DEFORMATION ANALYSIS OF GPS AUSCULTATION NETWORK BASED ON GENERALIZED REGRESSION NEURAL NETWORK (GRNN)

B GOURINE

Abstract


The present paper deals with the use of neural networks in the displacement and deformation fields modelling and analysis. The Generalized Regression Neural Network (GRNN) has proved its efficiency and reliability than the classical methods in the approximation of the displacement function. Based on the strain tensors, the deformation of GPS network is evaluated and represented according to a regular grid. In order to analyse this deformation, the concepts of deformability and deformation reliability are introduced, where the Monte Carlo method is employed to compute the significance degrees of the resulting tensors. At each stage of deformation field process, the GRNN neural network is used to perform an optimum interpolation of displacement field. The application concerns the GPS auscultation network of the Liquefied Natural Gas (LNG) underground tank of GL4/Z industrial complex (Arzew, Algeria). Composed of 119 points surrounding the LNG tank, the GPS network was observed between 2000 and 2006. The data concern the horizontal and vertical displacements of the network points according to local geodetic coordinates (E, N, U). The results show the performance of the adopted neural networks method in the generating and analysis of displacement and deformation fields. The most deformations measured are significant and at the deformability level. They support the physical interpretation by the presence of a rocky area at the WS side (landward) of the LNG tank which leads to compression and swelling, and the role of sea which acts as a warmer of freezing front causing important dilatations at the NE side (seaward).  


Keywords


Geodesy, Artificial Neural Networks, Deformability, Deformation Significance, GPS Auscultation Network, Generalized Regression Neural Network (GRNN).

Full Text:

PDF

Refbacks

  • There are currently no refbacks.


Creative Commons License
This work is licensed under a Creative Commons Attribution 3.0 License.