Open AccessSoft computing method for assessment of compressional wave velocity
Author(s) -
Rajesh Singh,
Vikram Vishal,
T. N. Singh
Publication year - 2012
Publication title -
scientia iranica
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.299
H-Index - 51
eISSN - 2345-3605
pISSN - 1026-3098
DOI - 10.1016/j.scient.2012.06.010
Subject(s) - schmidt hammer , compressive strength , soft computing , mean squared error , artificial neural network , geotechnical engineering , correlation coefficient , mean absolute percentage error , particle velocity , geology , mathematics , statistics , computer science , artificial intelligence , mechanics , materials science , physics , composite material
The physico-mechanical properties of rocks and rockmass are decisive for the planning of mining and civil engineering projects. The Schmidt hammer Rebound Number (RN), Slake Durability Index (SDI), Uniaxial Compressive Strength (UCS), Impact Strength Index (ISI) and compressive wave velocity (P-wave velocity) are important and pertinent properties to characterize rock mass, and are widely used in geological, geotechnical, geophysical and petroleum engineering. The Schmidt hammer rebound can be easily obtained on site and is a non-destructive test. The P-wave velocity and isotropic properties of rocks characterize rock responses under varying stress conditions. Many statistics based empirical equations have been proposed for the correlation between RN, SDI, UCS, ISI and P-wave velocity. The Artificial Neural Network (ANN), Fuzzy Inference System (FIS) and neuro-fuzzy system are emerging techniques that have been employed in recent years. So, in the present study, soft computing is applied to predict the P-wave velocity. 85 data sets were used for training the network and 17 data sets for the testing and validation of network rules. The network performance indices correlation coefficient, Mean Absolute Percentage Error (MAPE), Root Mean Square Error (RMSE), and Variance Account For (VAF) are 0.9996, 0.744, 25.06 and 99.97, respectively, which demonstrates the high performance of the predictive capability of the neuro-fuzzy system