CD Skripsi
Model Korelasi Empiris Sifat Fisik Dengan Permeabilitas Tanah Menggunakan Artificial Neural Network
Permeability coefficient is one of the significant parameters in the calculation of
seepage. Permeability also has a relationship with the physical properties of the
soil. Permeability can be calculated and predicted in various ways. In this era, the
use of artificial intelligence that is able to imitate the human nervous system has
begun to be widely used in the geotechnical field, one of which is using an Artificial
Neural Network which has the advantage of being able to model complex
relationships. In modeling an ANN model, there are several stages which is
training, testing, and simulation. This research will carry out various variations of
the distribution of the amount of data for training and testing to determine which
variation gives the best results. In this study, will conduct a model empirical
correlation of physical properties with soil permeability using ANN with
Backpropagation algorithm. With a total of 80 data, this data will be used in the
testing and training process with variations of 50:50, 60:40, 70:30, and 80:20. The
inputs used in this research are LL, PL, %sand, %fines, %silt, and %clay. In this
study, it was conclude that the variation that produces the best performing network
is the 50:50 variation with the network model using the backpropagation method,
the traingdx training function, the purelin activation function, using 1 hidden layer,
with the number of neurons in the hidden layer as many as 40, learning rate 0.001,
and epochs amount to 1000 and this network obtain R of 0.99509 and MSE of
7.28E-13.
Keywords: Permeability, soil physical properties, Artificial Neural Network,
Backpropagation
Tidak tersedia versi lain