Performance prediction of cryogenic/semi-cryogenic rocket engines by using multi-sensor data fusion techniques Mallappa1*, Ramesh S.2, Chandra D. Gilbert2, Nandi T. K.1 1Cryogenic Engineering Centre, IIT KharagpurKharagpur 2ISRO Propulsion Complex (IPRC), ISRO, Mahendragiri *E-mail ID: jabademallappa@gmail.com
Online published on 4 December, 2023. This paper proposes a computational technique for predicting the performance of a liquid rocket engine under varying operating conditions using ten sets of ground test data. Measuring various parameters from multiple location points during a ground test generates a significant amount of data, and the proposed technique aims to reduce the number of ground tests, especially for semi-cryogenic and cryogenic rocket engines. The technique employs multi-sensor data fusion (MSDF) to combine data from multiple sensors and extract unique features that cannot be achieved using the data from a single sensor. The paper provides a tutorial on data fusion applications and process models and their application to rocket engines. The proposed data fusion technique can effectively study the dynamic response of the engine and predict missing data or unknown values. The paper uses a deep neural network, specifically a long short-term memory-Recurrent neural network (LSTM-RNN), to analyze sample data from ISRO and predict the thrust of a rocket engine. The comparison with measured data shows an accuracy of around 97-99 percent. Top Keywords Deep neural networks, Multi-sensor data fusion, Rocket engine performance, Subsystems, LSTM-RNN. Top |