WCRnet

We propose a deep learning framework to estimate WCR using a cost-effective Frequency Domain Reflectometry (FDR) sensor and a deep model, WCRnet, which leverages residual connections. This repository provides TensorFlow-Keras code for training and testing WCRnet with deep learning on the FDR sensor data, which represents the physical and electrical properties of mortar.

A deep learning framework to estimate water-to-cement ratio in mortar exploiting frequency domain reflectometry sensors

Paper

1. Installation

This implementation requries the follwing dependencies (Tested on Window 10):

• Anaconda3
• Other required libraries will be automatically installed in steps 1-2.

1-1. Clone the WCRnet repository

git clone https://github.com/Hanyang-Robot/WCRnet.git
cd WCRnet/

1-2. Create the conda environment and install the requried packages.

conda env create -f wc_ratio.yaml
conda activate wc_ratio

2. Run the Jupyter Notebook and feel free to use the implementation code!

• If you want to use the WCRnet code, run "deep_learning_WCRnet.ipynb".
• If you want to use the machine learning code, run "machine_learning_[model name].ipynb".

jupyter notebook

Note if you use the WCRnet Framework in your work, please cite the following paper:

@article{yu2025deep,
  title={A deep learning framework to estimate water-to-cement ratio in mortar exploiting frequency domain reflectometry sensors},
  author={Yu, Seunghwan and Park, Homin and Ko, Byungjin and Lee, Han-Seung and Park, Taejoon and Yoon, Jong-Wan},
  journal={Construction and Building Materials},
  volume={462},
  pages={139896},
  year={2025},
  publisher={Elsevier}
}