An Improved Neural-Network-Based Calibration Method for Aerodynamic Pressure Probes

A new type of evolutionary algorithm, i.e., differential evolution (DE), is proposed and applied to the training of the neural networks, which is used to calibrate a multihole probe in the study and is illustrated and tested by a real case of calibrating a five-hole probe.

Abstract

Calibration of multihole aerodynamic pressure probe is a compulsory and important step in applying this kind of probe. This paper presents a new neural-network-based method for the calibration of such probe. A new type of evolutionary algorithm, i.e., differential evolution (DE), which is known as one of the most promising novel evolutionary algorithms, is proposed and applied to the training of the neural networks, which is then used to calibrate a multihole probe in the study. Based on the measured probe’s calibration data, a set of multilayered feed-forward neural networks is trained with those data by a modified differential evolution algorithm. The aim of the training is to establish the mapping relations between the port pressures of the probe being calibrated and the properties of the measured flow field. The proposed DE method is illustrated and tested by a real case of calibrating a five-hole probe. The results of numerical simulations show that the new method is feasible and effective.