A Data-driven Approach for qu Prediction of Laboratory Soil-cement Mixtures

• Autor(a) principal: Tinoco, Joaquim
• Data de Publicação: 2016
• Outros Autores: Correia, António Alberto, Venda, Paulo da, Correia, António Gomes, Lemos, Luís
• Tipo de documento: Artigo
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10316/108860; https://doi.org/10.1016/j.proeng.2016.06.073
• topic: Soil-cement mixtures; Laboratory formulations; Uniaxial compressive strength; Data mining; Neuronal networks; Sensitivity analysis

In this paper a new data-driven approach is proposed for uniaxial compressive strength (qu) prediction of laboratory soil-cement mixtures. The proposed model is able to predict qu over time under different conditions, e.g. different cement contents or soil types, and can be applied at the pre-design stage. This means that the model can be applied previously to the preparation of any laboratory formulation. The designer only needs to collect information about the main geotechnical soil properties (grain size, organic matter content, among other) and select the binder composition to prepare the mixture. Based on a sensitivity analysis, the key model variables were identified and its effect quantified. Thus, it was caught by the model the most relevant variables in qu prediction over time and very high prediction capacity with an overall regression coefficient higher than 0.95.