Measurement systems of electromagnetic field for aircraft with the use of a machine learning model

In this article, a new low-cost measurement system for measuring the electric component of electromagnetic field is presented. For the initial calibration of the new measurement system, a reference meter was used, and based on its readings, calibration was carried out using a machine learning model. Initial calibration was carried out in a GTEM 1000 with a Teseq ITS 6006 generator connected. Five models were compared, among which the K-Nearest Neighbors (KNN) model had the highest accuracy. The model was tested on 5 types of aircraft, and its readings were compared with a reference sensor. Test measurements were carried out in five types of aircraft: Cessna C172, Aero AT-3 R100, Tecnam P2006T, PZL M28 Bryza and the Mi-8 helicopter with the developed new measurement system and a reference meter (NHT3DL) with an 01E probe. The new measurement system is small in size and fits anywhere in the aircraft cockpit. To compare the models, the following metrics were used: the coefficient of determination, mean absolute error, mean square error and root mean square error. The Two-sample Kolmogorov-Smirnov tests were used for analysis, and the Bag of Words and Bag of Patterns methods were applied.