Machine Learning for Antenna Array Failure Analysis Lydia de Lange Under Dr DJ Ludick and Dr TL Grobler Dept. Electrical and Electronic Engineering, Stellenbosch University MML 2019
Outline 15/03/2019 3
Introduction 15/03/2019 4
Antenna Arrays 5 15/03/2019
Reconstructed Sky Im Image 6 15/03/2019
Square Kilometer Array (S (SKA) 7 15/03/2019
Lydia’s Arrays (LA) and Far -Field Patterns 8 15/03/2019
Problem Statement Distorted results Inaccurate far-field (e.g. in Element failure patterns reconstructed sky (beam patterns) image) Important applications: • Array failure correction • System health management of large antenna arrays
Previous work Failed antenna element detection and location possible with machine learning techniques e.g.: • Feedforward neural networks • Support vector models 11 15/03/2019
Methodology 15/03/2019 12
Methodology Sampling Simulate scenarios Train methods for input data FNN
Sampling Methods 𝜒 ( ° ) 𝜄 ( ° ) Name Number of Samples Single cut ( 𝜒 = 0) 0 𝜄 ∈ [−90, 90] 181 Single cut ( 𝜒 = 45) 45 Single cut ( 𝜒 = 90) 90 Single cut ( 𝜒 = 135) 135 Principle cuts 0, 90 362 Diagonal cuts 45, 135 All cuts 724 0, 45, 90, 135 3-D pattern (182 samples) 182 3-D far-field pattern sampled in a ( 𝜄, 𝜒 ) grid. 3-D pattern (361 samples) 361 3-D pattern (725 samples) 725
Training of FNN Multi-label feedforward neural network 𝑦 Sampled far-field observation of 1 failure scenario 𝑧 y = ON or OFF state of each antenna in the array “multi - label” – 1 label for each antenna (25)
Adapt parameters 𝛾 with each pass until f is as Training of FNN similar as possible to true relationship.
Results 15/03/2019 22
Nature of FNN*: • ↑ Number of samples (S) • ↑ Number of parameters ( 𝛾 ) to be estimated • ↓ Accuracy • If accuracy ↑ : sampling pattern found a useful region in the 3-D far-field pattern to accurately identify failure scenarios * # training iterations = const. FNN Results 23 15/03/2019
FNN Results Dataset Samples Training Time Accuracy (%) (sec) 90ᵒ cut 181 31.98 69.70 3-D pattern 182 32.17 87.88 Diagonal cuts 362 35.48 90.91 All cuts 724 40.73 75.76
25 Additional experiments
Additional Experiments • Compared 14 other classification algorithms 1 according to accuracy using the 10 sampling method datasets. • Best 4: • FNN • One vs Rest Classifier + Linear SVC • One vs Rest Classifier + Logistic Regression • One vs Rest Classifier + Logistic Regression CV x 1 Scikit-learn algorithms
Classification Algorithm Comparison 100 Additional 90 Experiments 80 70 ACCURACY (%) 60 • Best: One vs Rest + 50 Logistic Regression CV 40 • 100% accuracy 30 achieved 20 • Number of parameters 10 vs accuracy relationship is different 0 • 3-D sampling method contains more SAMPLING METHOD DATASETS information than combined single cuts FNN OvR+LinearSVC OvR+LogisticRegression OvR+LogisticRegressionCV 27 15/03/2019
Conclusion 15/03/2019 28
Conclusion • FNN used to detect and locate failed antenna elements in a bow-tie antenna array • Investigated choice of training data on FNN accuracy and training time • Diagonal cuts – 90.91% accuracy, 35.48 secs • 3-D pattern (182 samples) – 87.88% accuracy, 32.17 secs • On larger datasets with more scenarios, the difference in training time may become more significant. • Additional work: • Best algorithm: One vs Rest + Logistic Regression CV • Best sampling method: 3-D pattern
Future work • Manufacturing and measuring an antenna array with a spherical nearfield scanner! • Look at SVMs • Looking at other places in pipeline to do ML on: Power Spectral Density and Correlations
Acknowledgement The financial assistance of the South African SKA project (SKA SA) towards this research is hereby acknowledged (www.ska.ac.za). 15/03/2019 31
• [1] R. J. Mailloux, “Array Failure Correction With A Digitally Beamformed Array,” IEEE Trans. Antennas Propag ., vol. 44, no. 12, pp. 1543 – 1550, 1996. • [2] P. Hall, “The Square Kilometre Array: An Engineering Perspective,” Springer , 2010. • [3] J. A. Rodrìguez, et al ., “A Comparison Among References Several Techniques For Finding Defective Elements In Antenna Arrays,” 2nd European Conference on Antennas and Propagation (EUCAP), pp. 1 – 8, 2007. • [4] I. Goodfellow, Y. Bengio, and A. Courville, “Deep Learning,” MIT Press , pp. 164 – 167, 2016. 15/03/2019
Recommend
More recommend
