Artificial Neural Network-Based Normal Strength Prediction of Coir Fiber Concrete Hollow Blocks with Rice Husk Ash
Year: 2026
Type: conference-paper
Abstract: This study developed an artificial neural network (ANN) model to predict the normal compressive strength of coir fiber concrete hollow blocks (CHBs) incorporating rice husk ash (RHA) as a partial cement replacement. Twenty mix configurations were prepared using 0-20% RHA replacement and 0-3% coir fiber addition by cement weight, with compressive strength evaluated at 28 days. The experimental results showed moderate but structurally relevant compressive strength values ranging from 5.0 to 6.7 MPa. A feedforward backpropagation ANN was developed using 60 data arrays and optimized through a three-phase selection process involving the training algorithm, activation function, and number of hidden neurons. The best-performing model used the Levenberg-Marquardt training algorithm, the tansig-tansig activation function, and seven hidden neurons, achieving the lowest mean squared error (MSE = 0.305682) and mean absolute percentage error (MAPE = 8.058338%). The Akaike Information Criterion further supported the suitability of the seven-neuron architecture by balancing predictive accuracy and model complexity. Garson's algorithm indicated that cement was the most influential input variable, followed by RHA and coir fiber. The developed ANN model provides a useful decision-support tool for estimating the compressive strength of sustainable CHB mixtures and guiding preliminary mix proportioning.
Language: English
Subfield: Civil and Structural Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7202392049