Abstract:
In locating subsurface utilities, one known method is a surveying system towed by trailers employing electrical resistivity
tomography (ERT). However, the primary issue with subsurface surveying with towing mechanism is the change in speed caused by
unavoidable obstructions and sloping road surfaces since it affects the sampling logging of the system. With that, this study develops
a novel technique for fast exploration of extensive transects using optimized receiver sampling rate as a function of velocity, current,
power, slope angle, and voltage. Furthermore, regression models such as regression tree (RTree), gaussian process regression (GPR),
support vector machine (SVM), and ensemble regression (ER) were used for model optimization. The Nyquist rate optimization
network (NyRoNet) will be contemplated as the best-performing prediction model. To avoid data deformation in land surveying, the
intended output is a sampling rate that will adapt in slow-down or elevated road conditions. In modeling, the GPR outperforms the
RTree, SVM, and ER based on the RSME, SME, MAE, and R2 values, which were utilized as evaluation metrics in this study. Then,
the MSE values of the different models of GPR, such as the rational quadratic (RQ), square exponential (SE), Matern 5/2, exponential,
and optimized Gaussian process regression, were identified with 1.938e-10, 1.735e-10, 1.663e-10, 3.785e-6
, and 3.254e-10 values,
respectively. With this, Matern 5/2 regression model was considered as NyRoNet. Other evaluation criteria, such as the MAE and R2
,
were also used, demonstrating NyRoNet's efficiency. To further verify the efficiency of NyRoNet, Matplot in MATLAB was utilized
and enabled the sampling rate optimization and normalizing of the resistivity map.
