Modeling of time series of Earth crust velocity field in Azarbaijan using multilayer neural network with PSO training algorithm

The coordinates of the stations along with their velocity field and determination of the strain field are the most important parameters in determining the surface deformation of the shell. Preliminary estimation of the Earth's crust velocity field, especially in seismic areas and near faults, can provide valuable information on the geodynamic structure as well as how faults operate. Different solutions can be offered to solve such a problem. Paying attention to the reliability of the solution, its accuracy and efficiency, how to do it and most importantly the discussion of time and cost can be important and fundamental factors in this work. The purpose of this paper is to use modern and accurate methods to estimate and determine the velocity field and displacement field as well as strain tensor parameters in 3D. Artificial neural network (ANN) method with particle mass optimization training (PSO) algorithm for spatial estimation of crustal velocity changes in Iran has been studied. GPS measurements of Central Alborz network stations have been used to evaluate the method.The average relative error calculated in 4 test stations for the permanent base network in the VE component of the velocity field is 13%, in the VN component of the velocity field is 10/10% and in the Vz component of the velocity field is 15.18% of the artificial neural networks. For Central Alborz network, these values have been set as 18.41, 5.45 and 21.20% for VE, VN and Vz components, respectively. The results of this study show the high capability and efficiency of artificial neural network method in spatial estimation of the Earth's crust velocity field in this region.