Neural Network Calculator-Concrete Strength
Developed by: Sarthak Aaganja (sarthak.aaganja12@gmail.com)
Civil Engineer , Khwopa College of Engineering
Trained-Range(26.06 to 82.06 MPa)
This project involves developing a neural network model to predict the compressive strength of concrete using a dataset sourced from the Machine Learning Repository of the University of California, Irvine (UCI). The dataset, created by I-Cheng Yeh, provides valuable insights into how different components and factors contribute to the compressive strength of concrete.
Dataset Overview:
- Source: (Validate From here) https://archive.ics.uci.edu/dataset/165/concrete+compressive+strength ( UCI Machine learning Repo )
- Creator: I-Cheng Yeh
- Description: The dataset consists of 1,030 instances with 8 input variables (features) and one target output (compressive strength).
Input Variables:
- Cement (kg/m³) - Continuous
- Blast Furnace Slag (kg/m³) - Integer
- Fly Ash (kg/m³) - Continuous
- Water (kg/m³) - Continuous
- Superplasticizer (kg/m³) - Continuous
- Coarse Aggregate (kg/m³) - Continuous
- Fine Aggregate (kg/m³) - Continuous
- Age (days) - Continuous
All features are measured in kg/m³ except for Age, which is measured in days.
Project Description:
The neural network was trained using a dataset with 25,000 epochs to ensure thorough learning and convergence. The network architecture was robust, featuring over 500 neurons to capture the complexity of the input-output relationships.
The neural network functions by passing the input data through matrix neurons, followed by a ReLU activation function after each layer. This choice of activation function helps introduce non-linearity into the model, enabling it to learn complex patterns in the data effectively.
This project utilizes deep learning to predict concrete compressive strength, focusing on the relationships between various input features and the target outcome. The neural network was developed using the UCI dataset, chosen for its comprehensive real-world data on concrete mixtures. By training with 25,000 epochs and using an architecture of over 500 neurons, the model achieves high accuracy in predicting concrete strength based on its composition and age.
R(Squared)=0.9466