Identification on dynamic inverse model for sensor based on genetic neural network

The oil/water two-phase flow is a complicated two-component nonlinear system with time-variance, and the dynamic measuring system for water content in crude oil based on method of dielectric coefficient is affected by manufacturing technology of sensor itself and some non-object parameters, such as temperature and salinity content in oil-water mixture. Consequently, the sensor has serious non-linearity in its input-output characteristics, which is hard to be described by traditional mathematic models up to now. In this paper, a dynamic inverse model and its identification based on genetic neural network (GNN) is proposed for dealing with sensing mechanism under multi-factor influence, making full use of GNN’s advantages of nonlinear approximations with high accuracy, fast global convergence, self-adaptive and self-learning. The simulation result shows this method is effective to realize dynamic nonlinear error correction and eliminate the interference of non-object parameters and nonlinearity of sensor itself on the measurement, improving the nonlinear characteristics of the sensor and measuring accuracy for the dynamic testing system.