Machine Learning-based Approach for the Prediction of an Orifice size of Aerospace Vehicle RCS Thrusters during Cold Flow Calibration

Abstract

Artificial intelligence (AI) is expected to drive many advances in the aerospace industry, including cost reduction, reduced design cycle time, modeling, prototyping, optimization, production, and product updating. The efficiency of Reaction Control System (RCS) is one of the crucial parameters of aerospace vehicles and selection of proper orifice size of RCS thruster for a desire flow rate of fuel and oxidizer during cold flow calibration process is important to get better performance of RCS in terms of thrust during a flight test of an aerospace vehicle. In this experimental study, the size of the orifice used in thrusters is estimated by Machine Learning (ML) based mathematical techniques. MATLAB was used for the implementation of the ML techniques. In the present work, Artificial Neural Network (ANN), Support Vector Machines (SVM), and Relevance Vector Machines (RVM) tools are used to predict orifice size for a selected flow rate and pressure drop. The developed ML models such as LSSVM - Linear Kernel, LSSVM - Radial Basis Function kernel, LSSVM - Gaussian kernel, RVM - Gaussian, RVM - Laplace, RVM -Spline, ANN models are adopted to predict the size of the orifice used in rocket engines thrusters. In the present article, 55 data sets are chosen to model the network based on input parameters. For testing the trained model, 10 data sets are used. After comparing the results obtained by these models, it is noted that the LSSVM regression model with Gaussian add kernel performs better for an identified flow rate and pressure drop to forecast orifice size. The LSSVM-Gaussian add kernel has estimated the scale of the minimum mean absolute error orifice for the ML model, i.e. 0.057 relative to other ML models.