4. Data Processing/Signal Processing

Seismic Deconvolution using Hopfield Neural Network

Amin Roshandel Kahoo, Abdolrahim Javaherian and Babak Nadjar Araabi

University of Tehran, I. R. Iran.

Abstract

Seismic exploration projects include three steps: acquisition, processing, and interpretation. Seismic data processing, one of the most important stages in seismic exploration projects is very time consuming and expensive. Neural networks, according to parallel processing and hardware implementation characteristics, have potential uses that greatly speed up seismic data processing. In this paper, a commonly used neural network, Hopfield neural network, is employed to implement seismic deconvolution which is one of the most important stages involved in seismic processing. In this approach, deconvolution is decomposed into three subprocesses: reflectivity location detection, reflectivity magnitude estimation, and wavelet extraction. In this paper, Hopfield neural network is developed for each of the subprocesses. The obtained networks are combined by Block Component Method (BCM) for simultaneous estimation of reflectivity series and seismic wavelet. Connective weights and external inputs of each network can be determined by modifying cost function of optimization problem with energy function of Hopfield neural network which is the key step for using Hopfield neural network in optimization. For sensitivity examination of neural reflectivity estimator to noise, we implemented this algorithm on traces with variable signal to noise ratios. The value of SNR changed from infinite (trace without noise) to 0.1. This approach is applied to synthetic and real seismic data and results are compared with those of spiking deconvolution. With comparing the results: (1) unlike spiking deconvolution, deconvolution of seismic data using Hopfield neural network is not sensitive to noise and provides much better results than spiking deconvolution for a noisy trace; (2) there is no assumption for randomness reflectivity series; (3) the neural reflectivity estimator is not sensitive to frequency bandwidth of seismic source wavelet. We found that, neural estimator replaces wavelet with spike into the traces, this method increases temporal resolution of seismic section greatly.

Last modified: Wed May 10 17:03:11 2006