https://doi.org/10.1140/epjp/i2017-11623-5
Regular Article
Dry unit weight of compacted soils prediction using GMDH-type neural network
1
Department of Civil Engineering, Imam Khomeini International University, Qazvin, Iran
2
Department of Civil Engineering, Gonbadekavous University, Gonbadekavous, Iran
* e-mail: Afshin_geotec@yahoo.com
Received:
23
December
2016
Accepted:
30
June
2017
Published online:
21
August
2017
Dry unit weight ( 
of soils is usually determined by in situ tests, such as rubber balloon, sand cone, nuclear density measurements, etc. The elastic wave method using compressional wave has been broadly used to determine various geotechnical parameters. In the present paper, the polynomial neural network (NN) is used to estimate the of compacted soils indirectly depending on P -wave velocity ( 
, moisture content ( 
and plasticity index (PI as well as fine-grained particles (FC). Eight natural soil samples (88 data) were applied for developing a polynomial representation of model. To determine the performance of the proposed model, a comparison was carried out between the predicted and experimentally measured values. The results show that the developed GMDH-type NN has a great ability 
to predict the of the compacted soils and is more efficient (53% to 73% improvement) than the previous reported methods. Finally, the derived model sensitivity analysis has been performed to evaluate the effect of each input variable on the proposed model output and shows that the P -wave velocity is the most influential parameter on the predicted .
© Società Italiana di Fisica and Springer-Verlag GmbH Germany, 2017
