Prediction of geotechnical properties of treated fibrous peat by artificial neural networks

摘要

This paper concentrates on measuring the geotechnical properties of cement peat mixed with different dosages of well-graded sand as filler. Several geotechnical tests, namely unconfined compression strength (UCS), California bearing ratio (CBR) and compaction, were performed on the treated fibrous peat samples. The filler was used in a wide range of 50 to 400kg/m(3) of wet peat. In addition, time-dependent changes of geotechnical properties of treated peat were also studied after 14, 28 and 90 days of air curing. Besides, different artificial neural networks trained by a back-propagation algorithm (ANN-BP) and particle swarm optimization method (ANN-PSO) were used to estimate the UCS of stabilized fibrous peat. Results indicate that after a 90-day curing period, the UCS and CBR of treated samples with 300-kg/m(3) cement only, increased by a factor as high as 8.54 and 13.66, respectively, compared to untreated peat. Besides, in the compaction tests, adding filler content to the cement peat increased the maximum dry density (MDD) significantly. In addition, the results of soft computing techniques indicated that the performance indices of the ANN-PSO model was better compared to the ANN-BP model. Finally, sensitivity results showed that the filler content and curing time were the most influential factors on estimating UCS.