Search

IntroductionThe use of machine learning (ML) methods would improve the diagnosis of respiratory changes in systemic sclerosis (SSc). This paper evaluates the performance of several ML algorithms associated with the respiratory oscillometry analysis to aid in the diagnostic of respiratory changes in SSc. We also find out the best configuration for this task.MethodsOscillometric and spirometric exams were performed in 82 individuals, including controls (n = 30) and patients with systemic sclerosis with normal (n = 22) and abnormal (n = 30) spirometry. Multiple instance classifiers and different supervised machine learning techniques were investigated, including k-Nearest Neighbors (KNN), Random Forests (RF), AdaBoost with decision trees (ADAB), and Extreme Gradient Boosting (XGB).Results and discussionThe first experiment of this study showed that the best oscillometric parameter (BOP) was dynamic compliance, which provided moderate accuracy (AUC = 0.77) in the scenario control group versus patients with sclerosis and normal spirometry (CGvsPSNS). In the scenario control group versus patients with sclerosis and altered spirometry (CGvsPSAS), the BOP obtained high accuracy (AUC = 0.94). In the second experiment, the ML techniques were used. In CGvsPSNS, KNN achieved the best result (AUC = 0.90), significantly improving the accuracy in comparison with the BOP (p p ConclusionsOscillometric principles combined with machine learning algorithms provide a new method for diagnosing respiratory changes in patients with systemic sclerosis. The present study's findings provide evidence that this combination may help in the early diagnosis of respiratory changes in these patients.

Wind behavior in urban areas is receiving increasing interest from city planners and architects. Computational fluid dynamics (CFD) simulations are often employed to assess wind behavior around buildings. However, the accuracy of CFD simulations is often unknown. Measurements can be used to help understand wind behavior around buildings more accurately. In this paper, a model-based data interpretation framework is presented to integrate information obtained from measurements with simulation results. Multiple model instances are generated from a model class through assigning values to parameters that are not known precisely, including those for inlet wind conditions. The information provided by measurements is used to falsify model instances whose predictions do not match measurements and to estimate the parameter values of the simulation. The information content of measurement data depends on levels of measurement and modeling uncertainties at sensor locations. Modeling uncertainties are those associated with the model class such as effects associated with turbulent fluctuations or thermal processes. The model-based data interpretation framework is applied to the study of the wind behavior around the buildings of the Treelodge@Punggol estate, located in Singapore. The framework incorporates modeling and measurement uncertainties and provides probability-based predictions at unmeasured locations. This paper illustrates the possibility to improve approximations of modeling uncertainties through avoiding falsification of the entire set of model instances. It is concluded that the framework has the potential to infer time-dependent sets of parameter values and to predict time-dependent responses at unmeasured locations. � 2015 Elsevier Ltd.

The advent of new types of loads, such as power electronics and the increased penetration of low-inertia motors in the existing distribution grids alter the dynamic behavior of conventional power systems. Therefore, more accurate dynamic, aggregate, load models are required for the rigorous assessment of the stability limits of modern distribution networks. In this paper, a measurement-based, input/output, aggregate load model is proposed, suitable for dynamic simulations of distribution grids. The new model can simulate complex load dynamics by employing variable-order transfer functions. The minimum required model order is automatically determined through an iterative procedure. The applicability and accuracy of the proposed model are thoroughly evaluated under distinct loading conditions and network topologies using measurements acquired from a laboratory-scale test setup. Furthermore, the performance of the proposed model is compared against other conventional load models, using the mean absolute percentage error.

Submitted by Boris Flegr (boris@uerj.br) on 2021-01-06T19:01:36Z No. of bitstreams: 1 Genasil Francisco dos Santos.pdf: 1208112 bytes, checksum: c78265697b416e0be4e2330a11e982da (MD5) Made available in DSpace on 2021-01-06T19:01:36Z (GMT). No. of bitstreams: 1 Genasil Francisco dos Santos.pdf: 1208112 bytes, checksum: c78265697b416e0be4e2330a11e982da (MD5) Previous issue date: 2009-08-26 Coordenação de Aperfeiçoamento de Pessoal de Nível Superior Structural systems in a variety of applications including aerospace vehicles, automobiles and civil engineering structures such as tall buildings, bridges and offshore platforms, accumulate damage during their service life. In several situations, such damage may not be visually observable. From the standpoint of both safety and performance, it is desirable to monitor the occurrence, location and extent of such damage.System identification methods, which may be classified in a general category of nondestructive evaluation techniques, can be employed for this purpose. Using experimental data, such as eigenmodes, eigenvectors and static displacements, and an analytical structural model, parameters of the structures can be identified. The approach used in the present work is one where the structural properties of the analytical model are varied to minimize the difference between the analytically predicted and empirically measured response. This is an inverse problem where the structural parameters are identified. In this work a reduced number of vibration modes were used as the measured response. For the damage assessment problem a close analytical model of the structural system is available and the model of the damaged structure will be identified. Damage will be represented by a reduction in the elastic stiffness properties of the structure.The problem described above was solved using global methods of optimization due to the fact that depending on the number of variables or the location of damage the resulting design space is nonconvex presenting several local minima. In the present work, the Differential Evolution Optimization Technique (DE) was used. It is a metaheuristic inspired by a population of solutions that is successively updated by arithmetic operations such as mutation and recombination, until convergence. The approach was applied to simple truss structures with different levels of damage. Sistemas estruturais em suas variadas aplicações incluindo-se veículos espaciais, automóveis e estruturas de engenharia civil tais como prédios, pontes e plataformas off-shore , acumulam dano durante suas vidas úteis. Em muitas situações, tal dano pode não ser visualmente observado. Do ponto de vista da segurança e da performance da estrutura, é desejável monitorar esta possível ocorrência, localizá-la e quantificá-la. Métodos de identificação de sistemas, que em geral, são classificados numa categoria de Técnicas de Avaliação Não-Destrutivas, podem ser utilizados para esta finalidade. Usando dados experimentais tais como frequências naturais, modos de vibração e deslocamentos estáticos, e um modelo analítico estrutural, parâmetros da estrutura podem ser identificados. As propriedades estruturais do modelo analítico são modificadas de modo a minimizar a diferença entre os dados obtidos por aquele modelo e a resposta medida. Isto pode ser definido como um problema inverso onde os parâmetros da estrutura são identificados. O problema inverso, descrito acima, foi resolvido usando métodos globais de otimização devido à provável presença de inúmeros mínimos locais e a não convexidade do espaço de projeto. Neste trabalho o método da Evolução Diferencial (Differential Evolution, DE) foi utilizado como ferramenta principal de otimização. Trata-se de uma meta-heurística inspirada numa população de soluções sucessivamente atualizada por operações aritméticas como mutações, recombinações e critérios de seleção dos melhores indivíduos até que um critério de convergência seja alcançado. O método da Evolução Diferencial foi desenvolvido como uma heurística para minimizar funções não diferenciáveis e foi aplicado a estruturas planas de treliças com diferentes níveis de danos.