七孔探针的神经网络校准与制造偏差分析

The Neural Network Calibration of Seven-Hole Probe and Its Numerical Analysis of Manufacturing Deviations

作为一种可以同时获得流动速度的大小和方向、以及总压和静压的气动测量装置,七孔探针能够被广泛应用于各种大角度的流动测量。但是它的校准过程周期很长,代价昂贵,影响了探针的推广和批量制造。神经网络算法被用于该探针的校准过程,弥补常规校准方法的不足;在掌握足够多数据点的前提下,一个经过优化的神经网络结构使得对校准精度的提高和进一步改善大角度条件下的探针性能成为可能;同时本文利用CFD数值方法分别模拟了两根探针的不可压绕流流动,实现其数值校准过程,通过对结果的比较和对探针制造过程中产生的典型的制造偏差进行分析,研究了典型制造偏差对校准系数和校准精度的影响。

The seven-hole probe （SHP） is a device which can simultaneously measure the pressure and velocities of the steady flow, especially at large angles. However, the costly and time-consuming calibration impedes its mass production and applications. With the introduction of the artificial neural network into the calibration, the defects existed in the normal method are remedied and the accuracy is superior to those from the well-developed polynomial method. Providing enough amounts of data points, an optimized neural network structure makes it possible to promote the calibration accuracy and the performance of the seven-hole probe under the large angle condition. To gain the insight into the effect of the manufacturing deviations on the calibration accuracy of SHP, the incompressible flow around two probes are computed numerically at the calibrating condition. One is a perfectly symmetrical probe and the other features a typical deviation in shape. The comparison between the two cases indicates that the calibration accuracy is influenced directly by the manufacturing deviations, which can be seen in the change of the standard deviations of the calibration coefficients.