Showing papers by "INESC-ID published in 2023"

Flood simulation with the RiverCure approach: The open dataset of the Águeda 2016 flood event

Abstract: Abstract. Floods are among the most common natural disasters responsible for severe damages and human losses. Combining numerical modelling with user-friendly tools for geographically referenced data has been adopted to increase preparedness and reduce vulnerabilities. This paper describes the numerical dataset of hydrometric variables that characterize a flood event occurred in February 2016 in the Portuguese Águeda river, shortly defined as Agueda.2016Flood. The dataset was numerically produced and managed through the RiverCure Portal, a collaborative web platform connected to a validated shallow-water model featuring modelled dynamic bed geometries and sediment transport. The dataset Agueda.2016Flood can be used as a starting point to design other experiments and tools, and to learn and apply the proposed approach by directly using the RiverCure Portal. This dataset includes modelled hydrodynamic data (output data) and the topographic, geometrical, land-use and hydrologic data (input data) necessary to carry out the numerical simulation of the flood event.

Design Science in Information Systems and Computing

Abstract: Design science is a term commonly used to refer to the field of study that focuses on the research of artifacts, constructs, and other artificial concepts. Furthermore, the purpose of this article is to provide a definition of this domain of knowledge concerning information systems and computing, as well as to differentiate between what it is and what it is not, as well as to provide examples of these in ongoing research. In order to accomplish this goal, we conduct a bibliometric analysis on design science and its interaction with information systems and computation. This study aims to identify the primary aggregations of publications pertaining to design science and their chronological analyses. In addition, we clarify some common misconceptions about this field of study by defining what constitutes design science and what exactly does not constitute design science. In addition, we determined the primary stages of the methodological approach to design science.

Precision of Test Methods for Hot Mix Asphalt

Abstract: The paper presents the Portuguese experience in the precision evaluation of test methods for hot mix asphalt (AC 14 surf and AC 20 base). The repeatability and reproducibility values were determined based on Proficiency Test Schemes involving several laboratories. The tests were performed from 2007 until 2021. The paper analyses the soluble binder content, maximum density (Method A), bulk density (Procedure B), and Marshall properties (stability and flow). The results were compared with the European standards’ precision data. The analysis confirmed a wide variation of the precision data. Precision was not always constant, but, in some cases, it did seem to be influenced by the characteristics of the bituminous mixture, contrary to the variation presented in the test standards. In general, a lower reproducibility was observed.

Illustrating Algorithmic Design

Abstract: Architectural design is strongly based on visual and spatial reasoning, which is not easy to translate into algorithmic descriptions and, eventually, running programs, making it difficult for architects to use computational approaches, such as Algorithmic Design (AD). One of the most pressing problems is program comprehension. To overcome it, we propose an automatic illustration system for AD programs that produces annotated schemes of the program’s meaning. The illustration system focuses on a basic set of geometric elements used in most calculations to place geometry in space (points, distances, angles, vectors, etc.), and on the way they are manipulated to create more complex geometric entities. The proposed system automatically extracts the information from the AD program and the resulting illustrations can then be integrated into the AD program itself, intertwined with the instructions they intend to explain. This article presents the implementation of this solution using an AD tool to generate the illustrations and a computational notebook to intertwine the program and the illustrations. It discusses the choices made on the system’s implementation, the expected workflow for such a system, and potential future developments.

Polymer-Metal Adhesion of Single-Lap Joints Using Fused Filament Fabrication Process: Aluminium with Carbon Fiber Reinforced Polyamide

Abstract: Additive manufacturing (AM) is often used for prototyping; however, in recent years, there have been several final product applications, namely the development of polymer-metal hybrid (PMH) components that have emerged. In this paper, the objective is to characterize the adhesion of single-lap joints between two different materials: aluminium and a polymer-based material manufactured by fused filament fabrication (FFF). Single-lap joints were fabricated using an aluminium substrate with different surface treatments: sandpaper polishing (SP) and grit blasting (GB). Three filaments for FFF were tested: acrylonitrile butadiene styrene (ABS), polyamide (PA), and polyamide reinforced with short carbon fibers (PA + CF). To characterize the behaviour of these single-lap joints, mechanical tension loading tests were performed. The analysis of the fractured surface of the joints aimed to correlate the adhesion performance of each joint with the occurred failure mode. The obtained results show the impact of surface roughness (0.16 < Ra < 1.65 µm) on the mechanical properties of the PMH joint. The ultimate lap shear strength (ULSS) of PMH single-lap joints produced by FFF (1 < ULSS < 6.6 MPa) agree with the reported values in the literature and increases for substrates with a higher surface roughness, remelting of the primer (PA and PA + CF), and higher stiffness of the polymer-based adherent.

Quantifying Valence and Arousal in Text with Multilingual Pre-trained Transformers

Abstract: The analysis of emotions expressed in text has numerous applications. In contrast to categorical analysis, focused on classifying emotions according to a pre-defined set of common classes, dimensional approaches can offer a more nuanced way to distinguish between different emotions. Still, dimensional methods have been less studied in the literature. Considering a valence-arousal dimensional space, this work assesses the use of pre-trained Transformers to predict these two dimensions on a continuous scale, with input texts from multiple languages and domains. We specifically combined multiple annotated datasets from previous studies, corresponding to either emotional lexica or short text documents, and evaluated models of multiple sizes and trained under different settings. Our results show that model size can have a significant impact on the quality of predictions, and that by fine-tuning a large model we can confidently predict valence and arousal in multiple languages. We make available the code, models, and supporting data.

