MaaS4All Project Report

Report (2021)

Machine learning techniques for atmospheric pollutant monitoring

Poster (2012, January 27)

MAgNET: A Graph U-Net Architecture for Mesh-Based Simulations

E-print/Working paper (2023)

In many cutting-edge applications, high-fidelity computational models prove too slow to be practical and are thus replaced by much faster surrogate models. Recently, deep learning techniques have become ... [more ▼]

In many cutting-edge applications, high-fidelity computational models prove too slow to be practical and are thus replaced by much faster surrogate models. Recently, deep learning techniques have become increasingly important in accelerating such predictions. However, they tend to falter when faced with larger and more complex problems. Therefore, this work introduces MAgNET: Multi-channel Aggregation Network, a novel geometric deep learning framework designed to operate on large-dimensional data of arbitrary structure (graph data). MAgNET is built upon the MAg (Multichannel Aggregation) operation, which generalizes the concept of multi-channel local operations in convolutional neural networks to arbitrary non-grid inputs. The MAg layers are interleaved with the proposed novel graph pooling/unpooling operations to form a graph U-Net architecture that is robust and can handle arbitrary complex meshes, efficiently performing supervised learning on large- dimensional graph-structured data. We demonstrate the predictive capabilities of MAgNET for several non-linear finite element simulations and provide open-source datasets and codes to facilitate future research. [less ▲]

maintenance service logistics

in Operations Logistics and Supply Chain Management (2019)

Management of product characteristics uncertainty based on Formal Logic and Characteristics Properties Model

in CIRP Annals - Manufacturing Technology (2013), 62(1),

Uncertainty in product characteristics is ubiquitous in any engineering system at all the stages of product life-cycle. Considering uncertainty from different sources during the product design phase is ... [more ▼]

Uncertainty in product characteristics is ubiquitous in any engineering system at all the stages of product life-cycle. Considering uncertainty from different sources during the product design phase is critical to its reliable performance. This paper presents a framework integrating the uncertainty propagation through different product characteristics and its effect on product properties. The framework consists of three main parts: a descriptive model based on formal logic and characteristics properties model; a mathematical implementation through set theory and probabilistic approach; and an algorithm for design space evaluation and tolerancing. The application of framework is demonstrated through an industrial case study. [less ▲]

A Markov chain dynamic model for trip generation and distribution based on CDR

in Periodica Polytechnica Transportation Engineering (2015)

A Markov Chain Monte Carlo Approach for Estimating Daily Activity Patterns

Poster (2019, January)

Determining the purpose of trips brings is a fundamental information to evaluate travel demand during the day and to predict longer-term impacts on the population’s travel behavior. The concept of tours ... [more ▼]

Determining the purpose of trips brings is a fundamental information to evaluate travel demand during the day and to predict longer-term impacts on the population’s travel behavior. The concept of tours is the most suited to consider the value of a daily scheduling of individuals and travel interdependencies. However, the meticulous care required for both collecting data of high quality and interpret results of advanced demand models are frequently considered as major drawbacks. The objective of this study is to incorporate into a standard trip-based model some inherent concepts of activity-based models in order to enhance the representation of travel behavior. The main focus of this work is to infer, employing utility theory, the trip purpose of a population, at a zonal level. Making use of Markov Chain Monte Carlo, a set of parameters is estimated in order to retrieve tour-based primitives of the demand. The main advantage of this methodology is the low requirements in terms of data, as no individual information are used, and the good interpretation of the model. Estimated parameters of the priors set a utility-based probability function for departure time, which allows to have a dynamic overview of the demand. In order to account for the tour consistency of travel decisions, a duration constraint is added to the model. The proposed model is applied to the region of Luxembourg city and the results show the potential of the methodologies for dividing an observed demand based on the activity at destination. [less ▲]

Mathematical Modelling of Flux Decline due to Concentration Polarisation and Cake Layer Formation in Crossflow Filtration Systems

Scientific Conference (2007, May)

Crossflow membrane filtration is an effective way of removing both colloidal and dissolved organic matter from contaminated water supplies. Two phenomena domimate solute flux in crossflow systems ... [more ▼]

Crossflow membrane filtration is an effective way of removing both colloidal and dissolved organic matter from contaminated water supplies. Two phenomena domimate solute flux in crossflow systems; concentration polarization and cake layer formation. Many innovative mathematical models for predicting both flux decline and quasi-steady state flux have been produced in the literature. However, limited regime applicability and conflicting physical predictions have made choosing optimal performance parameters in design a challenging process. An overview of the current field of models was undertaken, including an assessment of mathematical assumptions, numerical computation workload and number and complexity of system constants. A new model incorporating viscosity dependence on concentration is developed for the concentration polarization regime. Preliminary results will also be presented from a Monte Carlo based molecular dynamics simulation of volume packing fraction in the cake layer regime. The novel aspects of these models will be compared with existing models and the experimental results of our collaborators. Laboratory ultrafiltration tests were undertaken using varying concentrations of Dextran. An outline of future model directions and refinements will be presented. [less ▲]

Mathematical relationships between representations of structure in linear interconnected dynamical systems

in The proceedings of the 2011 American Control Conference (ACC) (2011)

A dynamical system can exhibit structure on multiple levels. Different system representations can capture different elements of a dynamical system's structure. We consider LTI input-output dynamical ... [more ▼]

A dynamical system can exhibit structure on multiple levels. Different system representations can capture different elements of a dynamical system's structure. We consider LTI input-output dynamical systems and present four representations of structure: complete computational structure, subsystem structure, signal structure, and input output sparsity structure. We then explore some of the mathematical relation ships that relate these different representations of structure. In particular, we show that signal and subsystem structure are fundamentally different ways of representing system structure. A signal structure does not always specify a unique subsystem structure nor does subsystem structure always specify a unique signal structure. We illustrate these concepts with a numerical example. [less ▲]

Maximum-Entropy Meshfree Method for the Reissner-Mindlin Plate Problem based on a Stabilised Mixed Weak Form

Scientific Conference (2012)

Meshless methods, such as the Element Free Galerkin (EFG) method, hold various advantages over mesh-based techniques such as robustness in large-deformation problems and high continuity. The Reissner ... [more ▼]

Meshless methods, such as the Element Free Galerkin (EFG) method, hold various advantages over mesh-based techniques such as robustness in large-deformation problems and high continuity. The Reissner-Mindlin plate model is a particularly popular choice for simulating thin structures. It is well known in the Finite Element and Meshless literature that the simplest numerical treatments of the Reissner-Mindlin model lead to shear-locking which in turn produces erroneous results. This is due to the inability of the approximation functions to satisfy the Kirchoff constraint in the thin-plate limit. A recent advance in the area of meshless approximation schemes are Maximum-Entropy (MaxEnt) approximants. MaxEnt schemes provide a weak Kronecker-delta property on convex node sets which allows the direct imposition of Dirichlet (essential) boundary conditions. In this work, we derive a shear-locking free meshless method using MaxEnt approximants by consider- ing a stabilised mixed weak form. We include a scalar parameter which splits the energy from the shear bilinear form into two parts; the first is formed from the displacement fields only and the second from the independently interpolated shear strain field and the displacement fields. This splitting greatly eases the satisfaction of the LBB stability condition. We then eliminate the independently interpolated shear strain field using a localised projection operator, related to the “volume-averaged pressure” technique, which produces a final system of equations in the original displacement unknowns only. We show the good performance of the method for a variety of test problems. [less ▲]