Title:
Performance Density Diagram Developed with Machine Learning Models to Optimize Mixture Design of Sustainable UHPC
Author(s):
Tavares
Publication:
Web Session
Volume:
Issue:
Appears on pages(s):
Keywords:
DOI:
Date:
10/17/2021
Abstract:
The compressive strength of Ultra‐High-Performance Concrete (UHPC) is known to be mainly a function of the type, fraction, and quality of the raw materials used for its production. The amount of ingredients required to produce UHPC materials along with the synergistic relationship between them in dictating material performance makes linear regression a very ineffective tool for optimization purposes. Meanwhile, machine learning techniques have been gaining momentum in optimization studies and prediction models. However, the effectiveness of algorithms generated with this technique depend on the size, distribution, and quality of the training data. This study consists in using the Taguchi method to produce a strategic framework for experimental data collection. The experimental data are then used to generate and train a machine learning algorithm that estimates the compressive strength of several combinations of mix proportions within the range of the material contents produced experimentally. The generated matrix of predictors and outcomes is then used to produce a new tool described as Performance Density Diagrams. These diagrams intend to serve as a decision‐making aid to be used during the mix design phase, in which performance, durability and sustainability of different mix options can be evaluated simultaneously. In this study, the performance goal consisted of maximizing the compressive strength of the material, while minimizing the porosity levels aiming to improve durability. These diagrams exhibit regions with different strength levels, which allows one to identify multiple proportioning combinations that achieve the desired strength level while simultaneously maximizing the cement replacement with waste by‐product powders, aiming to reduce CO2 emissions by designing a greener product.