded a more extended use of risk assessment techniques with a significant need to further extend PSA’s scope and quality. Here is where the theory of stimulated dynamics (TSD) intervenes, as it is the mathematical foundation of the integrated safety assessment (ISA) methodology developed by the CSN(Consejo de Seguridad Nuclear) branch of Modelling and Simulation (MOSI). Such methodology attempts to extend classical PSA including accident dynamic analysis, an assessment of the damage associated to the transients and a computation of the damage frequency. The application of this ISA methodology requires a computational framework called SCAIS (Simulation Code System for Integrated Safety Assessment). SCAIS provides accident dynamic analysis support through simulation of nuclear accident sequences and operating procedures. Furthermore, it includes probabilistic quantification of fault trees and sequences; and integration and statistic treatment of risk metrics. SCAIS comprehensively implies an intensive use of code coupling techniques to join typical thermal hydraulic analysis, severe accident and probability calculation codes. The integration of accident simulation in the risk assessment process and thus requiring the use of complex nuclear plant models is what makes it so powerful, yet at the cost of an enormous increase in complexity. As the complexity of the process is primarily focused on such accident simulation codes, the question of whether it is possible to reduce the number of required simulation arises, which will be the focus of the present work. This document presents the work done on the investigation of more efficient techniques applied to the process of risk assessment inside the mentioned ISA methodology. Therefore such techniques will have the primary goal of decreasing the number of simulation needed for an adequate estimation of the damage probability. As the methodology and tools are relatively recent, there is not much work done inside this line of investigation, making it a quite difficult but necessary task, and because of time limitations the scope of the work had to be reduced. Therefore, some assumptions were made to work in simplified scenarios best suited for an initial approximation to the problem. The following section tries to explain in detail the process followed to design and test the developed techniques. Then, the next section introduces the general concepts and formulae of the TSD theory which are at the core of the risk assessment process. Afterwards a description of the simulation framework requirements and design is given. Followed by an introduction to the developed techniques, giving full detail of its mathematical background and its procedures. Later, the test case used is described and result from the application of the techniques is shown. Finally the conclusions are presented and future lines of work are exposed.

Multi-sensor background subtraction by fusing multiple region-based probabilistic classifiers

In the recent years, the computer vision community has shown great interest on depth-based applications thanks to the performance and flexibility of the new generation of RGB-D imagery. In this paper, we present an efficient background subtraction algorithm based on the fusion of multiple region-based classifiers that processes depth and color data provided by RGB-D cameras. Foreground objects are detected by combining a region-based foreground prediction (based on depth data) with different background models (based on a Mixture of Gaussian algorithm) providing color and depth descriptions of the scene at pixel and region level. The information given by these modules is fused in a mixture of experts fashion to improve the foreground detection accuracy. The main contributions of the paper are the region-based models of both background and foreground, built from the depth and color data. The obtained results using different database sequences demonstrate that the proposed approach leads to a higher detection accuracy with respect to existing state-of-the-art techniques.

Graphical User Interface (GUI) for the representation of GM surfaces (using the NRTL model) and curves, including tie-lines and Hessian matrix

A single and very easy to use Graphical User Interface (GUI- MATLAB) based on the topological information contained in the Gibbs energy of mixing function has been developed as a friendly tool to check the coherence of NRTL parameters obtained in a correlation data procedure. Thus, the analysis of the GM/RT surface, the GM/RT for the binaries and the GM/RT in planes containing the tie lines should be necessary to validate the obtained parameters for the different models for correlating phase equlibrium data.

A Quantile-Based Probabilistic Mean Value Theorem

For non-negative random variables with finite means we introduce an analogous of the equilibrium residual-lifetime distribution based on the quantile function. This allows us to construct new distributions with support (0, 1), and to obtain a new quantile-based version of the probabilistic generalization of Taylor's theorem. Similarly, for pairs of stochastically ordered random variables we come to a new quantile-based form of the probabilistic mean value theorem. The latter involves a distribution that generalizes the Lorenz curve. We investigate the special case of proportional quantile functions and apply the given results to various models based on classes of distributions and measures of risk theory. Motivated by some stochastic comparisons, we also introduce the “expected reversed proportional shortfall order”, and a new characterization of random lifetimes involving the reversed hazard rate function.

