Depth estimation of gravity anomalies using Artificial Neural Networks

The method of Artificial Neural Networks is used as a suitable tool for intelligent interpretation of gravity data in exploration; in this paper, we have designed a Hopfield Neural Network to estimate the gravity source depth. To calculate the weights and biasing values of the network first the network is designed for the models near to sphere or cylinder and these weights are fixed and the network will rotate so that finally get to its stable state . In this state the energy of the network will be in its minimum value. Thus the network will run for some different initial values of depths and the one which will have the least final energy will finally the depth of gravity source. It is very important to test the designed network we fed the noisy data to it and observed its behavior. This Artificial Neural network was used to estimate the depth of a qanat in north entrance of the Geophysics Institute of Tehran University and the result was very near to the real value of depth.

کلیدواژه ها:

Artificial neural network ، Gravity Exploration ، Depth estimation ، Hopefield