Probabilistic nonlinear lagged teleconnections of the sea surface temperature field

Abstract: The monthly anomaly sea surface temperature field over the global ocean exhibit probabilistic dependencies between remote points and lagged times, which are explained eventually by some oceanic or atmospheric bridge of information transfer. Despite much of the bivariate SST dependencies appear to be linear, others are characterized by robust and statistically significant nonlinear correlations. In order to enhance that, we present a general method of extracting bivariate (X,Y) dependencies, seeking for pairs of polynomials P(X) and Q(Y) which are maximally correlated. The method relies on a Canonical correlation Analysis (CCA) between sets of standardized monomials of X and Y, up to a certain (low) degree (e.g. 4). Polynomial coefficients are obtained from the leading CCA eigenvector. Polynomials are calibrated and validated over independent periods, being afterwards subjected to marginal Gaussian anamorphoses. The bivariate non-Gaussianity in the space of marginally Gaussianized polynomials remains residual because of the correlation concentration and maximization. Consequently, the bivariate Gaussian pdf or in alternative, a copula pdf in the space of maximally correlated polynomials can accurately approximate the bivariate dependency. That probabilistic model is then used to determine conditional pdfs, cdfs and probabilities of extremes.The method is applied to various (X,Y) pairs. In the first example, X is an optimized polynomial of the El-Ni&#241;o 3.4 index while Y is that index lagged to the future. For lags between 6 and 18 months, the nonlinear El-Ni&#241;o forecast clearly surpasses the linear one, contributing to lower the El-Ni&#241;o seasonal predictability barrier. In the second example, we relate El-Ni&#241;o (X) with the lagged Atlantic multidecadal oscillation index (Y). Nonlinear, robust correlations appear, both for positive and negative lags up to 5 years putting in evidence Pacific-Atlantic basin oceanic teleconnections.The above probabilistic (polynomial based) model appears to be a good candidate tool for the statistical (seasonal up to decadal) forecast of regime probabilities (e.g. dry/wet) and extremes, given certain antecedent precursors.This work was funded by the Portuguese Funda&#231;&#227;o para a Ci&#234;ncia e a Tecnologia (FCT) I.P./MCTES through national funds (PIDDAC) &#8211; UIDB/50019/2020- IDL and the project JPIOCEANS/0001/2019 (ROADMAP: &#8217;The Role of ocean dynamics and Ocean&#8211;Atmosphere interactions in Driving cliMAte variations and future Projections of impact&#8211;relevant extreme events&#8217;). Acknowledgements are also due to the International Meteorological Institute (IMI) at Stockholm University.

Reynolds stresses in gravity currents approaching an obstacle

Abstract: ABSTRACT:&#160;Gravity currents are flows generated by density differences within two contacting fluids.&#160;In this work the interaction between lock-release gravity currents propagating over a horizontal rectangular channel and an emergent cylinder is analyzed through velocity measurements obtained through PIV. Two-dimensional instantaneous velocity fields are measured in a plane perpendicular to the bottom along the center axis of the channel upstream of the obstacle. The experiments were also conducted without the cylinder for comparison purposes and ten repetitions were carried out for each configuration. The analyses focus on the effects that the presence of an adverse pressure gradient has on both the mean velocity field and the turbulence of the leading part of the current, the head, before the impact. The mean velocity field is not affected by the presence of the obstacle and since no differences were found in the spatial distribution of the mean velocity components, the necessary cylinder-induced deceleration occurs uniformly. Turbulence is studied through the components of the Reynolds stress tensor and their fluxes within the head. In the configuration with the cylinder, there are no fluxes of Reynolds stresses in the inner part of the section. Consequently, the Reynolds stress intensity decreases inside the head compared to the configuration without the obstacle. In conclusion, the presence of an adverse pressure gradient stops the mechanism of Reynolds stress distribution from the main source of production, i.e. the front region, to the inner region of the flow. This leads to a decrease in Reynolds stresses in the inner part of the head and an increase in the frontal region.Acknowledgements: This work was partially supported by Foundation for Science and Technology's through funding UIDB/04625/2020 (CERIS research unit).Keywords: Gravity currents, lock release, Particle Image Velocimetry, adverse pressure gradient, Reynolds stress.

Code Vectorization and Sequence of Accesses Strategies for Monolith Microservices Identification

Abstract: Migrating a monolithic application to a microservice architecture can benefit from automated methods that accelerate migration and improve the results of decomposition. One of the current approaches that guide software architects on the migration is to group monolith domain entities into microservices, using the sequences of accesses of the monolith functionalities to the domain entities. In this paper, we enrich the sequence of accesses solution by applying code vectorization to the monolith, using the Code2Vec neural network model. We apply Code2Vec to vectorize the monolith functionalities. We propose two strategies to represent a functionality, one by aggregating its call graph method vectors and the other by extending the sequence of accesses approach with vectorization of the accessed entities. To evaluate these strategies, we compare the proposed strategies with the sequence of accesses strategy and an existing approach that uses class vectorization. We run all these strategies over a large set of codebases and then compare the results of their decompositions in terms of cohesion, coupling, and complexity.