A carregar...
Publicação
Previsão da resistência à compressão de betões de referência com base em medições da condutividade
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 3.35 MB | Adobe PDF |
Orientador(es)
Resumo(s)
A resistência, tanto do betão como do cimento, é um parâmetro essencial na indústria cimenteira. O processo de determinação da resistência envolve a cura de provetes, em condições controladas, durante 28 dias e, sendo este um processo moroso, é crucial o desenvolvimento de métodos que permitam prever, com antecedência, a resistência do cimento e do betão. O CemTest e o ConcTest são equipamentos preditivos que preveem a resistência destes materiais através de medições da condutividade de soluções de cimento e/ou betão. Deste modo, numa fase inicial comparou-se a capacidade preditiva destes equipamentos, quanto à resistência do cimento a 2 e 28 dias. Foram testados cimentos do tipo CEM I 52,5 R e CEM II/A-L 42,5 R da fábrica 1 e 2. Observou-se uma melhoria nos resultados quando foram testados cimentos da fábrica 2, tal que os desvios a 2 e 28 dias diminuíram, em média, 0,5 e 2,9 MPa, respetivamente. De seguida desenvolveu-se e validou-se um método interno que prevê a resistência do betão a 7 e 28 dias por intermédio do equipamento ConcTest. Realizaram-se testes sob betões, com e sem agregados grossos tendo-se concluído que os resultados melhoram quando não existem britas na amostra. Deste modo os desvios observados para a resistência a 7 e 28 dias de betões da fábrica 1 diminuíram, em média, 2,9 e 3,1 MPa. Na fábrica 2 observou-se uma diminuição média de 0,5 e 1,2 MPa para a resistência a 7 e 28 dias. Por fim foram construíram-se dois modelos matemáticos (regressão parcial dos mínimos quadrados e redes neuronais) para prever a resistência do betão a 1, 7 e 28 dias. As redes neuronais apresentaram um desempenho superior, no entanto, o acompanhamento de novas previsões mostrou que se encontravam overfitted, pelo que a complexidade das mesmas foi reduzida. O erro médio absoluto (MAE) para 1, 7 e 28 dias foi: 1,20, 0,69 e 0,63 MPa, para betões CEM I 52,5 R, e 0,90, 0,65 e 0,60 MPa para betões CEM II/A-L 42,5 R.
The strength of cement and concrete is a crucial parameter within the industry. Nevertheless, measuring said resistance is time-consuming because it involves curing the specimens for 28 days in a controlled environment. This way, developing methods that can predict those strengths in a short time is essential. CemTest and ConcTest are two different instruments that predict the compressive strength of these materials by measuring the conductivity of cement and/or concrete solutions. In the early stages of this work, CemTest and ConcTest were compared according to their ability to predict the compressive strength of cement at 2 and 28 days. The tests were performed on two types of cement, CEM I 52,5 R and CEM II/A-L 42,5 R, from both plants 1 and 2. The results showed that the instruments worked better on cement from plant 2, leading to a decrease of 0,5 and 2,9 MPa in the predictive errors for the 2 nd and 28th day strength. Secondly, an internal method to predict the compressive strength of concrete at 7 and 28 days based on ConcTest was developed and validated. The tests were performed on both concrete with, and without coarse aggregates. On average, the predictive errors for the 7th and 28th day strength of plant 1 concrete were 2,9 and 3,1 MPa lower when the tests were performed without aggregates. For pant 2 concrete this errors were 0,5 and 1,2 MPa lower. Finally, a couple of partial least squares regressions and neural networks were developed to predict the compressive strength of concrete at 1, 7, and 28 days. As expected, the neural networks showed better performance, but when fed with a new set of data their performance was lost, indicating signs of overfitting. To reduce this problem, their complexity was lowered. The mean absolute error or MAE at 1,7 and 28 days was: 1,20, 0,69 and 0,63 MPa for CEM I 52,5 R concrete and 0,90, 0,65 and 0,65 for CEM II/A-L 42,5 R concrete.
The strength of cement and concrete is a crucial parameter within the industry. Nevertheless, measuring said resistance is time-consuming because it involves curing the specimens for 28 days in a controlled environment. This way, developing methods that can predict those strengths in a short time is essential. CemTest and ConcTest are two different instruments that predict the compressive strength of these materials by measuring the conductivity of cement and/or concrete solutions. In the early stages of this work, CemTest and ConcTest were compared according to their ability to predict the compressive strength of cement at 2 and 28 days. The tests were performed on two types of cement, CEM I 52,5 R and CEM II/A-L 42,5 R, from both plants 1 and 2. The results showed that the instruments worked better on cement from plant 2, leading to a decrease of 0,5 and 2,9 MPa in the predictive errors for the 2 nd and 28th day strength. Secondly, an internal method to predict the compressive strength of concrete at 7 and 28 days based on ConcTest was developed and validated. The tests were performed on both concrete with, and without coarse aggregates. On average, the predictive errors for the 7th and 28th day strength of plant 1 concrete were 2,9 and 3,1 MPa lower when the tests were performed without aggregates. For pant 2 concrete this errors were 0,5 and 1,2 MPa lower. Finally, a couple of partial least squares regressions and neural networks were developed to predict the compressive strength of concrete at 1, 7, and 28 days. As expected, the neural networks showed better performance, but when fed with a new set of data their performance was lost, indicating signs of overfitting. To reduce this problem, their complexity was lowered. The mean absolute error or MAE at 1,7 and 28 days was: 1,20, 0,69 and 0,63 MPa for CEM I 52,5 R concrete and 0,90, 0,65 and 0,65 for CEM II/A-L 42,5 R concrete.
Descrição
Palavras-chave
cimento betão resistência mecânica à compressão condutividade redes neuronais