Artificial Neural Networks for Predicting Surface Roughness in Milling GFRCs

This project aims to develop and implement an artificial neural network (ANN) to predict the surface roughness of milled glass fiber reinforced composites (GFRCs). The study involves data preprocessing, feature selection, and ANN architecture optimization to achieve accurate surface roughness predictions.

Key Features:
- Data Preprocessing: Imported experimental data and performed initial processing to handle categorical variables and remove outliers. Normalized the data to a range suitable for neural network input.
- Feature Selection: Conducted statistical analysis, including linear correlation coefficients and ANOVA, to identify significant input parameters affecting surface roughness.
- ANN Architecture Design: Experimented with various neural network architectures, adjusting the number of hidden layers and neurons to optimize performance. Implemented a feedforward neural network with Levenberg-Marquardt backpropagation for training.
- Cross-Validation: Employed k-fold cross-validation to evaluate the generalization ability of the ANN model and to avoid overfitting.
- Performance Evaluation: Assessed the model's performance using metrics such as Mean Absolute Error (MAE), Mean Squared Error (MSE), Mean Relative Error (MRE), and Coefficient of Determination (R²).