A probabilistic assessment of calcium carbonate export and dissolution in the modern ocean

The marine cycle of calcium carbonate (CaCO3) is an important element of the carbon cycle and co-governs the distribution of carbon and alkalinity within the ocean. However, CaCO3 export fluxes and mechanisms governing CaCO3 dissolution are highly uncertain. We present an observationally constrained, probabilistic assessment of the global and regional CaCO3 budgets. Parameters governing pelagic CaCO3 export fluxes and dissolution rates are sampled using a Monte Carlo scheme to construct a 1000-member ensemble with the Bern3D ocean model. Ensemble results are constrained by comparing simulated and observation-based fields of excess dissolved calcium carbonate (TA*). The minerals calcite and aragonite are modelled explicitly and ocean–sediment fluxes are considered. For local dissolution rates, either a strong or a weak dependency on CaCO3 saturation is assumed. In addition, there is the option to have saturation-independent dissolution above the saturation horizon. The median (and 68 % confidence interval) of the constrained model ensemble for global biogenic CaCO3 export is 0.90 (0.72–1.05) Gt C yr−1, that is within the lower half of previously published estimates (0.4–1.8 Gt C yr−1). The spatial pattern of CaCO3 export is broadly consistent with earlier assessments. Export is large in the Southern Ocean, the tropical Indo–Pacific, the northern Pacific and relatively small in the Atlantic. The constrained results are robust across a range of diapycnal mixing coefficients and, thus, ocean circulation strengths. Modelled ocean circulation and transport timescales for the different set-ups were further evaluated with CFC11 and radiocarbon observations. Parameters and mechanisms governing dissolution are hardly constrained by either the TA* data or the current compilation of CaCO3 flux measurements such that model realisations with and without saturation-dependent dissolution achieve skill. We suggest applying saturation-independent dissolution rates in Earth system models to minimise computational costs.

Dynamic Models under Uncertainty for Large-Scale Enterprise Systems

Thesis (Ph.D.)--University of Washington, 2016-06

A probabilistic switch model of information flow in social networks: Application for virtual circuits to organizational design

Network building and exchange of information by people within networks is crucial to the innovation process. Contrary to older models, in social networks the flow of information is noncontinuous and nonlinear. There are critical barriers to information flow that operate in a problematic manner. New models and new analytic tools are needed for these systems. This paper introduces the concept of virtual circuits and draws on recent concepts of network modelling and design to introduce a probabilistic switch theory that can be described using matrices. It can be used to model multistep information flow between people within organisational networks, to provide formal definitions of efficient and balanced networks and to describe distortion of information as it passes along human communication channels. The concept of multi-dimensional information space arises naturally from the use of matrices. The theory and the use of serial diagonal matrices have applications to organisational design and to the modelling of other systems. It is hypothesised that opinion leaders or creative individuals are more likely to emerge at information-rich nodes in networks. A mathematical definition of such nodes is developed and it does not invariably correspond with centrality as defined by early work on networks.

On the multicomponent polynomial solution models

The present study addresses the problem of predicting the properties of multicomponent systems from those of corresponding binary systems. Two types of multicomponent polynomial models have been analysed. A probabilistic interpretation of the parameters of the Polynomial model, which explicitly relates them with the Gibbs free energies of the generalised quasichemical reactions, is proposed. The presented treatment provides a theoretical justification for such parameters. A methodology of estimating the ternary interaction parameter from the binary ones is presented. The methodology provides a way in which the power series multicomponent models, where no projection is required, could be incorporated into the Calphad approach.

A graphical specification of model transformations with triple graph grammars

Models and model transformations are the core concepts of OMG's MDA (TM) approach. Within this approach, most models are derived from the MOF and have a graph-based nature. In contrast, most of the current model transformations are specified textually. To enable a graphical specification of model transformation rules, this paper proposes to use triple graph grammars as declarative specification formalism. These triple graph grammars can be specified within the FUJABA tool and we argue that these rules can be more easily specified and they become more understandable and maintainable. To show the practicability of our approach, we present how to generate Tefkat rules from triple graph grammar rules, which helps to integrate triple graph grammars with a state of a art model transformation tool and shows the expressiveness of the concept.

Hippocampal models for simultaneous localisation and mapping on an autonomous robot

To navigate successfully in a novel environment a robot needs to be able to Simultaneously Localize And Map (SLAM) its surroundings. The most successful solutions to this problem so far have involved probabilistic algorithms, but there has been much promising work involving systems based on the workings of part of the rodent brain known as the hippocampus. In this paper we present a biologically plausible system called RatSLAM that uses competitive attractor networks to carry out SLAM in a probabilistic manner. The system can effectively perform parameter self-calibration and SLAM in onedimension. Tests in two dimensional environments revealed the inability of the RatSLAM system to maintain multiple pose hypotheses in the face of ambiguous visual input. These results support recent rat experimentation that suggest current competitive attractor models are not a complete solution to the hippocampal modelling problem.

