MLP neural network in google colaboratory to predict mechanical properties of manufactured-sand concrete

Munzir Absa (Malikussaleh University, Indonesia)
Tulus Setiawan (Malikussaleh University, Indonesia)
Islami Fatwa (Malikussaleh University, Indonesia)
Amam Taufiq Hidayat (Malikussaleh University, Indonesia)

Jurnal Mantik
Vol. 6 No. 4 (2023): February: Manajemen, Teknologi Informatika dan Komunikasi (Mantik)

The use of manufactured sand instead of natural sand in concrete can help preserve rivers or beaches where these natural sand is sourced from. The mechanical properties of these manufactured-sand concretes are comparable to those produced using natural sand. To predict and help optimize the mechanical properties of concrete made with manufactured sand, a model of ANN was developed using python 3.8 programming language in Google Colaboratory environtment. It was found that ANN model with LBFGS weight optimization algorithm and 50 hidden layer nodes has the best performance, with RMSE = 0.067081. The accuracy of prediction made with this model is calculated to be 90.58% by means of R-squared value and 95.37% by mean absolute error (MAE) value.