Nonlinear correction method of pressure sensor based on data fusion

Abstract

Piezoresistive sensors are widely used in the design of underwater pressure transmitters, whose nonlinear output is easily affected by the temperature of oil and gas medium. Most of the published literature focuses on hardware, software, or hybrid correction methods, but they are inflexible, have low accuracy and large temperature drift, especially can not retain the nonlinear characteristics of the sensor itself. This paper presents a nonlinear correction method for pressure sensors based on data fusion. Firstly, the pressure and temperature coupling model is established by using ANSYS software, and the extra thermal stress caused by temperature is analyzed, which makes the pressure sensor have strong nonlinear. Secondly, a nonlinear correction method combining polynomial fitting, BP neural network activated by wavelet function, and BP optimization algorithm based on particle swarm optimization (PF-W_BP-PSO_BP) is proposed. This method combines the advantages of numerical calculation and machine learning, retains the nonlinear characteristics of sensor output and restrains the influence of temperature, and proves that this method is superior to the existing correction methods. Finally, pressure and temperature data acquisition experiments are designed, which verifies the scientific nature and effectiveness of the nonlinear correction method in this paper. The method has been applied to the development project of a pressure and temperature integrated transmitter 1500 m underwater, and its practicability has been verified.