Engineered Cementitious Composites Dataset / Strain Hardening Cementitious Composites Data for FDNN ensemble predictive model

Published: 2 March 2021| Version 1 | DOI: 10.17632/584ydjx3xm.1
Mohamedelmujtaba Altayeb,


The data is for the research titled "An Ensemble Method for Predicting the Mechanical Properties of Strain Hardening Cementitious Composites" in CONSTRUCTION AND BUILDING MATERIALS journal. The folder "Models" Contains the models in the project organized as follows: * "Comparison Models" is the folder Containing the models currently available for comparison with our FDNN. * "FDNN Ensemble" Folder contains the neural network models for the FDNN ensemble * "FRST" Folder contains the random forest regressors component of the FDNN ensemble Two ".xlsx" files are provided in this project. * "Full SHCC Dataset.xlsx" is the Raw dataset collected from literature * "sep_dataset.xlsx" is the processed dataset used for training the models Five other files are provided in this project, as: * "Data description.txt" contains a small description of the data * "FDNN Training.ipynb" contains the FDNN training algorithm * "FDNN predictor.ipynb" contains a small demo of how to use the FDNN models to predict the properties of SHCC * "" contains the normalization values that can be used to encode and decode the inputs for prediction purposes * "predict SHCC" is a python implementation of the "FDNN predictor.ipynb" file * "sep_dataset.xlsx" is the processed dataset used for training the models Finally a Description of the parameters is given in the file "Data description.txt" To develop the model, data were collected from a comprehensive literature review of articles. The dataset consists of several parameters related to the mix design and mechanical properties of SHCC. The mix design parameters included in the database are the amount of cement, water, fine aggregate, fly ash, silica fume, cenosphere, blast furnace slag, superplasticiser, coarse aggregate, accelerating admixtures, phase change materials, fibre weight, all represented in percentage weight. The parameters related to fibre properties in the dataset are fibre type, volume (% volume), diameter (Micro-meter), length (mm), tensile strength (MPa), and elastic modulus (GPa). The dataset also included rubber, light weight aggregate, viscosity agent hydroxypropyl methylcellulose, Air entraining admixture, oiling agent, graphene oxide, and defoamer represented in binary forms, i.e. 1 (was used in mix design) or 0 (was not used). Other experimental conditions are included, they are water curing time (days), air curing time (days), high frequency & velocity casting (1 or 0), and temperature (⸰C). Finally, the mechanical properties recorded in the dataset are the tensile stress at the first crack (MPa), tensile strain at the first crack (%), peak tensile stress (MPa), peak tensile strain (%), flexural strength at the first crack (MPa), flexural strain at the first crack (%), peak flexural Strength (MPa), peak flexural strain (%), and Compressive strength (MPa). In total, the resulting data set contains 38 parameters, divided into 29 features and nine targets.


Steps to reproduce

1) Install python 2) Install Jupyter notebook 3) Run the "predict SHCC" using python, or the "FDNN predictor.ipynb" file to make predictions on SHCC mix designs 4) By following the procedures described in the paper the results can be achieved


Composite Material, Artificial Neural Networks, Multivariate Regression, Artificial Intelligence Applications, Concrete Type, Properties of Concrete