PREDICTION OF GEOMECHANICAL PARAMETER USING ADAPTIVE NEURO FUZZY INFERENCE SYSTEMS (ANFIS) AND MULTIPLE REGRESSION ANALYSIS (MRA) METHOD

• Main Author: ARYANDA (NIM : 22116022), DADANG
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/26337
• Institution: Institut Teknologi Bandung

Evaluation of geomechanical parameters is an important part of every mining project, geology, petrology, and other geotechnical investigations. In practice, predictions are sometimes used in estimating these parameters. In addition to the Multiple Regression Analysis (MRA) method which is often used for prediction, the method that is now developing is ANFIS (adaptive neuro fuzzy inference systems). From the capabilities of MRA and ANFIS, the author tries to predict the uniaxial compressive strength test (UCS), Young's modulus (E), and point load index (PLI) from the physical properties and ultrasonic wave velocity (Vp). Rock samples used are limestone. From three parameters, only UCS and E have predicted because the PLI test results do not show a good correlation. To evaluate the prediction technique and the accuracy of the prediction results of a model, the Root Mean Square Error (RMSE) and Variance Account For (VAF) calculations are used. The MRA and ANFIS methods used in this research show good results and can be used to predict the value of UCS and E. This can also be seen in the calculation of RMSE and VAF values. The RMSE ANFIS value for UCS prediction is 7.73 and 7.54, while E is 1.34 and 1.76. The RMSE MRA value for UCS prediction is 8.65 and 14.99, while E is 1.94 and 3.00. The VAF ANFIS value for UCS prediction is 75.19 % and 90.18 %, while E is 88.26 % and 83.64 %. The VAF MRA value for UCS prediction is 63.54 % and 60.23 %, while E is 73.65 % and 68.41 %. Basically, the prediction is good if the RMSE value approaches 0 and the VAF value approaches 100%. Geomechanical parameters that have a high enough influence on the predicted value of UCS and E are dry density (ρd), porosity (n) and ultrasonic wave velocity (Vp). In addition, the study also made a prediction software for UCS and E values for limestone that can be used by other users.