Publicação
Application of data mining techniques to jet grouting columns design
| Resumo: | Jet Grouting (JG) is actually a reference method on soil improvement technologies, allowing to improve the strength, stiffness and permeability of soft soils. However, even after several years of practice and notable technology advances, there are still some limitations to overcome. In particular, the main limitation is the absence of efficient approaches for its design. Indeed, the actual design approaches are essentially based on empirically and less accurate methods that are often too conservatives. As a results, the economy and the quality of the treatment can be affected. Therefore, it is fundamental to develop new approaches able to accurately predict JG columns mechanical properties as well as its diameter. However, due to the high number of variables involved in JG process and the heterogeneity of the soils treated, the accomplishment of such complex task represents a major challenge. This challenge relies in the fact that a JG model design should be able to incorporate simultaneously the effect of different variables (e.g. soil and cement slurry properties). So far, the traditional statistical approaches were unable to deal with the complexity of JG data. However, in the past few years powerful tools have emerged for extracting useful information from large and complex data. These tools are currently known as Data Mining (DM) techniques and have been successfully applied in different application domains.. In the present research work, some of the most well known DM algorithms were applied in the prediction of the mechanical properties of JG mixtures as well as JG columns diameter. Therefore, and as a first step, a multiple regression, artificial neural network, support vector machine and functional network algorithms were trained to predict JG laboratory formulations stiffness and uniaxial compressive strength. Moreover, the analytical expressions proposed by Eurocode 2 and CEB-FIP Model Code 1990 for strength and stiffness prediction of concrete were adapted to JG mixtures. After that, the same methodologies were applied in the prediction of strength, stiffness and column diameter of real JG columns. As the main outcomes of this work, high quality predictive models were achieved, as well as a better understanding of the JG mixtures behaviour (given by a global sensitivity analysis). Such results are quite useful for JG design, being expecting an economic and technical improvement through a better optimization of the available resources and efficient design |
|---|---|
| Autores principais: | Tinoco, Joaquim Agostinho Barbosa |
| Assunto: | Soft soils Soil improvement Jet grouting Artificial intelligence Data mining Support vector machines Artificial neural networks Functional networks Sensitivity analysis Solos moles Tratamento de solos Mineração de dados Inteligência artificial Máquina de vetores de suporte Redes neuronais artificiais Redes funcionais Analises de sensibilidade |
| Ano: | 2012 |
| País: | Portugal |
| Tipo de documento: | tese de doutoramento |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade do Minho |
| Idioma: | inglês |
| Origem: | RepositóriUM - Universidade do Minho |
| Resumo: | Jet Grouting (JG) is actually a reference method on soil improvement technologies, allowing to improve the strength, stiffness and permeability of soft soils. However, even after several years of practice and notable technology advances, there are still some limitations to overcome. In particular, the main limitation is the absence of efficient approaches for its design. Indeed, the actual design approaches are essentially based on empirically and less accurate methods that are often too conservatives. As a results, the economy and the quality of the treatment can be affected. Therefore, it is fundamental to develop new approaches able to accurately predict JG columns mechanical properties as well as its diameter. However, due to the high number of variables involved in JG process and the heterogeneity of the soils treated, the accomplishment of such complex task represents a major challenge. This challenge relies in the fact that a JG model design should be able to incorporate simultaneously the effect of different variables (e.g. soil and cement slurry properties). So far, the traditional statistical approaches were unable to deal with the complexity of JG data. However, in the past few years powerful tools have emerged for extracting useful information from large and complex data. These tools are currently known as Data Mining (DM) techniques and have been successfully applied in different application domains.. In the present research work, some of the most well known DM algorithms were applied in the prediction of the mechanical properties of JG mixtures as well as JG columns diameter. Therefore, and as a first step, a multiple regression, artificial neural network, support vector machine and functional network algorithms were trained to predict JG laboratory formulations stiffness and uniaxial compressive strength. Moreover, the analytical expressions proposed by Eurocode 2 and CEB-FIP Model Code 1990 for strength and stiffness prediction of concrete were adapted to JG mixtures. After that, the same methodologies were applied in the prediction of strength, stiffness and column diameter of real JG columns. As the main outcomes of this work, high quality predictive models were achieved, as well as a better understanding of the JG mixtures behaviour (given by a global sensitivity analysis). Such results are quite useful for JG design, being expecting an economic and technical improvement through a better optimization of the available resources and efficient design |
|---|