| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 2.72 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
Em condições normais o escoamento dos rios realiza-se no leito principal, no entanto, em situações de cheia, a capacidade de vazão desses leitos é excedida e o escoamento faz-se também nos campos adjacentes, denominados leitos de cheia ou planícies de inundação, re-sultando uma configuração de secção composta. Como na zona do leito principal a altura de escoamento é maior do que nas planícies de inundação, existe uma diferença de velocidades entre elas, que gera interações e transferências de massa e de quantidade de movimento. Nes-tas condições, as características do escoamento não se encontram bem definidas devido ao elevado número de fenómenos associados que para elas contribuem.
A presente dissertação estuda os mecanismos envolvidos nos escoamentos em secção com-posta e estima alturas de água em caso de caudais de cheia fluvial, contribuindo para uma melhoria do mapeamento da inundação para um determinado caudal, ajudando a preparar melhor uma situação de cheia fluvial. Para tal foram realizadas duas análises complementares: avaliação e cálculo da tensão de corte aparente (comparando medições experimentais com va-lores teóricos) e estimativa dos coeficientes de Coriolis e Boussinesq em canais de secção com-posta. Ao calcular estes dois parâmetros com menores erros, a estimativa do caudal escoado por sua vez vai ter valores mais próximos da realidade melhorando o mape-amento num determino local. Nas duas análises realizadas recorreu-se a uma base de dados de escoamentos em canais de secção composta e a redes neuronais. Estas redes são modelos computacionais inspirados no funcionamento do cérebro humano, compostas por camadas de neurônios artificiais que processam informações através de conexões sinápticas ponderadas. Neste sentido, têm vindo a ganhar grande popularidade ao longo dos anos, como é evidenciado pelo aumento crescente do número de artigos sobre estas em revistas científicas.
Under normal circumstances, flow in rivers takes place in the central main channel but, during flood events, the flow capacity of the main channel is surpassed, and the flow inun-dates the adjacent floodplains, resulting in a compound section configuration. As flow depth in the main channel is higher than the one in the floodplains, there is a significant velocity gradient, which leads to interactions and transfers of mass and momen-tum. During these events, the flow characteristics are not well defined due to the high number of flow mechanisms associated to these flows. The present work studies the mechanisms of flow in a compound section and estimates water depths in the event of fluvial flood flows, contributing to the improvement of the inun-dation mapping for a given flow, therefore helping prepare a fluvial flood event. Two complementary analyses were carried out: evaluation and calculation of the appar-ent shear stress (comparing experimental measurements with theoretical values) and estima-tion of the Coriolis and Boussinesq coefficients in channels with compound section. By calcu-lating these two parameters with smaller errors, the estimated flow rate will in turn have val-ues closer to reality, improving the mapping at a given location. In both cases, a database of flows in compound channels and neural networks were used. These neural networks are com-putational models inspired by the functioning of the human brain, composed by layers of artificial neurons that process information through weighted synaptic connections. In this sense, they have been gaining great popularity over the years, as evidenced by the increasing number of articles about them in scientific journals.
Under normal circumstances, flow in rivers takes place in the central main channel but, during flood events, the flow capacity of the main channel is surpassed, and the flow inun-dates the adjacent floodplains, resulting in a compound section configuration. As flow depth in the main channel is higher than the one in the floodplains, there is a significant velocity gradient, which leads to interactions and transfers of mass and momen-tum. During these events, the flow characteristics are not well defined due to the high number of flow mechanisms associated to these flows. The present work studies the mechanisms of flow in a compound section and estimates water depths in the event of fluvial flood flows, contributing to the improvement of the inun-dation mapping for a given flow, therefore helping prepare a fluvial flood event. Two complementary analyses were carried out: evaluation and calculation of the appar-ent shear stress (comparing experimental measurements with theoretical values) and estima-tion of the Coriolis and Boussinesq coefficients in channels with compound section. By calcu-lating these two parameters with smaller errors, the estimated flow rate will in turn have val-ues closer to reality, improving the mapping at a given location. In both cases, a database of flows in compound channels and neural networks were used. These neural networks are com-putational models inspired by the functioning of the human brain, composed by layers of artificial neurons that process information through weighted synaptic connections. In this sense, they have been gaining great popularity over the years, as evidenced by the increasing number of articles about them in scientific journals.
Descrição
Palavras-chave
Canal de secção composta Tensão aparente Redes Neuronais Coeficientes de Coriolis e Boussinesq
