This study applied two feed-forward type computational methods to estimate the Unconfined Compression Strength (UCS) of stabilized peat soil with natural filler and cement. For this purpose, experimental data was obtained via testing of 271 samples at different natural filler and cement mixture dosages. The input parameters for the developed UCS (output) model were: 1) binder dosage, 2) coefficient of compressibility, 3) filler dosage, and 4) curing time. The model estimated the UCS through two types of feed-forward Artificial Neural Network (ANN) models that were trained with Particle Swarm Optimization (ANN-PSO) and Back Propagation (ANN-BP) learning algorithms. As a means to validate the precision of the model two performance indices i.e., coefficient of correlation (R 2 ) and Mean Square Error (MSE) were examined. Sensitivity analyses was also performed to investigate the influence of each input parameters and their contribution on estimating the output. Overall, the results showed that MSE (PSO) < MSE (BP) while R 2 (PSO) > R 2 (BP) ; suggesting that the ANN-PSO model better estimates the UCS compared to ANN-BP. In addition, on the account of sensitivity analysis, it is found that the binder and filler content were the two most influential factors whilst curing period was the least effective factor in predicting UCS.
- fibrous peat
- statistical analysis
- unconfined compressive strength
ASJC Scopus subject areas
- Civil and Structural Engineering