METHOD FOR FORMING AN OPTIMIZED NEURAL NETWORK MODULE FOR SIMULATING THE FLOW MODE OF A POLYPHASIC FLUID Vein

摘要

- Method to build a module (hydrodynamic or thermodynamic for example) intended to simulate in real time the flow mode at any point of a pipe, of a multiphase fluid vein comprising at least one liquid phase and at least one phase gas, so that it is best suited to fixed operating conditions relating to a certain number of defined structural and physical parameters relating to the pipe, and to a set of defined physical quantities (hydrodynamic or thermodynamic quantities for example), with ranges of variation fixed for the parameters and physical quantities. - It includes the use of a modeling system based on non-linear neural networks with each of the inputs for structure parameters and physical quantities, and outputs where quantities are available necessary for the estimation of the mode of flow, and at least one intermediate layer. The neural networks are determined iteratively to adjust to the values of a learning base with predefined tables connecting different values obtained for the output data to the corresponding values of the input data. We use a learning base adapted to the imposed operating conditions and we generate optimized neural networks that best adjust to the imposed operating conditions. - Applications to the modeling of hydrocarbon flows in pipes, for example.