Predicting Aircraft Monitoring Parameters Using Regression Polynomials with Process Similarity Criteria Fit
The paper presents a rapid predictive analysis for aircraft sensors monitoring data using Regression Polynomials with Process Similarity Criteria Fit (RPPSCF). Two necessary conditions of the method are formulated. It provides a comparison of machine learning methods for predicting aircraft health status. The Long Short-term Memory (LSTM) model is identified as the most efficient for aircraft engine health prediction, according to the analysis of the research papers on aircraft prediction. It is shown that machine learning models that have high prediction efficiency are suitable for long-term aircraft health prognostics and are not applicable for predicting parameters with rapidly changing characteristics. The paper introduces information about rapid machine learning algorithms and their limitations due to low efficiency. Due to the low computational complexity of the RPSCF, the method can be used for the maintenance and supply chain prognostics of the high volume of monitoring data. The method implementation is not limited to aircraft monitoring parameters abnormality detection. The results of the prediction with minimum least square approximation, LSTM, and RPPSCF are illustrated in the charts.The mathematical expression for method implementation for the arbitrary polynomial power is provided. The solution of the non-linear operator equations for defining polynomial coefficients is obtained using the hypernumber theory. Further directions are identified for automated cognition of the similarity criteria and combining the hypernumber method of solving operator equations with Kantorovich’s modified Newton method of solving non-linear equations. The second direction would allow us to increase the computational speed.
