Gravity anomaly separation based on cellular neural network

Key words:  geophysical prospecting  gravity anomaly  separation  cellular neural network  template  pseduo BP algorithm

Abstract:

The method of separating gravity anomalies using the cellular neural network and how to train the templates based on target local gravity anomalies were discussed. A pseduo BP algorithm was used to train the neural network templates. To ensure the minimum of global error function, weight revision formula was deduced by using the gradient descent algorithm and the derivation of global error function by weight. Molding all kinds of geological conditions, the applicable conditions of cellular neural network method were summarized. The results show that using the cellular neural network method to extract target anomalies is feasible. As long as suitable templates are chosen and target anomaly is prominented, lateral superimposed anomalies can be separated, which shows that the method has strong lateral resolution, and the boundary of local anomaly can be prominented. Based on this, igneous body can be delimited and local mineral resources can be discovered.