Identification of Aircraft Wake Vortex Based on SVM

Abstract

The aircraft wake vortex has important influence on the operation of the airspace utilization ratio. Particularly, the identification of aircraft wake vortex using the pulsed Doppler lidar characteristics provides a new knowledge of wake turbulence separation standards. This paper develops an efficient pattern recognition-based method for identifying the aircraft wake vortex measured with the pulsed Doppler lidar. The proposed method is outlined in two stages. (i) First, a classification model based on support vector machine (SVM) is introduced to extract the radial velocity features in the wind fields by combining the environmental parameters. (ii) Then, grid search and cross-validation based on soft margin SVM with kernel tricks are employed to identify the aircraft wake vortex, using the test dataset. The dataset includes wake vortices of various aircrafts collected at the Chengdu Shuangliu International Airport from Aug 16, 2018, to Oct 10, 2018. The experimental results on dataset show that the proposed method can identify the aircraft wake vortex with only a small loss, which ensures the satisfactory robustness in detection performance.

Recognition of Aircraft Wake Vortex Based on Random Forest

“…In this section, we conducted experiments on our wake vortex data set. To set a general comparison, the experimental results of KNN [11], SVM [12] and RF are considered and the results are summarized in Table 3.…”

“…To tackle these tasks, we have used the k-nearest neighbor (KNN) [11] and the support vector machine (SVM) [12] to solve the problem. In this paper, we provided a new perspective to further understand LiDAR data classification model based on random forests (RF) [13].…”

Characterizing Wake Vortices of Landing Aircraft Using Artificial Neural Networks and LiDAR Measurements

“…Not only have ANNs previously been employed for the evaluation of wake vortices of landing aircraft, they also outperform other -more fundamental -ML techniques, such as Support Vector Machines (SVMs). In a classification task, determining whether wake vortices are present in a LiDAR scan or not, ANNs obtain an accuracy of 94% [25] compared to 70% [26] with the SVMs. While linear regression is not applicable for the complex non-linear flow developing behind aircraft, other ML methods without temporal component (as single LiDAR scans are evaluated corresponding to a two-dimensional cross-section flow problem) such as the above mentioned SVMs or 'decision trees', are often not capable of recognizing complex flow patterns.…”

Conv-Wake: A Lightweight Framework for Aircraft Wake Recognition

“…For this purpose, several methods have been proposed to detect wake in wind fields. Previous publications such as [4][5][6] are usually based on low-level hand-crafted features, recognizing wake based on their symmetry on radar echoes. For the evolving near-Earth wake [7], wake-up detection methods using radial velocity methods based on machine learning techniques may perform poorly and are very sensitive to environmental changes.…”

“…Recently, some aircraft wake vortex recognition methods based on deep learning have been proposed [8][9][10]. Similar to [4][5][6], these deep learning-based recognition methods typically map aircraft wake vortex data obtained from LiDAR into pseudo-color maps with three channels to meet popular deep learning network models and then use classifiers to recognize aircraft wake vortex, which remains low computational performance. In other words, the performances of these existing methods depend on the visualization of LiDAR data and the colormap configuration at different speeds, and information loss resulted by improper color mapping during data-to-image conversion can lead to poor performance in wave vortex recognition.…”