Mixtures of probabilistic principal component analysers

Principal component analysis (PCA) is one of the most popular techniques for processing, compressing and visualising data, although its effectiveness is limited by its global linearity. While nonlinear variants of PCA have been proposed, an alternative paradigm is to capture data complexity by a combination of local linear PCA projections. However, conventional PCA does not correspond to a probability density, and so there is no unique way to combine PCA models. Previous attempts to formulate mixture models for PCA have therefore to some extent been ad hoc. In this paper, PCA is formulated within a maximum-likelihood framework, based on a specific form of Gaussian latent variable model. This leads to a well-defined mixture model for probabilistic principal component analysers, whose parameters can be determined using an EM algorithm. We discuss the advantages of this model in the context of clustering, density modelling and local dimensionality reduction, and we demonstrate its application to image compression and handwritten digit recognition.

A generative probabilistic oriented wavelet model for texture segmentation

This Letter addresses image segmentation via a generative model approach. A Bayesian network (BNT) in the space of dyadic wavelet transform coefficients is introduced to model texture images. The model is similar to a Hidden Markov model (HMM), but with non-stationary transitive conditional probability distributions. It is composed of discrete hidden variables and observable Gaussian outputs for wavelet coefficients. In particular, the Gabor wavelet transform is considered. The introduced model is compared with the simplest joint Gaussian probabilistic model for Gabor wavelet coefficients for several textures from the Brodatz album [1]. The comparison is based on cross-validation and includes probabilistic model ensembles instead of single models. In addition, the robustness of the models to cope with additive Gaussian noise is investigated. We further study the feasibility of the introduced generative model for image segmentation in the novelty detection framework [2]. Two examples are considered: (i) sea surface pollution detection from intensity images and (ii) image segmentation of the still images with varying illumination across the scene.

Variational inference for Student-t MLP models

This paper presents a novel methodology to infer parameters of probabilistic models whose output noise is a Student-t distribution. The method is an extension of earlier work for models that are linear in parameters to nonlinear multi-layer perceptrons (MLPs). We used an EM algorithm combined with variational approximation, the evidence procedure, and an optimisation algorithm. The technique was tested on two regression applications. The first one is a synthetic dataset and the second is gas forward contract prices data from the UK energy market. The results showed that forecasting accuracy is significantly improved by using Student-t noise models.

Equilibrium properties of disordered spin models with two-scale interactions

Methods for understanding classical disordered spin systems with interactions conforming to some idealized graphical structure are well developed. The equilibrium properties of the Sherrington-Kirkpatrick model, which has a densely connected structure, have become well understood. Many features generalize to sparse Erdös- Rényi graph structures above the percolation threshold and to Bethe lattices when appropriate boundary conditions apply. In this paper, we consider spin states subject to a combination of sparse strong interactions with weak dense interactions, which we term a composite model. The equilibrium properties are examined through the replica method, with exact analysis of the high-temperature paramagnetic, spin-glass, and ferromagnetic phases by perturbative schemes. We present results of replica symmetric variational approximations, where perturbative approaches fail at lower temperature. Results demonstrate re-entrant behaviors from spin glass to ferromagnetic phases as temperature is lowered, including transitions from replica symmetry broken to replica symmetric phases. The nature of high-temperature transitions is found to be sensitive to the connectivity profile in the sparse subgraph, with regular connectivity a discontinuous transition from the paramagnetic to ferromagnetic phases is apparent.

Hidden state models for time series

Amongst all the objectives in the study of time series, uncovering the dynamic law of its generation is probably the most important. When the underlying dynamics are not available, time series modelling consists of developing a model which best explains a sequence of observations. In this thesis, we consider hidden space models for analysing and describing time series. We first provide an introduction to the principal concepts of hidden state models and draw an analogy between hidden Markov models and state space models. Central ideas such as hidden state inference or parameter estimation are reviewed in detail. A key part of multivariate time series analysis is identifying the delay between different variables. We present a novel approach for time delay estimating in a non-stationary environment. The technique makes use of hidden Markov models and we demonstrate its application for estimating a crucial parameter in the oil industry. We then focus on hybrid models that we call dynamical local models. These models combine and generalise hidden Markov models and state space models. Probabilistic inference is unfortunately computationally intractable and we show how to make use of variational techniques for approximating the posterior distribution over the hidden state variables. Experimental simulations on synthetic and real-world data demonstrate the application of dynamical local models for segmenting a time series into regimes and providing predictive distributions.