| dc.description.abstract | Reliability in geotechnical engineering is related to uncertainty. Decreasing uncertainty will affect increasing reliability and low risk. Moreover, evaluating the seismic refraction method is fundamental to reducing uncertainty. Human error in seismic refraction is one of the important issues related to uncertainty in this method. The main research purpose is to evaluate seismic refraction and reduce uncertainty in seismic refraction. Two steps are essential for seismic refraction to determine the boundaries of different layers with different P wave velocity (Vp), picking first arrival data and grouping data into a unified layer with identical wave velocity based on the time-distance curve (T-D curve). This study utilized the forward modeling results of two synthetic geological models. The boundaries of layers (gently dip straight line) and the P wave velocity of each layer was given. The first model is made up of five layers, with velocity increasing with depth (Vp = 700 m/s, 1100 m/s, 1500 m/s, 1800 m/s, and 2200 m/s from top to bottom), while the second model is made up of six layers with a blind layer (Vp = 1000 m/s, 1500 m/s, 1700 m/s, 1400 m/s, 1800 m/s, and 2200 m/s). According to the pick results of seven operators, we discovered high uncertainty (11.2 %) in the first layer for Model 1 and (13.7%) for Model 2. High uncertainty in the boundary layer and low precision of the P wave velocity on Model 2 was also founded on human error in the first arrival picking processing. On the Grouping layer processing, high uncertainty of P wave velocity in the first layer (11.8 %) on the Model 1. However, low accuracy of the P wave velocity value on Model 2. The blind layer amplifies uncertainty and decreases accuracy. Based on different first arrival picking, Seismic refraction tomography will have the narrow zone of the boundary data. Moreover, the uncertainty of seismic refraction tomography is small compared with first arrival picking and grouping layer processing. Thickness data effectively reduces uncertainty (7.1%), and P wave velocity data can reduce uncertainty until (1.7 %) on Model 1 and (3.5 %) on Model 2. Thickness data and velocity data effectively reduce uncertainty in grouping layer processing. However, uncertainty from first arrival picking still exists. The analyzed geological model could have deviated significantly from the true model. | en_US |