First break refraction event picking using fuzzy logic systems

The authors apply a backpropagation fuzzy logic system (BPFLS) to first break picking. Several features extracted from seismic traces are used as the inputs to a fuzzy logic system to determine whether the peak is a first-break peak. Intertrace continuity is also taken into account. An error backpropagation supervised training scheme is used to determine the full design parameters in the BPFLS. Experimental results indicate that the fuzzy logic system achieves picking accuracy similar to that of a backpropagation neural network (BPNN). However, the BPFLS has a much faster convergence rate in supervised learning than the BPNN. This suggests that a BPFLS should be the first candidate for supervised learning problems, especially when training time is limited